Latest / The Pragmatic Engineer / Design Engineering with Maggie Appleton

Transcript

Design Engineering with Maggie Appleton · The Pragmatic Engineer

0:00
SPEAKER_01: What makes for a great designer or a great design engineer
0:03
SPEAKER_01: and what can us software engineers learn from them?
0:05
SPEAKER_01: Maggie Appleton is one of the most thoughtful design engineers I know.
0:09
SPEAKER_01: She's worked as an illustrator,
0:11
SPEAKER_01: as a designer at an AI startup, and is currently prototyping at GitHub Next.
0:15
SPEAKER_01: Today we talk about what is a design engineer and why there's often a tension
0:20
SPEAKER_01: between engineering and design.
0:22
SPEAKER_01: The tools she's used before and now, and why pen and pencil is not going away even with agents.
0:27
SPEAKER_01: Do you still need a designer when AI can generate 20 high fidelity prototypes?
0:31
SPEAKER_01: And is AI design obvious to Spot?
0:33
SPEAKER_01: And many more.
0:34
SPEAKER_01: If you're an engineer wanting to know where design is important and how
0:37
SPEAKER_01: to get better at design yourself, this episode is for you.
0:41
SPEAKER_01: This episode is presented by TurboPuffer, Vector and Full Text Search built on object storage.
0:46
SPEAKER_01: It's fast, cheap, and extremely scalable.
0:48
SPEAKER_01: Today's episode will be about design and design engineering,
0:51
SPEAKER_01: and so I wanted to share something visually interesting about our season sponsor,
0:55
SPEAKER_01: Antisysys.
0:56
SPEAKER_01: We already know that Antisysys verifies your system correctness
0:59
SPEAKER_01: by running your whole system in hostile simulation and finding bugs.
1:03
SPEAKER_01: Here's a UI for casualty analysis.
1:05
SPEAKER_01: You can open a report for a bug and see the probability of
1:08
SPEAKER_01: a bug occurring throughout the timeline of the simulation.
1:12
SPEAKER_01: In this case,
1:12
SPEAKER_01: we can see that at virtual time 25,
1:15
SPEAKER_01: something happened that makes this bug close to 100% to occur.
1:19
SPEAKER_01: So we can jump to this point in virtual time simulation to read the logs.
1:23
SPEAKER_01: This kind of bug probability visualization is one that I've never seen before.
1:27
SPEAKER_01: There's also this neat log explorer.
1:29
SPEAKER_01: You can filter on error messages and then visualize how common or uncommon
1:33
SPEAKER_01: the error is over time.
1:35
SPEAKER_01: For example, here's looking for failing linearization failures, the purple line.
1:40
SPEAKER_01: And you can understand how rare or common a specific failure was.
1:43
SPEAKER_01: Again,
1:44
SPEAKER_01: I've yet to see this kind of error visualization and I really like the innovation on
1:47
SPEAKER_01: the UI side.
1:48
SPEAKER_01: And finally, the simplified multiverse debugger.
1:52
SPEAKER_01: You can go back in time and replay a debug timeline.
1:55
SPEAKER_01: And you can inject bash commands at any time without affecting the playback of the bug.
1:59
SPEAKER_01: How cool is that?
2:00
SPEAKER_01: For example,
2:01
SPEAKER_01: we're listing files in the current directory,
2:03
SPEAKER_01: but as you can imagine, you can go debug the whole environment so much easier.
2:07
SPEAKER_01: I love how the team at Antistices are pushing what's possible with both debugging
2:11
SPEAKER_01: and verifying software.
2:12
SPEAKER_01: Head to antistices.com slash pragmatic to learn more.
2:15
SPEAKER_01: Maggie, welcome to the podcast.
2:17
SPEAKER_00: Thanks, I'm thrilled to be here.
2:18
SPEAKER_01: I've had...
2:19
SPEAKER_01: So many guests on the podcast.
2:22
SPEAKER_01: No one had quite the introductory story into tech like you.
2:25
SPEAKER_01: How did you get into tech?
2:27
SPEAKER_01: And you came from a very different background, right?
2:29
SPEAKER_00: Yeah.
2:30
SPEAKER_00: I mean, I came in because it was where the money was, I guess.
2:33
SPEAKER_00: I mean, not really.
2:35
SPEAKER_00: But in the sense that like I in university,
2:37
SPEAKER_00: I studied cultural anthropology, which is like my one true love.
2:40
SPEAKER_00: Like I adore cultural anthropology.
2:42
SPEAKER_00: I totally fell in love with it in university.
2:44
SPEAKER_01: What is anthropology?
2:45
SPEAKER_01: Yeah.
2:45
SPEAKER_00: Okay, it's the study of human beings,
2:47
SPEAKER_00: which sounds impossibly broad and like how could that be a discipline?
2:50
SPEAKER_00: But it does it in a particular way where you go and you live with people intensely
2:53
SPEAKER_00: in this thing called participant observation.
2:55
SPEAKER_00: Usually it was done kind of when the field was born of different cultures, of course.
2:59
SPEAKER_00: It was like anthropologists from the West primarily going
3:01
SPEAKER_00: to places like Papua New Guinea or Australia and living among kind
3:05
SPEAKER_00: of traditional peoples and then realizing how different their cultures were.
3:09
SPEAKER_00: Not just like, oh, they eat different food, you know, they have different houses.
3:11
SPEAKER_00: But They have completely different understandings of what color is.
3:15
SPEAKER_00: They have completely different understandings of time.
3:17
SPEAKER_00: It was kind of alongside the birth of psychology,
3:19
SPEAKER_00: like understanding how flexible is the human mind about constructing understandings
3:24
SPEAKER_00: of the world.
3:25
SPEAKER_00: Anthropology is really one of these eye-opening I love it because I grew up
3:28
SPEAKER_00: an expat kid overseas so I think I was exposed early to lots of different cultures
3:31
SPEAKER_00: and so it felt very natural to me to realize like
3:44
SPEAKER_00: Oh, the way things,
3:45
SPEAKER_00: you know,
3:45
SPEAKER_00: people do things at home,
3:46
SPEAKER_00: you know,
3:46
SPEAKER_00: quote unquote home would be London for me,
3:48
SPEAKER_00: but I left at age six, is completely different to the way they do it elsewhere.
3:52
SPEAKER_00: And there's no fixed way for humans to kind of construct a society or live life.
3:57
SPEAKER_00: And the bounds of what we think is possible is much wider than we originally assume,
4:00
SPEAKER_00: which is what I loved about it.
4:02
SPEAKER_00: So I studied it.
4:03
SPEAKER_00: But of course,
4:04
SPEAKER_00: come senior year,
4:05
SPEAKER_00: kind of go, right, what jobs are there available in cultural anthropology?
4:09
SPEAKER_01: I guess living with natives isn't all.
4:12
SPEAKER_00: It's not very lucrative.
4:14
SPEAKER_00: So the options were like, go get a PhD and become a professor.
4:17
SPEAKER_00: Or the military hires lots of anthropologists to come up with torture techniques
4:21
SPEAKER_00: for people in different countries.
4:23
SPEAKER_00: So when presented with these options by our professors,
4:26
SPEAKER_00: we were like, OK, OK, we'll go think about that.
4:29
SPEAKER_00: And I had always loved design growing up.
4:31
SPEAKER_00: So I had been doing like, you know, I was a kid of the 90s.
4:35
SPEAKER_00: I grew up on Neopets with HTML and CSS and MySpace.
4:38
SPEAKER_00: And I learned HTML and CSS probably around And knew how to do it,
4:42
SPEAKER_00: but in the way where there wasn't that much complexity to it in whatever this was,
4:45
SPEAKER_00: 1999.
4:46
SPEAKER_01: Did you have a MySpace?
4:47
SPEAKER_00: I did.
4:48
SPEAKER_00: Oh, oh yeah.
4:49
SPEAKER_01: And you customized it with all the...
4:50
SPEAKER_00: Yeah, all crazy animations and the little sparkle trail at the end of your cursor.
4:54
SPEAKER_00: And I think mine was quite goth at the time.
4:56
SPEAKER_00: But I learned a lot.
4:57
SPEAKER_00: But of course, at the time, the web was so new.
5:00
SPEAKER_00: It wasn't like people thought this was a career or you didn't really think of it that way.
5:03
SPEAKER_00: But I came out of university with this degree that wasn't necessarily employable.
5:06
SPEAKER_00: And I just started doing freelance web design work because that was how I knew how
5:10
SPEAKER_00: to make money.
5:10
SPEAKER_00: Like throughout school,
5:11
SPEAKER_00: I was doing IT tech work for the university and just naturally went into this
5:16
SPEAKER_00: because it was like, well, I need to pay rent somehow.
5:19
SPEAKER_00: And I gravitated towards illustration originally because I love drawing.
5:22
SPEAKER_00: So I was originally an illustrator for the first couple of years.
5:25
SPEAKER_00: And the way that I went more into the tech side,
5:27
SPEAKER_00: because you can kind of be an illustrator that's like,
5:29
SPEAKER_00: you know, editorial or you'd go into branding or like all kinds of types.
5:33
SPEAKER_00: But I started working for a startup.
5:36
SPEAKER_00: I guess I wouldn't call it a startup.
5:37
SPEAKER_00: It was more like a design dev shop in Prague that was making UI,
5:40
SPEAKER_00: UX apps for like SF startups at the time.
5:43
SPEAKER_00: They were doing work for like Tinder and Uber, like in the early days, like in their beginnings.
5:47
SPEAKER_00: So it was there I got exposed to UX design and I was making the illustrations
5:51
SPEAKER_00: for these apps and doing logos and branding.
5:53
SPEAKER_00: But I like that was my first exposure to like,
5:55
SPEAKER_00: oh, there are people who design buttons and sidebars.
5:58
SPEAKER_00: And that's kind of interesting.
5:59
SPEAKER_00: I didn't end up going into that for a while,
6:01
SPEAKER_00: but that was my first introduction to like what is product design in tech as
6:05
SPEAKER_00: a field.
6:07
SPEAKER_00: And eventually I joined this company called Egghead, which does developer education.
6:10
SPEAKER_00: So they taught JavaScript.
6:13
SPEAKER_00: Sure, you've seen them around.
6:14
SPEAKER_00: And I was their illustrator for four years.
6:16
SPEAKER_00: And then I became an art director and like art directed other people's illustrations for them.
6:21
SPEAKER_00: And that was really where I learned JavaScript, React.
6:23
SPEAKER_00: Like that was,
6:24
SPEAKER_00: I think my real front end engineering education was like doing illustrations
6:28
SPEAKER_00: for them.
6:28
SPEAKER_00: But in doing the illustrations,
6:29
SPEAKER_00: I had to understand the material I was illustrating,
6:32
SPEAKER_00: which turned out to be like React components and use effect
6:35
SPEAKER_00: and like JavaScript functions.
6:37
SPEAKER_00: And in a weird way,
6:38
SPEAKER_00: I just ended up moving more into front-end engineering and visual design
6:42
SPEAKER_00: because it just felt like a natural move.
6:43
SPEAKER_01: You worked at Elicit as well, right?
6:46
SPEAKER_01: Yeah.
6:46
SPEAKER_01: Was that before Eckhead?
6:48
SPEAKER_00: After.
6:48
SPEAKER_01: That was after.
6:49
SPEAKER_01: Yeah.
6:49
SPEAKER_01: So you were at Eckhead as their learned design there.
6:53
SPEAKER_01: And then Elicit was an AI startup, right?
6:57
SPEAKER_00: Yeah.
6:57
SPEAKER_00: Early.
6:58
SPEAKER_00: Pretty early.
6:58
SPEAKER_01: Well, okay.
6:59
SPEAKER_01: Pretty early.
6:59
SPEAKER_01: Yeah.
7:00
SPEAKER_00: I mean,
7:01
SPEAKER_00: it was like the founders there were really kind of incredible people and that one
7:04
SPEAKER_00: of them had been studying language models for 10 years.
7:07
SPEAKER_00: He had like seen the writing on the wall way before anyone else,
7:10
SPEAKER_00: but like kind of out of MIT,
7:11
SPEAKER_00: like PhD kind of like,
7:13
SPEAKER_00: you know,
7:13
SPEAKER_00: machine learning stuff and kind of had this realization
7:16
SPEAKER_00: of like language models are going to revolutionize science.
7:18
SPEAKER_00: He was really big on like, how could this speed up the scientific process?
7:22
SPEAKER_00: So Elicit originally was,
7:25
SPEAKER_00: and it still is actually,
7:25
SPEAKER_00: I mean they've expanded,
7:27
SPEAKER_00: but at its core it is using language models to speed up
7:29
SPEAKER_00: the scientific literature review process.
7:32
SPEAKER_00: It's usually a very manual slow thing of all these academics reading thousands
7:35
SPEAKER_00: of papers and extracting data about them into spreadsheets,
7:37
SPEAKER_00: perfect for models, like a perfect use case.
7:40
SPEAKER_00: So I joined them, I think it was late 2021, early 2022.
7:43
SPEAKER_01: And you were the only designer, there was no product manager, you were the founding designer?
7:47
SPEAKER_00: Yeah, yeah.
7:48
SPEAKER_00: And there were six of us in the beginning when I joined,
7:51
SPEAKER_00: maybe seven, I might have been the seventh, I forget.
7:54
SPEAKER_00: A very small team.
7:54
SPEAKER_00: And it stayed small most of the time I was there.
7:57
SPEAKER_00: And yeah, I was the only designer the whole time.
7:59
SPEAKER_00: And then we had a PM at some point who left,
8:01
SPEAKER_00: and then we didn't get another one, so it's no PM required.
8:03
SPEAKER_00: And it was a really wonderful experience because it was a classic early startup.
8:06
SPEAKER_00: But like it was like we were family.
8:08
SPEAKER_00: We would like stay together for weeks at a time during retreats.
8:10
SPEAKER_00: Like the founders had really strong conviction and they're really smart people.
8:14
SPEAKER_00: And I trusted them so much.
8:15
SPEAKER_00: So it was sort of like get on board with the vision kind of deal.
8:18
SPEAKER_00: And it was there I really learned product design end to end,
8:20
SPEAKER_00: I'd say,
8:21
SPEAKER_00: in the terms of we had lots of users because we had a free prototype
8:24
SPEAKER_00: that was like very, very popular.
8:25
SPEAKER_00: So you had all this data you could collect about what people were searching for
8:28
SPEAKER_00: and clicking on and what worked and what didn't and A-B tests.
8:31
SPEAKER_00: And And we just went really fast.
8:33
SPEAKER_00: Like we shipped a feature every week for a year.
8:35
SPEAKER_00: So it was like design, build, like ship it, measure, design, build, ship it, measure on repeat.
8:42
SPEAKER_00: Over and over for like the solid,
8:44
SPEAKER_00: I think I was there a little over two years and I'd learned a ton in that time just
8:47
SPEAKER_00: about the core mechanics of product design in the sense of like really going
8:51
SPEAKER_00: from what are the user needs,
8:52
SPEAKER_00: what's their domain through to doing all the front end engineering
8:55
SPEAKER_00: and like tying it all together.
8:57
SPEAKER_00: So it was a really wonderful experience.
8:58
SPEAKER_00: I learned a lot.
8:59
SPEAKER_00: It was intense.
8:59
SPEAKER_00: By the end of it, I was like, step back from startups for a minute.
9:02
SPEAKER_01: So you got into design,
9:05
SPEAKER_01: I guess, kind of more like self-taught, like figuring out there's a need for this.
9:09
SPEAKER_01: You worked at Eckhead where you learned to design or explain developer concept,
9:13
SPEAKER_01: educational concept for developers.
9:15
SPEAKER_01: You worked at an AI startup as a designer.
9:18
SPEAKER_01: What does a designer do?
9:20
SPEAKER_00: It's kind of a simple answer.
9:21
SPEAKER_01: Now that you've actually done several roles,
9:24
SPEAKER_01: and I get a sense that,
9:25
SPEAKER_01: of course,
9:26
SPEAKER_01: there's going to be,
9:27
SPEAKER_01: it depends,
9:29
SPEAKER_01: but at startups that you've worked with, that you know, how would you describe?
9:35
SPEAKER_01: And of course, many of us developers have worked with designers, some have not at all.
9:40
SPEAKER_00: I mean, I kind of describe it, it's not that different to engineering.
9:42
SPEAKER_00: It's problem solving.
9:44
SPEAKER_00: But the materials are different is the way I would describe it.
9:47
SPEAKER_00: It's like you go through the same process of like you have defining what your problem is.
9:50
SPEAKER_00: Like, are you sure this is the correct problem?
9:52
SPEAKER_00: Have you defined it well and scoped it well?
9:54
SPEAKER_00: You know, researching possible solutions, doing wide exploration.
9:57
SPEAKER_00: What are all the ways we could solve this?
9:58
SPEAKER_00: What are the tradeoffs of them?
10:00
SPEAKER_00: You know, prototyping solutions, validating those are the right solutions.
10:03
SPEAKER_00: Do they work for users?
10:04
SPEAKER_00: Do they make sense?
10:05
SPEAKER_00: I don't think it's that different to engineering in the sense that I do some of both
10:09
SPEAKER_00: in my job, but it's just the material is different.
10:11
SPEAKER_00: Instead of working with code,
10:13
SPEAKER_00: although nowadays with developers,
10:15
SPEAKER_00: you're working in higher level kind of architecture data flow.
10:18
SPEAKER_00: You're not necessarily writing the syntax, but at the time.
10:20
SPEAKER_00: What syntax are you going to use to write this problem?
10:23
SPEAKER_00: And with design,
10:23
SPEAKER_00: the materials are like space, size, weight, color,
10:27
SPEAKER_00: you know,
10:28
SPEAKER_00: things you would see in an interface in terms of the visuals,
10:30
SPEAKER_00: motion, you know, prominence.
10:32
SPEAKER_00: Is this big enough?
10:33
SPEAKER_00: Are these the right words for the user to understand what this button is going to do?
10:37
SPEAKER_00: Like everything from copywriting to visual graphic design.
10:40
SPEAKER_00: But with product design,
10:41
SPEAKER_00: you're also dealing with what we would call nouns and verbs of a product.
10:44
SPEAKER_00: So it's easy when your product is like a sneaker store.
10:48
SPEAKER_00: It's like the nouns are like sneaker, cart, money.
10:52
SPEAKER_00: If you're designing AWS, the nouns get extremely difficult.
10:56
SPEAKER_00: And I've primarily worked in...
10:58
SPEAKER_00: What I'd call like power user tools,
10:59
SPEAKER_00: like scientists,
11:00
SPEAKER_00: developers, where the nouns are extremely hard because they get very abstract.
11:04
SPEAKER_00: It's sort of like, what's the right container for a set of data?
11:06
SPEAKER_00: What's the right container or noun to point you to like, oh,
11:09
SPEAKER_00: this is your whatever set of data or like this set of data connects to that set
11:13
SPEAKER_00: of data or like here's a function that transforms data into another set.
11:16
SPEAKER_00: You need a noun and verbs to give to users so they can understand how
11:20
SPEAKER_00: to manipulate whatever you're trying to get them to do.
11:23
SPEAKER_00: It's really difficult in dev tools sometimes because there's so much malleability in
11:26
SPEAKER_00: a way that What is
11:28
SPEAKER_01: malleability?
11:31
SPEAKER_00: It could take many different forms and shapes.
11:34
SPEAKER_00: Versus if you're designing things for the real world,
11:35
SPEAKER_00: like I have friends who work for government design,
11:38
SPEAKER_00: there's restrictions there where it's like you're trying to get someone to fill out
11:41
SPEAKER_00: a form.
11:42
SPEAKER_00: It can be a hard design challenge, but it's not complex in the same way.
11:46
SPEAKER_00: Versus like at the moment, right, I'm of course trying to design like agentic tools.
11:50
SPEAKER_00: And it's like, what are the nouns of agents?
11:52
SPEAKER_00: Like we don't know.
11:53
SPEAKER_00: There's like chat sessions.
11:54
SPEAKER_00: We have these things called plans.
11:55
SPEAKER_00: There's something called an MCP.
11:56
SPEAKER_00: There's skills.
11:57
SPEAKER_00: Like could we make new primitives that like connect a bunch of sessions all the way
12:01
SPEAKER_00: to a PR that becomes a new noun that contains it?
12:03
SPEAKER_00: Like there's all kinds of boundaries you could draw that would make the user think
12:08
SPEAKER_00: about your experience differently.
12:09
SPEAKER_00: And then you have to define what verbs can they take on which nouns, right?
12:12
SPEAKER_00: Can you Sure, edit, rename, delete.
12:14
SPEAKER_00: This is like classic CRUD stuff you have to figure out.
12:16
SPEAKER_00: But I don't know.
12:17
SPEAKER_00: Can you like fork an agent session?
12:19
SPEAKER_00: What are the implications of that?
12:20
SPEAKER_00: Like, I think the hard bit of product design is designing a coherent system
12:25
SPEAKER_00: that takes all this complexity that could exist,
12:27
SPEAKER_00: especially in something like DevTools,
12:29
SPEAKER_00: and reducing it to the simplest possible form you can,
12:32
SPEAKER_00: which is a very easy to say, extremely hard to do every time.
12:36
SPEAKER_00: To like a really canonical set of nouns that the user can go, okay, I can point at that.
12:41
SPEAKER_00: I understand what that's going to do.
12:42
SPEAKER_00: I can, when I click a button, I don't get surprised at the outcome.
12:45
SPEAKER_00: This takes time to do this kind of like hard, reduce it to its best, most elegant form.
12:50
SPEAKER_01: This is so interesting because what you've described of like,
12:53
SPEAKER_01: you know,
12:54
SPEAKER_01: we have a new product that has not existed before,
12:56
SPEAKER_01: like a deaf tool in a digital space,
12:59
SPEAKER_01: which lives in our head or inside of a computer.
13:01
SPEAKER_01: I just see that you're kind of drawing up a map in your head,
13:04
SPEAKER_01: which is not all that different to when you're building a new system
13:07
SPEAKER_01: a software engineer.
13:21
SPEAKER_01: Ken Beck,
13:22
SPEAKER_01: who talked about how with Ward Cunningham,
13:26
SPEAKER_01: they came up with some of the basics of that might have been domain driven design
13:31
SPEAKER_01: or it might have been some related concepts,
13:35
SPEAKER_01: but they had a thesaurus in front of them trying to,
13:38
SPEAKER_01: they searched for the right words on how to, oh, it was the design patterns.
13:43
SPEAKER_01: Design patterns later came out of it,
13:44
SPEAKER_01: but they were trying to put a name on these constructs and these things and how
13:49
SPEAKER_01: Which feels like a very similar thing to what you're describing.
13:54
SPEAKER_00: Yeah, yeah, yeah.
13:55
SPEAKER_00: That's design.
13:55
SPEAKER_00: This is where I kind of get into, sure, there's types of design.
13:58
SPEAKER_00: Like you could be like, there's interior design, there's brand design, there's product design.
14:02
SPEAKER_00: But product design, when we talk about it in tech, is actually software design.
14:05
SPEAKER_00: And software design is engineering.
14:06
SPEAKER_00: It's actually the same skill.
14:08
SPEAKER_00: Like you might be working with slightly different materials in the sense of one
14:10
SPEAKER_00: of you cares more about The colors and the size and the shadows and the shape
14:14
SPEAKER_00: of things and the motion design.
14:16
SPEAKER_00: But that person also has to understand, what's the shape of the database?
14:20
SPEAKER_00: What are these APIs?
14:22
SPEAKER_00: Is the data in the right place?
14:23
SPEAKER_00: Are we passing data through this component in the right way?
14:26
SPEAKER_00: They have to also understand.
14:28
SPEAKER_01: This is interesting because designers I worked with years back at places like Uber
14:32
SPEAKER_01: and at Skype.
14:33
SPEAKER_01: They had their design tools, which was Sketch, later Figma.
14:38
SPEAKER_01: They typically work with the product folks.
14:40
SPEAKER_01: They did explorations or UX prototypes.
14:44
SPEAKER_01: They often sat into user testing.
14:46
SPEAKER_01: But in the end, they had a design.
14:48
SPEAKER_01: They had visuals.
14:49
SPEAKER_01: They had animations.
14:51
SPEAKER_01: That was their thing.
14:52
SPEAKER_01: And they handed this over to us engineers together with the PRD or here's the product.
14:56
SPEAKER_01: Here's how it's going to look.
14:58
SPEAKER_01: And then we built it and we would build the UI,
15:00
SPEAKER_01: let's say on mobile,
15:01
SPEAKER_01: and we would then go maybe sit with them or show them and they would say,
15:05
SPEAKER_01: oh, this motion doesn't feel good.
15:06
SPEAKER_01: But my view of the designer is like, well, it was very visual.
15:10
SPEAKER_01: And for example,
15:11
SPEAKER_01: those designers,
15:12
SPEAKER_01: maybe they didn't need to,
15:14
SPEAKER_01: but they didn't need to get involved in like how the database works on how
15:17
SPEAKER_01: the flows.
15:17
SPEAKER_01: They were very much aware of the flows of the user journey.
15:20
SPEAKER_00: Yeah.
15:21
SPEAKER_01: So like is that type of design,
15:24
SPEAKER_01: how would you characterize as being different or is it just that their domain is
15:29
SPEAKER_01: a little bit different?
15:30
SPEAKER_01: It's more of the business level that the more kind of mobile,
15:33
SPEAKER_01: these were like mobile and web and some of those.
15:35
SPEAKER_00: Yeah, I think it gets into kind of like how we label different parts of design.
15:38
SPEAKER_00: Like the kind of design I'm talking about is obviously biased by my experience,
15:42
SPEAKER_00: which is much more design engineering stuff because it is like
15:45
SPEAKER_00: when I hear design engineer, that's kind of a catchphrase now.
15:47
SPEAKER_00: And what does that mean?
15:48
SPEAKER_00: But I think of it as a designer who really engages in the engineering
15:51
SPEAKER_00: and understands how the product works.
15:53
SPEAKER_00: And it's required.
15:54
SPEAKER_00: It's not required in all domains.
15:55
SPEAKER_00: Again, if you're doing government forms,
15:56
SPEAKER_00: I don't think you need to understand the database properly.
15:59
SPEAKER_00: But if you're designing for developers or you're designing in
16:01
SPEAKER_00: a brand new field like AI where so much of how the product works is determined
16:07
SPEAKER_00: by what is possible on the back end and the shape of the data,
16:11
SPEAKER_00: then you do need to engage with it a lot.
16:13
SPEAKER_00: But if you're working on an app where actually a designer is probably better served
16:17
SPEAKER_00: by advocating for the user, I think that's a kind of more traditional approach.
16:21
SPEAKER_00: Philosophy of product design is that you represent the user and you represent trying
16:25
SPEAKER_00: to get the best experience possible for the user.
16:27
SPEAKER_00: And that means user interviews, caring about flows, like does this button feel big enough?
16:32
SPEAKER_00: Does it have the right words?
16:34
SPEAKER_00: And then they can spend much more energy there if they don't have to care about
16:37
SPEAKER_00: the technical backend.
16:38
SPEAKER_00: Like maybe it's actually irrelevant when really what you need to care about is like,
16:41
SPEAKER_00: does this flow make sense to people?
16:42
SPEAKER_00: So that's not necessarily a different kind of designer,
16:45
SPEAKER_00: but if you kind of think of the whole stack, right,
16:47
SPEAKER_00: but expand it all the way out,
16:48
SPEAKER_00: not just backend and frontend,
16:49
SPEAKER_00: but The interface and users and then the product in the context of a business
16:54
SPEAKER_00: and like a product in the context of an economy,
16:56
SPEAKER_00: like there are people who lean way more on this side.
16:58
SPEAKER_00: And I'm a bit more the straddling of the product design and the engineering,
17:02
SPEAKER_00: but you can have valid designers all along this.
17:04
SPEAKER_00: And some people, again, you know, just visuals or just animation, like there's lots of niches.
17:09
SPEAKER_00: I just like a bit more of the half engineering, half design, like slice of it.
17:14
SPEAKER_01: To help understand a bit more of what you do, can you talk about the tools that you use?
17:18
SPEAKER_01: Like, you know,
17:19
SPEAKER_01: as engineers,
17:19
SPEAKER_01: we're kind of used to,
17:20
SPEAKER_01: our tools used to be the ID and the code editor and,
17:23
SPEAKER_01: you know, some of the hardcore people like Vim and some of those things.
17:26
SPEAKER_01: And these days, of course, it's changing a bit more.
17:27
SPEAKER_01: But, you know, those are the tools that you use.
17:29
SPEAKER_01: What are your tools?
17:30
SPEAKER_00: It changes all the time.
17:31
SPEAKER_00: I mean, to some degree,
17:32
SPEAKER_00: like some things stay constant, like you're and you're doing similar tasks.
17:36
SPEAKER_00: But of course,
17:37
SPEAKER_00: at the moment,
17:37
SPEAKER_00: I'm trying to try everything because part of my job is like trying
17:40
SPEAKER_00: to Figure out what GitHub should do next.
17:43
SPEAKER_00: That's what our team does.
17:44
SPEAKER_01: That's actually the name, right?
17:45
SPEAKER_00: GitHub Next.
17:45
SPEAKER_00: GitHub Next, hey?
17:46
SPEAKER_00: It was well-named.
17:48
SPEAKER_00: So a lot of the time,
17:49
SPEAKER_00: of course,
17:50
SPEAKER_00: I'm trying out Codex and Claude and dogfooding stuff internally at GitHub.
17:55
SPEAKER_00: I've tried Conductor.
17:56
SPEAKER_00: I mean, name any agentic harness, like OpenCode, Pi.
18:00
SPEAKER_00: I try them all.
18:01
SPEAKER_00: I do love Codex at the moment.
18:03
SPEAKER_00: I do think OpenAI is on to some really good stuff, at least in there.
18:07
SPEAKER_00: Their design of their desktop app is beautiful.
18:10
SPEAKER_00: They've really thought it through.
18:11
SPEAKER_00: So I think they're doing some really good stuff.
18:12
SPEAKER_00: I still use paper and pen for initial brainstorming.
18:15
SPEAKER_01: You have your notebooks here.
18:17
SPEAKER_00: Yeah, I brought some along because notebooks aren't dead.
18:19
SPEAKER_00: I don't think they're going to be dead forever.
18:22
SPEAKER_00: And I still use Figma because I know how to use it and I can do some brainstorming
18:26
SPEAKER_00: in it before I pass it off to an agent.
18:28
SPEAKER_00: But to be honest,
18:29
SPEAKER_00: there's a point where once you've figured out what you need to build,
18:32
SPEAKER_00: which is actually all the work,
18:34
SPEAKER_00: once you hand it off to an agent,
18:36
SPEAKER_00: It's not that implementation is solved,
18:37
SPEAKER_00: but we've reached a point where implementation is like good enough that
18:41
SPEAKER_00: if I spec it out really well and I list out how the agent should verify for me
18:45
SPEAKER_00: that it actually did the work,
18:47
SPEAKER_00: I can hand it to an agent and just like be like,
18:49
SPEAKER_00: right, let me know when you've got a PR up.
18:51
SPEAKER_00: You know, I don't really look at code that much anymore.
18:53
SPEAKER_00: I do look at PR code when I'm reviewing it,
18:56
SPEAKER_00: but skim, skim, skim, you know, okay, that looks sensible.
19:00
SPEAKER_00: Merge.
19:03
SPEAKER_00: So most of the work I find is everything leading up to telling the agent
19:06
SPEAKER_00: to implement something.
19:07
SPEAKER_00: And then the PR review is kind of a separate piece of work.
19:09
SPEAKER_00: But like the bulk of my tools deal in this first section now versus we used to have
19:14
SPEAKER_00: the IDE,
19:14
SPEAKER_00: you know, once it was like, OK, you've decided this is the right thing to design.
19:17
SPEAKER_00: Now you actually start the work of like implementing it.
19:20
SPEAKER_00: It's kind of beautiful that that's no longer part of it as much.
19:24
SPEAKER_00: Open VS Code sometimes, but not really.
19:25
SPEAKER_01: Yeah, but beforehand you would prototype, right?
19:27
SPEAKER_01: Like you build prototypes.
19:29
SPEAKER_00: Yeah, yeah, yeah.
19:29
SPEAKER_00: As part of the like, what should we build?
19:31
SPEAKER_00: I include in this lots of prototyping.
19:33
SPEAKER_00: And I kind of have the privilege of being on a team where we don't have
19:36
SPEAKER_00: to build like production quality software.
19:38
SPEAKER_00: We are mostly trying to prototype and validate ideas.
19:41
SPEAKER_00: So we have a lower quality bar than someone shipping to proper github.com.
19:45
SPEAKER_00: That's a very higher quality bar.
19:47
SPEAKER_01: Yeah, which makes sense.
19:48
SPEAKER_01: I mean, that's kind of, you get the directional things, right?
19:52
SPEAKER_01: And then once it's great, you might build it or you might decide to build it.
19:55
SPEAKER_00: Yeah, yeah.
19:56
SPEAKER_00: But I definitely, I mean, we can hopefully show these later.
19:59
SPEAKER_00: I do a lot of stuff sketching.
20:00
SPEAKER_00: I mean, even just interface design,
20:01
SPEAKER_00: a lot of what you're doing is just drawing boxes and then being like,
20:04
SPEAKER_00: does this slide over from the front, from the bottom?
20:07
SPEAKER_00: Like if I click this button.
20:08
SPEAKER_01: So this is like you kind of drawing up a, UIs.
20:11
SPEAKER_01: How an interface you think could look like.
20:13
SPEAKER_01: So I see a mix of UIs, and we'll put these onto the screen.
20:21
SPEAKER_01: But UIs, description of what they do.
20:24
SPEAKER_01: Can you just talk to one of these?
20:27
SPEAKER_01: Sure, sure, sure.
20:27
SPEAKER_01: Or one that's interesting or memorable?
20:29
SPEAKER_00: Yeah.
20:30
SPEAKER_00: So this is a new, I should explain the context.
20:32
SPEAKER_00: I'm prototyping at the moment something where my theory is like,
20:35
SPEAKER_00: one of the bottlenecks with agents is planning.
20:38
SPEAKER_00: And by question 20, you're like quite tired and your brain starts shutting down.
21:04
SPEAKER_01: That was my experience.
21:04
SPEAKER_01: I went through and I got 36 questions and I was kind of,
21:07
SPEAKER_01: they were good, but I was starting to get annoyed around like, yeah, like 20 or 25.
21:11
SPEAKER_01: I'm like,
21:12
SPEAKER_01: and I had no idea when it was stopped.
21:14
SPEAKER_00: Right.
21:14
SPEAKER_00: It's endless.
21:15
SPEAKER_00: And like the human brain, you get tired.
21:17
SPEAKER_00: You can't make this many decisions in this shorter time.
21:19
SPEAKER_00: And also you don't have enough information about most of those decisions
21:22
SPEAKER_00: because it's given you a question and three options and it's told you number
21:26
SPEAKER_00: A is recommended.
21:26
SPEAKER_00: Then you just start being like, yep, A, A, enter, A, I agree with you.
21:31
SPEAKER_00: So this is not an ideal experience.
21:32
SPEAKER_00: This gets into like agents love to output reams and reams of text.
21:36
SPEAKER_00: And that's an ideal output for agents, but it is not an ideal input for humans.
21:40
SPEAKER_00: So we have a mismatch with like,
21:42
SPEAKER_00: what do humans need to be able to digest large amounts of information
21:45
SPEAKER_00: and truly understand it and be able to make informed decisions?
21:49
SPEAKER_00: That is not the interface for this.
21:50
SPEAKER_00: So I'm trying to explore how would we make an interface that got us
21:53
SPEAKER_00: to do better planning and better decision making,
21:56
SPEAKER_00: but in a way that was easier for us to comprehend.
21:59
SPEAKER_00: So this is what I'm prototyping at the moment.
22:01
SPEAKER_00: And part of this is like, OK,
22:02
SPEAKER_00: it's going to be multiplayer because,
22:03
SPEAKER_00: of course, everything should be with your team planning together.
22:06
SPEAKER_01: So several people can do it together.
22:08
SPEAKER_00: Exactly.
22:09
SPEAKER_00: But also just like how do you make it so that you have more space for each decision?
22:14
SPEAKER_00: So I'm trying to prototype like,
22:15
SPEAKER_00: OK, so first of all, this has to definitely happen in a GUI, not in a CLI.
22:19
SPEAKER_01: So these are just like rough ideas of how you can have more space.
22:22
SPEAKER_00: Exactly.
22:23
SPEAKER_00: So I'm thinking of it as like each decision maybe has its own little mini document or card.
22:27
SPEAKER_00: And depending on the decision, some of them maybe you can just have three multiple choices.
22:32
SPEAKER_00: That's fine.
22:32
SPEAKER_00: It's maybe a simple decision, right?
22:33
SPEAKER_00: A, B or C.
22:34
SPEAKER_00: But sometimes it'll ask me things like,
22:36
SPEAKER_00: do you think the borders should be like gray 10%, 12% or 15%?
22:41
SPEAKER_00: And I'm like, well, show me, this is a visual question.
22:45
SPEAKER_00: Or sometimes it's like, how do you want the architecture to be structured?
22:47
SPEAKER_00: And I'm like, well,
22:48
SPEAKER_00: show me an architecture diagram,
22:49
SPEAKER_00: show me a data flow diagram, show me a state machine.
22:51
SPEAKER_00: Like I'm trying to prototype like,
22:54
SPEAKER_00: Decisions that have come with diagrams and prototypes and HTML embedded in them so
22:58
SPEAKER_00: that depending on what question is being asked,
23:01
SPEAKER_00: it shows me the correct interface to make that decision.
23:03
SPEAKER_00: So what I'm prototyping here is like, okay, you've got a plan that's a big document.
23:06
SPEAKER_00: You've maybe got these little decision docs embedded within them that you can expand
23:10
SPEAKER_00: to see more of.
23:11
SPEAKER_00: And then looking at like,
23:13
SPEAKER_00: I think each decision should have a human assigned to it who made that decision so
23:16
SPEAKER_00: that you have an audit trail later.
23:18
SPEAKER_00: Not necessarily to hunt people down,
23:19
SPEAKER_00: but to be like, okay, Why did we make this decision about the backend?
23:23
SPEAKER_00: Well, let's go see.
23:23
SPEAKER_00: Okay, Luke made that decision like six months ago.
23:25
SPEAKER_00: Let's go open up his decision card and be like,
23:28
SPEAKER_00: what information did he have available in order to make this decision?
23:31
SPEAKER_00: Hopefully this becomes useful later.
23:32
SPEAKER_00: But here I'm just sketching.
23:33
SPEAKER_00: What are the possible shapes this could take?
23:35
SPEAKER_00: Is this like a stack of cards that expands on the screen?
23:37
SPEAKER_00: Is this like all one big linear thing?
23:39
SPEAKER_00: Are we like swiping through stuff like it's Tinder?
23:41
SPEAKER_00: Like what interfaces are going to be useful for this?
23:44
SPEAKER_00: So this is what a lot of the sketching is just trying to figure out.
23:47
SPEAKER_00: Like possible, possible shapes of things.
23:50
SPEAKER_00: I have all kinds.
23:51
SPEAKER_00: Oh, these are like characters because I was trying to make an ESP character.
23:54
SPEAKER_01: Oh, can you show me the character?
23:55
SPEAKER_00: Well, this is like, do you know, oh, this is like my kids scribbling all over my notebook.
23:59
SPEAKER_00: This is like, you know, ESP32s, you know, Steve Ruiz has like sold everyone on buying an ESP32.
24:04
SPEAKER_00: It's like this little device with a screen and a very small microchip that has Wi-Fi
24:07
SPEAKER_00: and Bluetooth.
24:08
SPEAKER_00: Yeah.
24:08
SPEAKER_00: And so I was designing this little character who like tells me the weather
24:11
SPEAKER_00: and like I'm trying to hook it up to my agent
24:13
SPEAKER_00: so it can be like you have this much capacity left before your reset limit happens.
24:16
SPEAKER_00: So like this is just all kinds of things.
24:18
SPEAKER_01: It's a mix of like like some serious stuff and some fun stuff.
24:23
SPEAKER_00: Yeah.
24:23
SPEAKER_00: Yeah.
24:23
SPEAKER_00: And I can show like the whole thing with notebooks and thinking on paper.
24:26
SPEAKER_00: I think all designers do this.
24:27
SPEAKER_00: But coming from illustration,
24:29
SPEAKER_00: the reason I do this is because I started in a world where you have
24:32
SPEAKER_00: to draw everything to like figure out what you're going to do.
24:34
SPEAKER_00: So when I was an illustrator,
24:35
SPEAKER_00: I had tons of these notebooks where you're figuring out like What's the composition?
24:40
SPEAKER_00: What's the physical shape of things?
24:42
SPEAKER_00: How are you working out the shape of the grass in a scene?
24:46
SPEAKER_00: And so I think I started problem solving on paper very early on in my career,
24:51
SPEAKER_00: figuring out compositions and layouts.
24:53
SPEAKER_01: Was this in college?
24:55
SPEAKER_00: When was this?
24:56
SPEAKER_00: I think I was working for Egghead when I was doing this.
24:59
SPEAKER_01: And you drew all of these?
25:00
SPEAKER_00: Yeah.
25:01
SPEAKER_00: Yeah, I was in LA for a while, which is a terrible place to live.
25:05
SPEAKER_00: But I trained with people who are concept artists on films.
25:09
SPEAKER_00: And they have a really beautiful way of working that's very technical.
25:12
SPEAKER_00: It's very constructing things from 3D shapes and drawing in space.
25:16
SPEAKER_01: Yeah, because these are really 3D.
25:17
SPEAKER_00: Yeah, yeah.
25:19
SPEAKER_00: And they kind of just teach you how to do...
25:22
SPEAKER_00: So I think doing this set me up to when I moved into like UI design to do...
25:43
SPEAKER_00: A lot of this kind of sketching because,
25:46
SPEAKER_01: you know, it just comes naturally.
25:47
SPEAKER_01: I mean, these are a little bit easier to sketch as well in terms of the user interfaces.
25:52
SPEAKER_00: Yeah.
25:52
SPEAKER_00: But then I think it's like that it helps you think on paper that all
25:54
SPEAKER_00: of this is just sort of like you have an idea in your head.
25:57
SPEAKER_00: And sure, I could like go into Claude code and be like, hey, Claude, here's an idea I have.
26:01
SPEAKER_00: It's like a stack of cards and it's an accordion.
26:04
SPEAKER_00: But it's much faster to just get a piece of pen,
26:07
SPEAKER_00: like pen or pencil on the desk next to me and draw that with my hands.
26:12
SPEAKER_00: It's like way few effort.
26:13
SPEAKER_00: So I feel like people online keep being like,
26:15
SPEAKER_00: oh, like everyone's just going to prompt everything to create.
26:19
SPEAKER_00: But you need something that is like quick feedback and very loose in
26:22
SPEAKER_00: the early stages to figure out the shape of something before you can put
26:26
SPEAKER_00: into words what you want an agent to do.
26:27
SPEAKER_00: And also it's visual.
26:29
SPEAKER_00: It's not text.
26:30
SPEAKER_01: And is it not a bit more satisfying doing it with your hand?
26:32
SPEAKER_01: So much better.
26:33
SPEAKER_01: Yeah.
26:33
SPEAKER_00: Yeah.
26:34
SPEAKER_00: And then you can look at it.
26:35
SPEAKER_00: It doesn't go away on your screen.
26:36
SPEAKER_00: It can sit on your desk and you can like the next day be like, oh,
26:38
SPEAKER_00: yes, I remember I was trying to like figure out what shape this feature should be.
26:44
SPEAKER_00: Or like you can draw like data diagrams, whatever you want.
26:47
SPEAKER_00: It doesn't have to be visual,
26:48
SPEAKER_00: but it's just a way of externalizing thoughts before they're not linguistic.
26:53
SPEAKER_00: I think this is the key thing is like, again, agents only accept text as inputs.
26:58
SPEAKER_00: OK, they can read images like I do take photos and put them in.
27:01
SPEAKER_00: But they're not as good at images.
27:02
SPEAKER_00: They're very bad at spatial reasoning.
27:04
SPEAKER_00: They're very bad at visual design.
27:06
SPEAKER_00: I mean, trying to get them to do design,
27:07
SPEAKER_00: they just make mistakes where like they just don't put spacing around things
27:10
SPEAKER_00: and things of the wrong size and they make text overlap.
27:12
SPEAKER_00: They can't see, right?
27:14
SPEAKER_00: Like trying to explain a visual idea in text to an agent is really challenging
27:19
SPEAKER_00: and doesn't work very well.
27:21
SPEAKER_00: So I find I still end up doing a lot of my design without agents upfront
27:25
SPEAKER_00: because all it is is thinking through the visual pieces of it.
27:29
SPEAKER_00: And then when I'm like, okay,
27:30
SPEAKER_00: I know this is the shape of the thing I want, then I can tell an agent to do it.
27:33
SPEAKER_00: But I find it hard to like involve it earlier in a way that I think some people
27:37
SPEAKER_00: on Twitter are claiming they do, but I'm skeptical.
27:40
SPEAKER_01: But this is interesting because even in software design,
27:43
SPEAKER_01: I'm talking about architecture design,
27:44
SPEAKER_01: some of the,
27:45
SPEAKER_01: I guess,
27:45
SPEAKER_01: most productive sessions I've observed have been in person around the whiteboard
27:50
SPEAKER_01: where people, again,
27:52
SPEAKER_01: we're talking about like components,
27:54
SPEAKER_01: databases,
27:54
SPEAKER_01: networking connections,
27:56
SPEAKER_01: retry logics,
27:56
SPEAKER_01: whatever,
27:57
SPEAKER_01: you know, like these things you could describe or you could put it on a computer.
28:00
SPEAKER_01: But when you put it on a board,
28:02
SPEAKER_01: when someone puts it on there and then someone else takes and,
28:05
SPEAKER_01: you know,
28:05
SPEAKER_01: they kind of They're forced to take their ideas into 2D space
28:10
SPEAKER_01: because we don't do 3D.
28:13
SPEAKER_01: I know you can draw like cool three, but we can also just box scenarios.
28:17
SPEAKER_01: Someone does that and the other people understand and they go in
28:20
SPEAKER_01: and they add their own thing or they circle or they add new components.
28:25
SPEAKER_01: And even in digital space, I think Meijer was a good example.
28:28
SPEAKER_01: They became so popular because they figured out a way to do collaborative whiteboarding.
28:34
SPEAKER_01: You don't have to be in the same room, but they give you somewhat similar tools.
28:37
SPEAKER_01: So I wonder if this whole thing of Of getting your ideas either to
28:42
SPEAKER_01: a physical medium,
28:43
SPEAKER_01: which I think for us engineers is the whiteboard, for you is the notebook.
28:48
SPEAKER_01: Maybe it just helps maybe rethink or solidify or also it does have
28:53
SPEAKER_01: a forcing function about the verbs, the shapes that you...
28:57
SPEAKER_00: Yeah, yeah, yeah.
28:58
SPEAKER_00: And this gets into kind of like the interfaces we have to agents right now are so primitive.
29:03
SPEAKER_00: I think we know this, right?
29:04
SPEAKER_00: We're all like a couple of years into this entire thing, which is wild.
29:07
SPEAKER_00: Maybe five years.
29:08
SPEAKER_00: I forget when GPT-4 came out.
29:11
SPEAKER_00: Was it three or four years ago?
29:14
SPEAKER_01: ChatGPT 3.5 came out in three years.
29:16
SPEAKER_01: It was November or December.
29:17
SPEAKER_01: November 2022.
29:18
SPEAKER_01: Yeah, yeah, yeah.
29:19
SPEAKER_00: This is when I was at Elicit and then because we were debating doing
29:22
SPEAKER_00: a chat interface and then they did it and we were like,
29:23
SPEAKER_00: oh, they like stole our idea.
29:26
SPEAKER_01: Not that Elicit.
29:27
SPEAKER_00: Is necessarily competing with them.
29:29
SPEAKER_00: But that is no time at all.
29:30
SPEAKER_00: And it's kind of wild we've made it as far as we have.
29:32
SPEAKER_00: But it's also like, I always say, software design is an extremely young field.
29:37
SPEAKER_00: Like, what are we, 60 years into it at most?
29:39
SPEAKER_00: So even that,
29:40
SPEAKER_00: we haven't figured out a lot of things about how to design the best interfaces
29:43
SPEAKER_00: for people and machines to like communicate with each other.
29:45
SPEAKER_00: And agents barely at all.
29:47
SPEAKER_00: I feel like there's this world that agents live in.
29:50
SPEAKER_00: There's like weights and Models and skills and MCPs.
29:52
SPEAKER_00: And then you have your human side that is like physicality and texture and light
29:56
SPEAKER_00: and materials and all these things agents don't understand and trying
29:59
SPEAKER_00: to find like artifacts that allow us to meet in the middle
30:03
SPEAKER_00: and create stuff together is like the really hard challenge
30:05
SPEAKER_00: because you've got two totally different types of beings.
30:08
SPEAKER_00: Not that agents are conscious beings.
30:10
SPEAKER_00: I'm not in that camp.
30:12
SPEAKER_00: But they're a type of intelligence that like wants to think and act in a certain way
30:16
SPEAKER_00: that is not the way humans want to think and act.
30:18
SPEAKER_00: And it's hard to like translate between the two.
30:20
SPEAKER_00: I just find myself very frustrated that they can't sort of,
30:23
SPEAKER_00: you know,
30:23
SPEAKER_00: be looking over my shoulder,
30:24
SPEAKER_00: looking at my notebook and be understanding what I'm drawing
30:27
SPEAKER_00: and like helping me move my ideas along.
30:29
SPEAKER_00: This is like the eventual dream is they understand space and light and shape
30:34
SPEAKER_00: and lines and, you know, but I think we're quite a ways away from that.
30:38
SPEAKER_01: So one thing,
30:39
SPEAKER_01: we've talked about your notebooks and digital tools,
30:42
SPEAKER_01: but one thing that you've posted about recently is woodworking.
30:46
SPEAKER_01: You said, I'm at the stage of my software design journey.
30:50
SPEAKER_01: Well, no, I'm at the stage of software design where you start taking woodworking courses.
30:53
SPEAKER_01: Can you tell me a little bit about that experience?
30:56
SPEAKER_00: Well, the little bit of the joke is like,
30:58
SPEAKER_00: I just feel like at some point,
30:59
SPEAKER_00: every engineer,
30:59
SPEAKER_00: you have a choice of things you can get into
31:01
SPEAKER_00: because you feel like you're like disembodied from the world of working in software.
31:04
SPEAKER_00: So you have to pick ceramics, bread baking.
31:07
SPEAKER_00: You can maybe pick like motorcycle repair or something.
31:09
SPEAKER_00: But woodworking is a pretty popular one because it's like engineering.
31:12
SPEAKER_00: There's a lot of like measurement and being precise.
31:15
SPEAKER_00: So, yeah,
31:15
SPEAKER_00: I'm learning woodworking because I bought a house here and all the houses
31:19
SPEAKER_00: in London are very old.
31:20
SPEAKER_00: So you buy a house that needs a lot of work.
31:22
SPEAKER_00: That is like always what happens.
31:23
SPEAKER_00: Maybe the same in Amsterdam.
31:24
SPEAKER_00: I think old housing stuff.
31:25
SPEAKER_00: So then you suddenly become like, oh,
31:27
SPEAKER_00: I need to learn like DIY skills,
31:29
SPEAKER_00: which I will say,
31:30
SPEAKER_00: you know, Claude,
31:31
SPEAKER_00: ChatGPT, very helpful coaches in this regard, along with YouTube and Instagram.
31:35
SPEAKER_00: I can learn a lot.
31:36
SPEAKER_00: But I just was like, oh,
31:37
SPEAKER_00: there's so many things I need solutions to that I need to figure out how to make.
31:41
SPEAKER_00: I need to like make shelves here.
31:42
SPEAKER_00: I need to like fix this banister.
31:44
SPEAKER_00: So I was like, well,
31:46
SPEAKER_00: I have no DIY skills,
31:47
SPEAKER_00: no wordworking, but just signed up for a course being like, well, I'll learn.
31:50
SPEAKER_00: I'm pretty sure I can like, I'm not too old.
31:52
SPEAKER_00: I can acquire new skills and I'm sure I'll find lots of parallels to software,
31:56
SPEAKER_00: but in a more satisfying way where you actually like touch the thing you make.
31:59
SPEAKER_01: One thing that you've talked about recently,
32:02
SPEAKER_01: You've been talking about is the concept of design engineers.
32:06
SPEAKER_01: For a few years ago,
32:07
SPEAKER_01: you posted,
32:08
SPEAKER_01: I'll quote you,
32:10
SPEAKER_01: I'm having a strong, should I become a full-blown design engineer day?
32:12
SPEAKER_01: I'm not even sure what it means,
32:14
SPEAKER_01: but I just want to touch a lot of code and solve tangible problems on screens
32:18
SPEAKER_01: and make beautiful animated stuff.
32:20
SPEAKER_00: Yeah.
32:20
SPEAKER_01: And then you went over and you tried to collect names of people who you knew,
32:25
SPEAKER_01: who you felt were kind of doing this design engineering.
32:27
SPEAKER_01: That was a few years ago.
32:28
SPEAKER_00: Yeah.
32:29
SPEAKER_01: What have you figured out about design engineers,
32:34
SPEAKER_01: if they exist, what they do, what places they work in?
32:38
SPEAKER_00: Yeah, I think they do exist.
32:40
SPEAKER_00: I think Twitter might have a different definition or like I think there's a lot
32:43
SPEAKER_00: of people where not that I should call it X.
32:45
SPEAKER_00: Sorry, X. Twitter, X, all the same.
32:49
SPEAKER_00: On X, I feel like a lot of people who get called design engineers or
32:52
SPEAKER_00: who present themselves
32:52
SPEAKER_00: as design engineers actually are like very good micro interaction designers,
32:57
SPEAKER_00: like a little like here's a cool hover effect on a button and like here's
33:01
SPEAKER_00: a cool loading transition state.
33:02
SPEAKER_00: And those things are definitely cool.
33:04
SPEAKER_00: I don't consider that design engineering because you could Achieve most
33:08
SPEAKER_00: of those things by just telling an agent to do those without looking at
33:10
SPEAKER_00: a single piece of code.
33:12
SPEAKER_00: To me, that's just like very advanced, sophisticated motion design and visual design.
33:18
SPEAKER_00: And that's cool,
33:19
SPEAKER_00: but I don't consider that design engineering because I don't think that's like
33:22
SPEAKER_00: a full job for you to do that as a career.
33:25
SPEAKER_00: But design engineering now,
33:27
SPEAKER_00: like the people who I think of as good design engineers are Do the kind
33:30
SPEAKER_00: of thing I was talking about before where you step much more into
33:33
SPEAKER_00: the engineering side of work.
33:34
SPEAKER_00: So you're still a designer.
33:35
SPEAKER_00: You're still caring about like product nouns and verbs and like the visual design.
33:39
SPEAKER_00: But then you really work with engineers closely and or are like directly involved
33:44
SPEAKER_00: in implementing and writing code yourself.
33:47
SPEAKER_00: And you fully understand, well, you know, not fully, you don't have to be like full, full stack.
33:52
SPEAKER_00: But like you have a deep understanding of the technical...
33:55
SPEAKER_00: The technical architecture of the product you're building.
33:57
SPEAKER_00: You really are like,
33:58
SPEAKER_00: okay, given the shape of the backend data, what is possible in the interface?
34:01
SPEAKER_00: Like that sort of design works.
34:04
SPEAKER_00: Given what models are capable of and what kind of custom skills we're building
34:08
SPEAKER_00: into this product,
34:10
SPEAKER_00: how do I explain to users what the capabilities of this product are?
34:14
SPEAKER_00: Truly digging into the technicals and not living in the world
34:17
SPEAKER_00: of just user interviews and the market and what color is the sidebar,
34:22
SPEAKER_00: but caring much more and working much more closely with the engineers.
34:25
SPEAKER_00: And almost always, I feel like implementing a lot of it yourself.
34:29
SPEAKER_00: I've always done my own front-end work just because it's easier.
34:32
SPEAKER_00: And then the engineers I work with are usually thrilled because they hate CSS.
34:36
SPEAKER_00: And then they get to go work on like the more interesting, difficult stuff, I'd say.
34:39
SPEAKER_00: It's like the syncing or like the back of the front end work,
34:44
SPEAKER_00: you know, like the more logic stuff.
34:45
SPEAKER_00: They really get to engage in that.
34:46
SPEAKER_00: And they don't have to worry about like, is this the right border radius on something?
34:50
SPEAKER_00: Like, I don't think they very rarely care.
34:52
SPEAKER_00: And I do care.
34:53
SPEAKER_00: So it's always worked out well to kind of do that.
34:55
SPEAKER_00: So I just define it as someone who's like deeply into the engineering side of the design.
35:00
SPEAKER_01: What's in your past positions,
35:02
SPEAKER_01: current positions,
35:03
SPEAKER_01: through friends who are also designers,
35:06
SPEAKER_01: what have you seen great engineering and design collaboration look like?
35:11
SPEAKER_01: May that be with designers or design engineers?
35:14
SPEAKER_00: I mean, I do find being a design engineer,
35:16
SPEAKER_00: I've always had amazing collaborations with engineers because,
35:19
SPEAKER_00: again, you do the bit they didn't ever want to do.
35:21
SPEAKER_00: And you take away what I assume is most of the tension between designers
35:24
SPEAKER_00: and engineers that I've never experienced because I've always been
35:27
SPEAKER_00: a front-end person.
35:28
SPEAKER_00: But I've heard of people,
35:29
SPEAKER_00: you know,
35:29
SPEAKER_00: there being the tension where you have like someone's made some perfect Figma file
35:33
SPEAKER_00: and the engineer hasn't implemented it exactly to spec.
35:35
SPEAKER_00: Yes, I've been there.
35:38
SPEAKER_00: Yeah.
35:38
SPEAKER_00: And obviously there's a tension there because someone made something in a medium.
35:42
SPEAKER_01: Or there's a nice, great.
35:43
SPEAKER_01: Gradient and it's mobile and doing gradients would absolutely wreck the scroll performance.
35:49
SPEAKER_01: And again, these are like we're talking about years back when this was a thing.
35:53
SPEAKER_01: And now you're going back like can I use a single color or you try
35:57
SPEAKER_01: to explain like what you can do with your either performance limitations or think
36:01
SPEAKER_01: about low on devices on Android or iOS versus Android where a designer might be more
36:06
SPEAKER_01: into the iOS world and Again,
36:08
SPEAKER_01: a good designer would not,
36:09
SPEAKER_01: but there's oftentimes the engineer would come like,
36:11
SPEAKER_01: all right, we have constraints here for whatever reason.
36:15
SPEAKER_00: Yeah, I think this gets into like,
36:17
SPEAKER_00: I don't want to blame the designers because I think it's a failure of tools because
36:21
SPEAKER_00: the design tools that everyone used to use,
36:23
SPEAKER_00: Sketch and Figma,
36:24
SPEAKER_00: these like pixel mock-ups,
36:26
SPEAKER_00: they have no relationship to the constraints of the medium,
36:28
SPEAKER_00: which is like whatever you're building for, iOS or desktop or the web.
36:31
SPEAKER_00: You have to understand the materials you're building with in any design role.
36:35
SPEAKER_00: A table designer would not be oblivious to how oak performs in certain contexts
36:41
SPEAKER_00: or how pine dents in a certain way.
36:43
SPEAKER_00: And in the same way,
36:44
SPEAKER_00: I think if you're designing for the web
36:45
SPEAKER_00: and you don't understand performance It's friction, yeah.
37:09
SPEAKER_00: The engineers come back and told me
37:11
SPEAKER_00: that we have this like error state I've never heard of in my life.
37:14
SPEAKER_00: Like explain to me why that error state would occur,
37:16
SPEAKER_00: you know, and help me build a state machine.
37:18
SPEAKER_00: Right.
37:18
SPEAKER_00: Like there's like all these new tools available,
37:19
SPEAKER_00: but those old frictions I have to assume were just like tool based things.
37:24
SPEAKER_00: And I would say designers that are maybe stuck in the past who don't want to let go
37:29
SPEAKER_00: of like, oh, I make pixel mockups.
37:31
SPEAKER_00: I don't know if anyone's there anymore,
37:32
SPEAKER_00: but if they still are, I mean,
37:35
SPEAKER_00: you're just, yeah, I can't imagine that relationship going well in the future.
37:38
SPEAKER_01: And then you mentioned Figma.
37:40
SPEAKER_01: Figma was such a popular tool for a while.
37:43
SPEAKER_01: That was the interface between engineers and designers.
37:47
SPEAKER_01: How do you use Figma today or how have you used it in the past?
37:50
SPEAKER_00: I think I've always used it.
37:52
SPEAKER_00: Like not to get to high fidelity mockups at all.
37:54
SPEAKER_00: I think to get like the rough shape of things past a notebook.
37:57
SPEAKER_00: It's like a notebook.
37:57
SPEAKER_00: You can be like, OK, the shape of it is like this, but it's such a rough sketch.
38:01
SPEAKER_00: And Figma is useful, I find,
38:02
SPEAKER_00: for being like, OK,
38:03
SPEAKER_00: exactly what color is like the right amount of contrast to like draw attention
38:06
SPEAKER_00: to this element on the page or like exactly what size does this text need to be
38:10
SPEAKER_00: to like flow well.
38:12
SPEAKER_00: But then I would never take it to high fidelity because,
38:14
SPEAKER_00: of course, everything always looks different.
38:16
SPEAKER_00: I mean, I primarily design for the web.
38:18
SPEAKER_00: Everything looks different in the browser, right?
38:19
SPEAKER_00: Depends on the font rendering and all this kind of stuff.
38:22
SPEAKER_00: Responsiveness, and then when exactly your breakpoints.
38:25
SPEAKER_00: So I would always take it into browsers pretty early on.
38:28
SPEAKER_00: You get like medium fidelity in Figma,
38:31
SPEAKER_00: like the shape of it,
38:31
SPEAKER_00: and then you take it into a prototype and then you can really tweak and refine.
38:36
SPEAKER_00: And of course,
38:36
SPEAKER_00: with agents now,
38:37
SPEAKER_00: it's just like,
38:37
SPEAKER_00: I point an agent at my Figma mock-up and I'm like,
38:40
SPEAKER_00: just get all of that in there and then make,
38:43
SPEAKER_00: you know,
38:44
SPEAKER_00: what's called a jig,
38:45
SPEAKER_00: which is where you get like little sliders and variables attached to like
38:47
SPEAKER_00: a little... A jig.
38:48
SPEAKER_00: A jig comes from woodworking.
38:50
SPEAKER_00: Or you make like a little device that helps you do one specific job.
38:54
SPEAKER_00: So with design,
38:55
SPEAKER_00: when you're working on a live prototype,
38:57
SPEAKER_00: you say like, okay, here are the variables I'm not sure about.
38:59
SPEAKER_00: I'm not sure about like my headline size.
39:01
SPEAKER_00: I'm not sure about these colors.
39:03
SPEAKER_00: Give me sliders and color pickers and like all these things or like
39:06
SPEAKER_00: the animation curve I'm not sure about.
39:08
SPEAKER_00: Give me and then I'll tweak it live.
39:10
SPEAKER_00: And then when I have the values just right in the live version,
39:13
SPEAKER_00: then we'll commit those to be the actual values.
39:15
SPEAKER_00: Much faster.
39:16
SPEAKER_00: It's like build your own Figma as needed.
39:18
SPEAKER_01: And inside of GitHub, how do designers work?
39:22
SPEAKER_01: How do yourself and other designers work?
39:24
SPEAKER_01: Because it's a bigger organization.
39:26
SPEAKER_01: You're building a tool for developers with a bunch of different parts.
39:30
SPEAKER_00: I mean, I don't spend that much time with like the GitHub design org proper.
39:33
SPEAKER_00: I know some people in there, but...
39:35
SPEAKER_01: No, because you're next.
39:36
SPEAKER_00: So the GitHub next team is a little bit isolated.
39:38
SPEAKER_00: Like we still have good relationships with the rest of the org,
39:40
SPEAKER_00: but a little bit by design,
39:41
SPEAKER_00: we're supposed to be kind of like the R&D team out on the side,
39:45
SPEAKER_00: kind of doing weird stuff and then trying to convince the rest of the org
39:48
SPEAKER_00: that we're right and they should like pay attention and do our thing,
39:51
SPEAKER_00: which is a whole different politics thing.
39:53
SPEAKER_00: So I like kind of see the designers,
39:54
SPEAKER_00: but I think they have a very different job to me in that they have to like...
39:58
SPEAKER_00: Uphold quality standards and they're working on like a big design system and like
40:01
SPEAKER_00: an old Ruby on Rails app.
40:02
SPEAKER_00: Like they've got very different constraints to me.
40:04
SPEAKER_01: Yeah, and they have like millions or even more, like tens of millions of customers.
40:08
SPEAKER_01: Yes.
40:10
SPEAKER_01: That's now a different thing.
40:11
SPEAKER_01: They've been used to certain things or finding certain things or certain patterns.
40:14
SPEAKER_00: Exactly, exactly.
40:15
SPEAKER_00: Changing anything on .com is like a whole political thing because customers expect,
40:20
SPEAKER_00: you know, them to come in and they have critical workflows.
40:22
SPEAKER_00: You can't move their button.
40:23
SPEAKER_00: But the team I work on, there's me and one other guy who are both design engineers.
40:27
SPEAKER_00: So we both like do engineering work, but we are more the design people.
40:30
SPEAKER_00: And everyone else is much more engineer.
40:31
SPEAKER_00: And they're all really senior.
40:33
SPEAKER_00: So it's fun to work with them because like they've seen everything.
40:35
SPEAKER_00: They like know how to architect an app.
40:37
SPEAKER_00: But we don't work that differently to engineers, I find.
40:39
SPEAKER_00: I mean, we're putting in PRs the same as them.
40:41
SPEAKER_00: We're working on the same tools and materials.
40:43
SPEAKER_00: It's much more that thing of what bits of the app we care about.
40:46
SPEAKER_00: I, of course, care about the interface part.
40:49
SPEAKER_00: And then my engineer partners,
40:50
SPEAKER_00: of course, care about the backend and the data flow and everything.
40:53
SPEAKER_00: So same tools, but different considerations, I guess.
40:58
SPEAKER_01: I wanted to take some time to mention our presenting sponsor, TurboBuffer.
41:01
SPEAKER_01: TurboPuffer couldn't be a better sponsor for this episode
41:04
SPEAKER_01: as they have what I think is one of the most refreshing brands in tech right now.
41:08
SPEAKER_01: TurboPuffer's co-founder Simon describes their brand as hardcore and whimsical
41:13
SPEAKER_01: and their design perfectly reflects this.
41:15
SPEAKER_01: TurboPuffer, the database, is extremely hardcore AI infrastructure.
41:20
SPEAKER_01: It supports critical search workloads for Entropic, Notion,
41:23
SPEAKER_01: Bridgewater,
41:23
SPEAKER_01: Cognition and many more of the largest and fastest growing AI companies in
41:27
SPEAKER_01: the world.
41:28
SPEAKER_01: They take reliability, performance, and scalability very seriously.
41:32
SPEAKER_01: But TurboPuffer, the team, refuses to be another sterile, boring enterprise tech company.
41:37
SPEAKER_01: Just look at their website.
41:39
SPEAKER_01: It's exactly what you would want the website for a database to look like.
41:43
SPEAKER_01: It's simple, modest, and transparent.
41:45
SPEAKER_01: They even put the limits on the homepage.
41:48
SPEAKER_01: But it also isn't boring.
41:50
SPEAKER_01: It's nice to look at,
41:51
SPEAKER_01: it's very well thought out,
41:52
SPEAKER_01: and their little ASCII diagrams perfectly blend hardcore engineering
41:56
SPEAKER_01: with whimsical design.
41:57
SPEAKER_01: Plus, if you've ever tried to generate a good ASCII diagram with AI,
42:01
SPEAKER_01: you'll quickly realize that they create all of these by hand, not with AI.
42:06
SPEAKER_01: TurboBuffer's brand is the perfect reflection of who they are, and that's why it works so well.
42:11
SPEAKER_01: The brand is hardcore and whimsical, and so is the team.
42:13
SPEAKER_01: I highly recommend Turbo Puffer not only as the extremely scalable search engine
42:17
SPEAKER_01: for AI,
42:18
SPEAKER_01: but also as a group of people that will go to exceptional lengths to deliver
42:22
SPEAKER_01: for their customers.
42:24
SPEAKER_01: You can check them out at turbopuffer.com slash pragmatic.
42:27
SPEAKER_01: I also want to mention our season sponsor, Entire.
42:30
SPEAKER_01: Maggie mentioned that as a design engineer,
42:32
SPEAKER_01: she's still putting in PRs the same as everyone else.
42:35
SPEAKER_01: And every one of those PRs is pushed to Git hosting.
42:38
SPEAKER_01: But there is a problem with Git agents.
42:41
SPEAKER_01: Git is increasingly becoming a bottleneck for modern, agent-heavy software development.
42:45
SPEAKER_01: Devs are creating more code with agents,
42:47
SPEAKER_01: these agents are pushing more code,
42:49
SPEAKER_01: and many devs are running parallel agents that are pushing even more code.
42:53
SPEAKER_01: GitHub is clearly struggling to keep up and has frequent outages.
42:56
SPEAKER_01: So what is the solution?
42:58
SPEAKER_01: Entire was founded by GitHub's last CEO,
43:00
SPEAKER_01: Tomasz Dömke, and he rebuilt GitHub hosting for the Agenda Kara from scratch.
43:05
SPEAKER_01: Entire was built to be very fast and have your repos regionally close to you
43:09
SPEAKER_01: to reduce latency, allowing for fleets of agents to push in parallel.
43:13
SPEAKER_01: And Entire's performance is next level.
43:15
SPEAKER_01: The repo can handle 418 pushes per second,
43:19
SPEAKER_01: and that's up to 89 times faster than every other competitor on the market.
43:23
SPEAKER_01: When GitHub is down,
43:24
SPEAKER_01: you can still keep working, and you don't need to migrate away from GitHub.
43:27
SPEAKER_01: You just sign up to Entire, and the platform mirrors your repo.
43:31
SPEAKER_01: Oh, and one more thing.
43:32
SPEAKER_01: Have you ever wondered what prompt resulted in this code being generated?
43:36
SPEAKER_01: I find that the prompt and conversation with the agents carries more information
43:39
SPEAKER_01: than the PR itself.
43:41
SPEAKER_01: Entire captures all the prompt history with your agent right in the repo, easy to check back.
43:46
SPEAKER_01: And it's got a pretty innovative UI to show all of this.
43:49
SPEAKER_01: If you're looking for get hosting that works even when GitHub is down,
43:52
SPEAKER_01: head to entire.io slash pragmatic,
43:54
SPEAKER_01: install the CLI and mirror your repo with a click.
43:57
SPEAKER_01: I've already done it.
43:58
SPEAKER_01: Oh, and did I mention that it works with any agent and
44:01
SPEAKER_00: is open source?
44:02
SPEAKER_01: And then with LLM, since they came out, how has your design process changed?
44:06
SPEAKER_01: You've already mentioned how it's now a lot easier.
44:09
SPEAKER_01: You can like,
44:10
SPEAKER_01: for example,
44:11
SPEAKER_01: if you either have an interface of design,
44:13
SPEAKER_01: you can ask, It's definitely all faster and it definitely all feels a bit easier.
44:32
SPEAKER_00: I will say, like,
44:32
SPEAKER_00: I definitely remember not suffering through, like,
44:35
SPEAKER_00: making elaborate prototypes in Figma because that was going to be faster than,
44:39
SPEAKER_00: like, building more complex things in the web.
44:41
SPEAKER_00: And then trying to show them to users with these,
44:43
SPEAKER_00: like, Figma click-through prototypes that obviously, like, everything is faked.
44:48
SPEAKER_01: Oh, I remember that, like, on the user thing.
44:50
SPEAKER_01: And then sometimes you have to, like, have these weird things that click here and then...
44:53
SPEAKER_00: So you're like in these user interviews and like the user's trying
44:56
SPEAKER_00: to click something that you haven't hooked up to like a fake new screen.
44:59
SPEAKER_00: And the whole thing just feels a bit insincere.
45:02
SPEAKER_00: And then even when you did make prototypes, you couldn't spend any time making them look nice.
45:05
SPEAKER_00: So then sometimes the users would get like a bit distracted or confused by
45:09
SPEAKER_00: the fact it looked Terrible.
45:10
SPEAKER_00: But you're like, yeah,
45:11
SPEAKER_00: but we only had a few days and then we've got to move on to the next prototype.
45:14
SPEAKER_00: And now I wouldn't have that problem at all, right?
45:16
SPEAKER_00: Like now I can build something that really looks quite sophisticated.
45:19
SPEAKER_00: And not with the agents doing it on their own.
45:21
SPEAKER_00: They need a lot of direction,
45:22
SPEAKER_00: I find, to make something that I would think was an acceptable interface.
45:26
SPEAKER_00: But it's so much faster.
45:28
SPEAKER_00: Sometimes I have even mocked up quite high fidelity stuff in Figma and
45:31
SPEAKER_00: then just been like, hey, agent, here's the Figma file, you know, goal mode.
45:35
SPEAKER_00: You...
45:36
SPEAKER_00: Playwright, take screenshots.
45:38
SPEAKER_00: If it doesn't look like the Figma, keep going.
45:40
SPEAKER_00: Keep looping until it looks exactly like the Figma, and it'll get there.
45:44
SPEAKER_00: And so that was just like, I can run that overnight.
45:46
SPEAKER_00: I don't have to do anything in the morning.
45:47
SPEAKER_00: The interface is pretty much there as I specced it out.
45:50
SPEAKER_01: And to you, I guess this was grunt work, right?
45:53
SPEAKER_00: Totally.
45:53
SPEAKER_00: It's totally.
45:54
SPEAKER_00: Because you're like, this isn't complex.
45:56
SPEAKER_00: This is like, okay, again, we're building another sidebar, right?
45:58
SPEAKER_00: Let's get the React components and like...
46:00
SPEAKER_01: And you don't really care about the cool quality or tech depth or any of it.
46:03
SPEAKER_01: It's going to be a throwaway anyway, right?
46:06
SPEAKER_01: If it works because of that, because then you'll build it.
46:08
SPEAKER_01: And if it doesn't work, like, create.
46:10
SPEAKER_01: It doesn't matter.
46:11
SPEAKER_00: So prototyping is a whole different game now.
46:13
SPEAKER_00: I feel like you can prototype much more ambitious things.
46:16
SPEAKER_00: And then I do love the whole jig thing.
46:18
SPEAKER_00: It feels like this malleable software thing where you can make
46:20
SPEAKER_00: the software whatever you want it to be,
46:21
SPEAKER_00: where I'm not constrained by Figma's available tools.
46:24
SPEAKER_00: I can be like, okay,
46:25
SPEAKER_00: I'm trying to make this animated constellation map with something I recently did.
46:28
SPEAKER_00: What speed are the stars moving at?
46:30
SPEAKER_00: What zoom level are we at?
46:32
SPEAKER_01: You did that?
46:33
SPEAKER_00: Yeah.
46:33
SPEAKER_00: And what degree is the gradient at on the screen?
46:36
SPEAKER_00: And you just put these all on little sliders and you just tweak it.
46:39
SPEAKER_00: And then you think, oh, I really want a slight...
46:43
SPEAKER_00: So this is
46:51
SPEAKER_00: something I built for the new GitHub Next website.
46:55
SPEAKER_00: We have a kind of basic one up, but I wanted something a bit fancier.
46:57
SPEAKER_00: We want to make it a little bit easier to publish research frequently.
47:00
SPEAKER_00: So I was playing around with some jazzy homepages.
47:05
SPEAKER_00: And this whole thing, this is like an animated spinning constellation.
47:08
SPEAKER_00: And these are some projects that the teams worked on over the years.
47:12
SPEAKER_00: And like this kind of thing to prototype this, right?
47:14
SPEAKER_00: Of course,
47:15
SPEAKER_00: I started with the sketch and notebook,
47:17
SPEAKER_00: but then made a bunch of jigs where I was like, okay, what color is the background?
47:21
SPEAKER_00: And like, how fast is this map spinning?
47:23
SPEAKER_00: And then there's a bunch of like physics,
47:24
SPEAKER_00: cool physics on the stars that I didn't have to do any physics.
47:27
SPEAKER_00: You know, I didn't have to know anything about that.
47:30
SPEAKER_00: You know, how big are these when you hover over them?
47:32
SPEAKER_00: And then I did this like cool scroll effect down into the proper page.
47:35
SPEAKER_00: Yeah.
47:36
SPEAKER_00: And all of that, right, I didn't write any code.
47:39
SPEAKER_00: I made a bunch of different jigs and tweaked and adjusted exactly how much.
47:42
SPEAKER_01: And so these were,
47:43
SPEAKER_01: just imagine,
47:44
SPEAKER_01: these were like things on the screen,
47:45
SPEAKER_01: for example,
47:45
SPEAKER_01: the color you could go from like blue, green, red,
47:49
SPEAKER_01: and then you figure out,
47:50
SPEAKER_01: all right, you like this the most, the sizes, all of those, right?
47:53
SPEAKER_00: Exactly.
47:54
SPEAKER_00: I can be like,
47:54
SPEAKER_00: make me a slider for controlling the animation variables on
48:03
SPEAKER_00: the constellation experiment.
48:06
SPEAKER_00: Correct.
48:06
SPEAKER_00: So we'll let that work.
48:07
SPEAKER_00: But that's the kind of, it's like really simple.
48:09
SPEAKER_00: If you tell it like make me controls or make me a slider
48:12
SPEAKER_00: for whatever it is you're trying to do,
48:13
SPEAKER_00: you just think about like,
48:14
SPEAKER_00: okay, what are the ways I need to like tweak this to make it work?
48:17
SPEAKER_00: It's pretty good about just making them.
48:18
SPEAKER_00: I mean, I have a skill, so it like knows how to like make them so that they look decent, right?
48:22
SPEAKER_00: Right.
48:23
SPEAKER_01: And based on your preferences, you know, what you figured out works.
48:27
SPEAKER_00: Yeah.
48:27
SPEAKER_00: Yeah.
48:28
SPEAKER_00: But then it's great.
48:28
SPEAKER_00: Just hooks it up and then you can just kind of like play with it.
48:31
SPEAKER_00: And it's a whole different way to design.
48:32
SPEAKER_00: It's very, you know, Brett Victor, like the live programming stuff.
48:37
SPEAKER_00: Brett Victor was, he's an engineer, also a designer.
48:41
SPEAKER_00: I mean, he's off doing some crazy stuff now.
48:43
SPEAKER_00: He did a set of talks.
48:45
SPEAKER_00: Between, I don't know, it was like 2010, 2013, one of them's like, stop drawing dead fish.
48:50
SPEAKER_00: And it's about how like programming is not a very live medium.
48:53
SPEAKER_00: Like, you know,
48:54
SPEAKER_00: you code in the editor,
48:55
SPEAKER_00: you do your whole build,
48:56
SPEAKER_00: and then you like look at it in the browser or wherever it is.
48:58
SPEAKER_00: And these two things feel very disconnected.
49:00
SPEAKER_00: And it's very hard to like,
49:01
SPEAKER_00: You're tweaking variables in code and then you have to go see the effect
49:04
SPEAKER_00: and it's not a direct connection.
49:06
SPEAKER_00: And so his whole thing was like you need to have instant direct feedback
49:09
SPEAKER_00: at all times to be able to like look at the artifact you're making
49:12
SPEAKER_00: and directly tweak it.
49:14
SPEAKER_00: And it's like this kind of stuff come to life.
49:16
SPEAKER_00: Like now we finally can do it.
49:17
SPEAKER_00: But before you couldn't do that.
49:19
SPEAKER_00: There was like no programming system in existence that could give you like
49:22
SPEAKER_00: a live preview of every possible variable or thing.
49:26
SPEAKER_00: Like the closest we have is like dev tools in the browser.
49:28
SPEAKER_01: Well, another topic, which is very related to this one, while we're waiting for it.
49:32
SPEAKER_01: One thing that with AI,
49:33
SPEAKER_01: of course,
49:33
SPEAKER_01: you're saying it's easier to build stuff, but with AI, it's also easier to design.
49:37
SPEAKER_01: When I was asking people how they work with designers,
49:40
SPEAKER_01: an engineer replied to something,
49:42
SPEAKER_01: which I hear a lot more,
49:43
SPEAKER_01: which is,
49:43
SPEAKER_01: I'll quote this person, Amir,
49:46
SPEAKER_01: As an engineer,
49:46
SPEAKER_01: I asked Cloud Design to generate 20 high fidelity alternatives and I keep iterating
49:51
SPEAKER_01: until I landed a perfect design.
49:52
SPEAKER_01: It's very enjoyable and feels much faster than interacting with the design team.
49:56
SPEAKER_01: So it is a thing saying, oh, I have these tools.
49:59
SPEAKER_01: I don't necessarily need a designer.
50:00
SPEAKER_01: I can just ask for all these.
50:03
SPEAKER_01: And in this case, it will just give me 20 alternatives.
50:05
SPEAKER_01: I'll choose the perfect one.
50:06
SPEAKER_01: And boom, I'm done.
50:07
SPEAKER_01: As someone who is a designer, what's your take on this?
50:11
SPEAKER_01: And by the way, this is not the first time.
50:13
SPEAKER_01: I think there's always the thing like, do we need designers?
50:16
SPEAKER_01: As engineers, do we need product managers?
50:17
SPEAKER_01: And of course, product managers will also ask, do we need engineers?
50:20
SPEAKER_01: And so on.
50:21
SPEAKER_01: But what do you think?
50:23
SPEAKER_01: Where could this be valid?
50:24
SPEAKER_01: What could people be missing?
50:26
SPEAKER_00: I think in cases where you don't have a designer to hand and you're trying
50:29
SPEAKER_00: to like validate a product or like prove a hypothesis,
50:32
SPEAKER_00: but you need an interface for it,
50:33
SPEAKER_00: it's great to just use a model to be like, sure, make me 20 high fidelity designs.
50:37
SPEAKER_00: I'm sure I might look at those designs and be like,
50:40
SPEAKER_00: wow, like it all depends on taste too, right?
50:42
SPEAKER_00: Like I might look at them and think, okay, well,
50:44
SPEAKER_00: these look obviously generated by AI to me and I don't know that they're going to do
50:48
SPEAKER_00: the job that they need to do it.
50:50
SPEAKER_00: Again, it depends on the context.
50:51
SPEAKER_00: If it's something simple, needs a button and a sidebar, fine.
50:54
SPEAKER_00: But if it's like,
50:55
SPEAKER_00: you know,
50:55
SPEAKER_00: we have some new primitive we're trying to figure out the shape of,
50:58
SPEAKER_00: that's when you really need a designer to come in.
51:00
SPEAKER_00: And like, really,
51:01
SPEAKER_00: it's just someone who's assigned to like think through the problem properly,
51:04
SPEAKER_00: like put in the brain work and be like, okay,
51:07
SPEAKER_00: do the experiments,
51:08
SPEAKER_00: like show them to users, like figure out what people actually understand and don't.
51:12
SPEAKER_00: And that's really like the labor a designer should come in and do.
51:15
SPEAKER_00: But I think it's completely fine for developers without access to one
51:18
SPEAKER_00: to like use models as much as they can.
51:20
SPEAKER_00: But when models do designs for me,
51:23
SPEAKER_00: I, of course, just look at them and think that is terrible quality.
51:26
SPEAKER_00: That is awful.
51:27
SPEAKER_00: Even to the extent of like,
51:29
SPEAKER_00: I think they've been prompted
51:30
SPEAKER_00: with some what we would call like universal design principles.
51:33
SPEAKER_00: But they sort of don't understand nuance and context.
51:37
SPEAKER_00: And then I think they'd like stick to them too strictly.
51:40
SPEAKER_00: So like I find the agents want to put a label on everything in the interface.
51:44
SPEAKER_01: Label meaning?
51:46
SPEAKER_00: Like a small bit of text.
51:47
SPEAKER_00: Like there might be like a little button that's like to close the sidebar.
51:49
SPEAKER_00: And we would usually use an icon button for that because most people are trained
51:52
SPEAKER_00: that this little button near the sidebar means it'll close it.
51:55
SPEAKER_00: And they'll experiment with it and they'll click it and they'll figure out
51:57
SPEAKER_00: that is correctly what it does.
51:58
SPEAKER_00: But the agent will write close sidebar.
52:01
SPEAKER_00: Or close modal in the top right hand of the modal.
52:03
SPEAKER_00: And you're like, there's now a lot of text on this page.
52:08
SPEAKER_00: And they'll just like put like four lines of instructional text over a button
52:11
SPEAKER_00: and just things where you're like,
52:13
SPEAKER_00: I understand in the model's mind,
52:15
SPEAKER_00: it's thinking, oh,
52:16
SPEAKER_00: this is how we like make the interface explainable to the human,
52:19
SPEAKER_00: you know, but actually it's really bad design.
52:22
SPEAKER_01: Now, talking about the UX and UI for AI.
52:27
SPEAKER_01: What do you think UX and good UX and UI looks like?
52:31
SPEAKER_01: And I know this is a bigger question,
52:33
SPEAKER_01: but so far, what have you learned of what works and what doesn't?
52:37
SPEAKER_00: I think when it comes to this, like,
52:39
SPEAKER_00: what is the ideal UI of AI,
52:41
SPEAKER_00: which is like a big question I don't know that we'll figure out the answer to for a
52:44
SPEAKER_00: while.
52:44
SPEAKER_00: But it's something like some things haven't changed,
52:47
SPEAKER_00: like in terms of what is good interface design for software.
52:52
SPEAKER_00: There's still a lot of principles here.
52:54
SPEAKER_00: What formats do we have to show users data?
52:57
SPEAKER_00: It's like dashboards and sidebars and documents.
52:59
SPEAKER_00: There's a lot of things that haven't changed.
53:01
SPEAKER_00: We actually had a good story about this from Illicit where
53:05
SPEAKER_00: the app was helping scientists extract data from papers.
53:08
SPEAKER_00: And they were very used to doing this in Excel spreadsheets.
53:10
SPEAKER_00: They'd get a big Excel spreadsheet,
53:12
SPEAKER_00: put the papers in,
53:13
SPEAKER_00: one on each row, and then they'd extract each piece of data into the cells, right?
53:16
SPEAKER_00: Pretty standard process.
53:18
SPEAKER_00: But of course, in the beginning of Elicit, we were like, oh, that's so old school.
53:21
SPEAKER_00: There must be some much better interface for seeing the data extracted from papers.
53:27
SPEAKER_00: So we tried all this crazy stuff.
53:28
SPEAKER_00: There was infinite canvases with cards spread everywhere.
53:31
SPEAKER_00: There was one that's a bunch of cards that are all linearly stacked.
53:35
SPEAKER_00: We were like, maybe it's more like Notion with these composable documents with rich interfaces.
53:38
SPEAKER_00: We tried all this stuff.
53:40
SPEAKER_00: And every user interview I did, users were just like, this is very confusing.
53:43
SPEAKER_00: Can I just have a table?
53:45
SPEAKER_00: Like every time.
53:46
SPEAKER_00: And so after a couple months of this like, oh, what's the new UI of AI?
53:52
SPEAKER_00: We were like, oh, it's a table.
53:53
SPEAKER_00: Well, at least in our use case,
53:56
SPEAKER_00: it turned out the interface people were using was the best because it was
54:00
SPEAKER_00: the most familiar to them and it caused the least amount of like cognitive load
54:03
SPEAKER_00: for them to use the tool.
54:04
SPEAKER_00: So we were like, right.
54:05
SPEAKER_00: Back to tables.
54:06
SPEAKER_00: That's fine.
54:07
SPEAKER_00: Like it was good to go on that journey.
54:09
SPEAKER_00: But sometimes we have this notion of like the UI of AI will be so wildly different
54:13
SPEAKER_00: from our current imaginings.
54:15
SPEAKER_00: And sometimes it's really not at all.
54:17
SPEAKER_00: It's like you can do more powerful things.
54:18
SPEAKER_00: I think there's lots of leverage,
54:19
SPEAKER_00: but I expect it to all be built with documents and sidebars and cards and tables
54:24
SPEAKER_00: and like all the all the classic things we've worked out do work and people know how
54:28
SPEAKER_00: to use.
54:29
SPEAKER_01: Yeah, interesting.
54:30
SPEAKER_01: There's a cognitive load in learning any new UI paradigm.
54:34
SPEAKER_01: It can be confusing if your existing tools don't work like that because, again,
54:38
SPEAKER_01: people will still use operating systems and Google Sheets or Excel or Office Docs.
54:44
SPEAKER_01: And even if like a small subset of people are like,
54:48
SPEAKER_01: I don't know,
54:48
SPEAKER_01: familiar with this new UI stuff,
54:50
SPEAKER_01: you remember like probably like parents or grandparents teaching them how to use
54:55
SPEAKER_01: the phones, the interfaces.
54:56
SPEAKER_01: And once they learned it, they're good.
54:58
SPEAKER_01: But then do you really want to do that again?
55:00
SPEAKER_00: Yeah, no.
55:00
SPEAKER_01: It's interesting.
55:02
SPEAKER_01: There's a cost.
55:03
SPEAKER_01: There will be something similar-ish.
55:05
SPEAKER_00: Yeah, it's kind of like how we started with chatbots, right?
55:07
SPEAKER_00: It's like chatbots were not the final form.
55:08
SPEAKER_00: Like if we could look at codecs, like there's a lot going on here.
55:11
SPEAKER_00: Sure, there's a chat window here.
55:13
SPEAKER_00: But there's all kinds of things going on with Git.
55:14
SPEAKER_00: Work trees over here,
55:16
SPEAKER_00: and then we've got this whole interface, and I can do annotations and debug in this.
55:20
SPEAKER_00: There's a lot of extra stuff where you take the familiar primitive
55:23
SPEAKER_00: and you expand outwards from it.
55:25
SPEAKER_00: And we did similar things at Elicit.
55:26
SPEAKER_00: There was still a table, but then we baked a lot of stuff into the table and around it.
55:31
SPEAKER_00: But you start with familiar primitives.
55:33
SPEAKER_00: So we have our jig ready.
55:36
SPEAKER_00: Well, yeah, let's expand this.
55:37
SPEAKER_00: So here's a jig.
55:38
SPEAKER_00: This is using Dial Kit.
55:40
SPEAKER_00: This is built with this.
55:41
SPEAKER_00: I can change the speed of the animation to see like, oh, is that too fast?
55:45
SPEAKER_00: Like, is that too slow?
55:47
SPEAKER_00: The intensity is like how much gravity is being applied to the stars here.
55:52
SPEAKER_00: How much variation is there in the stars?
55:53
SPEAKER_00: This is sizing and scrolls.
55:55
SPEAKER_00: This is like when I scroll down to have this zoom out.
55:57
SPEAKER_00: Like what's the resting height?
55:59
SPEAKER_00: I think this is of the... And this is here, like the resting height of this.
56:04
SPEAKER_00: Resting logo size.
56:05
SPEAKER_00: Let's see, what's this?
56:06
SPEAKER_00: I have some glass effects on the surface that it's like... Oh, it's on the...
56:10
SPEAKER_00: On the top part, yes.
56:12
SPEAKER_00: How much is that glass effect?
56:14
SPEAKER_00: Like, let's see, that's too bright, then you can't read the words.
56:17
SPEAKER_00: All the way down here, you can barely tell it's glass.
56:19
SPEAKER_00: There's some sort of like happy middle ground you want to figure out.
56:22
SPEAKER_00: Like, how much contrast between the background...
56:25
SPEAKER_01: You weren't kidding that this is like your own almost personal Figma,
56:29
SPEAKER_01: if you will, or like design tool.
56:31
SPEAKER_00: Yeah.
56:32
SPEAKER_00: And because it's really difficult to mock up static versions of this in Figma,
56:35
SPEAKER_00: which I have done.
56:36
SPEAKER_00: Yeah.
56:36
SPEAKER_00: Like I was for a minute,
56:37
SPEAKER_00: we were thinking maybe like a gray background,
56:40
SPEAKER_00: trying to figure out how much like grain and stuff, maybe like a wavy one.
56:43
SPEAKER_00: Yeah.
56:44
SPEAKER_00: And then I'm trying to figure out here, like, OK, is it like halfway off the screen?
56:47
SPEAKER_00: Maybe there's like a slice out of it here.
56:49
SPEAKER_00: So you can play around with layouts and ideas,
56:52
SPEAKER_00: but you can't get a sense until it's live in the browser of how this is going
56:57
SPEAKER_00: to feel and how much scroll you need to do here.
57:00
SPEAKER_00: And then the themes and the colors and textures and everything.
57:04
SPEAKER_00: So I work with this a lot now.
57:06
SPEAKER_00: Whenever I'm designing something visual,
57:08
SPEAKER_00: it's so much easier to have live variables,
57:11
SPEAKER_00: which again, would not have been possible before.
57:12
SPEAKER_00: You could have built this by hand, I guess, in the olden days.
57:15
SPEAKER_01: It just would have been practical.
57:16
SPEAKER_00: Just coding the thing itself was slow enough.
57:19
SPEAKER_01: One interesting thing you talked about is capability gaslighting.
57:23
SPEAKER_01: What is capability gaslighting?
57:25
SPEAKER_00: Okay, so I think this is similar to like, Ethan Mollick has this phrase, the jagged frontier.
57:29
SPEAKER_00: Yeah.
57:30
SPEAKER_00: It's pretty much that, but I came up with capabilities gaslighting before I read his work.
57:34
SPEAKER_00: But so that people are familiar with the concept,
57:36
SPEAKER_00: it's that models are really, really good at some things and really bad at others.
57:40
SPEAKER_00: And it's really hard to predict when you give it a certain task,
57:43
SPEAKER_00: which of those it's going to fall into.
57:45
SPEAKER_00: And then capabilities gaslighting was the feeling I got early on from using them
57:48
SPEAKER_00: where like they convince you they're so capable because they'll really impress you
57:52
SPEAKER_00: on one task.
57:53
SPEAKER_00: And then you try them on something else and like they fail and you kind
57:55
SPEAKER_00: of feel like,
57:55
SPEAKER_00: I feel like you kind of gaslit me,
57:56
SPEAKER_00: like imagining that you're this extremely intelligent person.
57:59
SPEAKER_00: So it's such an inconsistent experience.
58:01
SPEAKER_00: And it's like, it's so different to working with a human.
58:03
SPEAKER_00: Like if you find a human who you feel is like really an expert in a topic,
58:22
SPEAKER_00: And you work with them.
58:23
SPEAKER_00: It's very unusual for them to be inconsistent in their performance, right?
58:27
SPEAKER_00: Like it's very rare for them to suddenly forget all of their like expert knowledge on a topic.
58:32
SPEAKER_00: And if they did, you would be very like, are you having like a mental breakdown?
58:35
SPEAKER_00: Like this is very weird behavior.
58:37
SPEAKER_00: But this is how models behave every day.
58:40
SPEAKER_00: Like one day they will actually perform really well on a task and they could fail at
58:43
SPEAKER_00: that same task the next day because you prompted differently
58:45
SPEAKER_00: or they had different contexts available or like random whatever it is,
58:48
SPEAKER_00: like stochastic outputs they just didn't do as well.
58:52
SPEAKER_00: So I think it's hard to work with them because you never quite know what you're going to get.
58:55
SPEAKER_01: Yeah, I guess we just need to keep this in mind, right?
58:59
SPEAKER_01: And this is ongoing.
59:00
SPEAKER_01: Like, of course,
59:01
SPEAKER_01: we always say and we have seen that the models improve,
59:04
SPEAKER_01: but you want to be skeptical.
59:06
SPEAKER_00: Yeah, yeah.
59:07
SPEAKER_00: Of course,
59:07
SPEAKER_00: there's some baseline with Frontier models,
59:09
SPEAKER_00: like they're not going to perform below a certain bit.
59:11
SPEAKER_00: But, you know,
59:12
SPEAKER_00: of course,
59:12
SPEAKER_00: the models are changing all the time and people complain about this,
59:15
SPEAKER_00: that you upgrade to a new model in it and it changes what it's good at
59:17
SPEAKER_00: or it changes like it doesn't perform to your expectations of certain things.
59:21
SPEAKER_00: It's like it's hard to predict.
59:23
SPEAKER_01: You did a talk titled,
59:25
SPEAKER_01: One Developer,
59:26
SPEAKER_01: Two Dozen Agents, Zero Alignment, Why We Need Collaborative Engineering.
59:31
SPEAKER_01: Can you talk about just this observation, like one developer, two dozen agents, zero alignment?
59:36
SPEAKER_00: Yeah.
59:36
SPEAKER_00: So this is like a problem that the GitHub Next team has been focused on trying
59:40
SPEAKER_00: to find various ways to solve.
59:42
SPEAKER_00: I think for well over a year now is this acknowledgement that we are now working
59:46
SPEAKER_00: with these agents locally on machines and it speeds up each individual person.
59:49
SPEAKER_00: Like us alone with the agent, we can go really fast.
59:53
SPEAKER_00: Software is always built on a team, right?
59:55
SPEAKER_00: You're always trying to align with like the product managers and the designers and
59:58
SPEAKER_00: the other engineers, like what decisions you're all making.
1:00:00
SPEAKER_00: And we don't actually have good tools in place to do that.
1:00:02
SPEAKER_00: It's sort of like we have Slack.
1:00:04
SPEAKER_00: People might have something like Linear or GitHub issues where they're like doing issues.
1:00:07
SPEAKER_00: But there's tons of pre-planning work before you write an issue,
1:00:10
SPEAKER_00: because usually when you write an issue,
1:00:11
SPEAKER_00: you're like ready to hand it off to an agent.
1:00:13
SPEAKER_00: But before that point,
1:00:14
SPEAKER_00: you have to have agreed upon your approach and like, should we build this feature?
1:00:18
SPEAKER_00: Is this feature the right shape?
1:00:19
SPEAKER_00: Does it have the right interface?
1:00:20
SPEAKER_00: Like, Do we have some sort of database migration we have to do?
1:00:23
SPEAKER_00: There's all this like upfront work and not many good tools to do that work in is how we felt.
1:00:29
SPEAKER_00: Especially not ones with agents involved in them.
1:00:31
SPEAKER_00: So the biggest gap seems to be like,
1:00:33
SPEAKER_00: we have agentic coding tools, but none of them are real-time multiplayer.
1:00:36
SPEAKER_00: All your sessions are private and shared to you,
1:00:38
SPEAKER_00: often locally on your machine, so you couldn't possibly share it with someone.
1:00:42
SPEAKER_00: This is starting to change now.
1:00:43
SPEAKER_00: We have seen some products come out that are trying
1:00:46
SPEAKER_00: to do live multiplayer agent work,
1:00:47
SPEAKER_00: like Buzz from Jack Dorsey and Ace,
1:00:50
SPEAKER_00: the prototype that we made at GitHub Next was in this direction,
1:00:52
SPEAKER_00: was like shared compute, shared sandboxes in a Slack-like interface.
1:00:56
SPEAKER_00: So you're like coding and talking at the same time.
1:00:58
SPEAKER_00: So people I think are beginning to realize this is like the next big thing we need
1:01:02
SPEAKER_00: to push on to improve in like our agentic coding tools.
1:01:05
SPEAKER_00: But it's still totally unsolved.
1:01:06
SPEAKER_00: Like even once you have a Slack interface and there's an agent in there,
1:01:10
SPEAKER_00: you still need tools to agree on things.
1:01:12
SPEAKER_00: This is where I'm kind of pushing on like maybe decisions need to become sort
1:01:15
SPEAKER_00: of first class primitive.
1:01:16
SPEAKER_00: Like when an agent presents a decision to a human.
1:01:20
SPEAKER_00: You first need to have the decision be like large enough on screen to give you
1:01:24
SPEAKER_00: the information about it.
1:01:25
SPEAKER_00: But then also you probably need someone else on your team to come like help you make
1:01:28
SPEAKER_00: that decision or at least like give input on it.
1:01:31
SPEAKER_00: And then someone has to be responsible for having had made that decision.
1:01:34
SPEAKER_00: You need like a record of like,
1:01:35
SPEAKER_00: here's the things we decided,
1:01:36
SPEAKER_00: like here's the context we're going to feed to the agent.
1:01:39
SPEAKER_00: You know,
1:01:39
SPEAKER_00: there's some moment where we've written like a clear set of issues or specs
1:01:43
SPEAKER_00: that we're going to hand off to be implemented.
1:01:45
SPEAKER_00: And then some verification check on the other side.
1:01:47
SPEAKER_00: But you all need to be so aligned up to the point of implementation because in
1:01:51
SPEAKER_00: the old world, implementation took so long you could like adjust along the way.
1:01:55
SPEAKER_00: But now because there's a sort of hard handover point to an agent,
1:01:58
SPEAKER_00: you need to all be aligned up front in a way that we're not at the moment.
1:02:01
SPEAKER_00: Like even our team, we feel this all the time.
1:02:03
SPEAKER_00: It's like we're all individually running on our machines and it's hard
1:02:06
SPEAKER_00: to stay like aligned on what we're doing.
1:02:08
SPEAKER_01: And you mentioned GitHub Next and the prototyping that you're doing.
1:02:11
SPEAKER_01: Can you talk a little bit about the types of projects you've built,
1:02:15
SPEAKER_01: that you've experimented, some learnings that you've had?
1:02:19
SPEAKER_01: Maybe give us a direction of things you're now excited about exploring.
1:02:24
SPEAKER_00: Yeah, yeah.
1:02:25
SPEAKER_00: So I joined the team a little less than a year ago,
1:02:28
SPEAKER_00: but most of the time we were working on this prototype ACE,
1:02:30
SPEAKER_00: which is Slack,
1:02:33
SPEAKER_00: Plus cloud compute,
1:02:34
SPEAKER_00: like sandboxes and micro VMs and a bunch of other stuff in there.
1:02:37
SPEAKER_00: Like you can like open PRs and review your code and that kind of stuff.
1:02:41
SPEAKER_00: So it was a kind of an all in one multiplayer workspace.
1:02:44
SPEAKER_00: And it was a really like useful prototype,
1:02:47
SPEAKER_00: but it turned out to be extremely ambitious for our small team.
1:02:49
SPEAKER_00: We had like three or four people working on it at a time.
1:02:52
SPEAKER_00: And we wanted at some point to take it to like a real product,
1:02:55
SPEAKER_00: but it just turned out that was like not feasible with the number of people we had.
1:02:59
SPEAKER_00: But we're trying to take bits of it and ship it to the rest of GitHub.
1:03:02
SPEAKER_00: Sandboxes have now gone out to be part of the GitHub desktop app.
1:03:05
SPEAKER_00: That's being used in other ways.
1:03:07
SPEAKER_00: And I think it's also made leadership take the multiplayer thing more seriously
1:03:11
SPEAKER_00: and try and find other ways we might ship that to the product.
1:03:13
SPEAKER_01: This is a really promising direction.
1:03:14
SPEAKER_01: I mean, we already have a deep dive out by the time this podcast is out about Ramp
1:03:19
SPEAKER_01: who have built just this inside and they made it collaborative.
1:03:23
SPEAKER_01: It's forced collaborative.
1:03:25
SPEAKER_01: You cannot make anything private.
1:03:26
SPEAKER_01: Oh, yeah, we took that.
1:03:28
SPEAKER_01: They have multiple interfaces.
1:03:29
SPEAKER_01: They have this running.
1:03:30
SPEAKER_01: It's called Inspect and it runs in Slack.
1:03:32
SPEAKER_01: There's a Chrome plugin as well.
1:03:34
SPEAKER_01: There's a web interface.
1:03:36
SPEAKER_01: And they did find that in the Slack channels with product managers,
1:03:40
SPEAKER_01: designers, they can talk, the agent inspect gets all this input.
1:03:44
SPEAKER_01: And then all the sessions,
1:03:46
SPEAKER_01: anyone can join a session and anyone can prompt and change the direction,
1:03:50
SPEAKER_01: which actually was a really...
1:03:52
SPEAKER_01: They weren't sure if they want this because people previously,
1:03:56
SPEAKER_01: but they were like, well, it's just civilized things.
1:03:59
SPEAKER_01: So it turned out to not be an issue.
1:04:01
SPEAKER_01: But because of this, it spreads better.
1:04:03
SPEAKER_01: And it has all this feedback loops.
1:04:05
SPEAKER_01: And the reason it works really well for them is it's integrated this agent to all
1:04:09
SPEAKER_01: of their internal systems.
1:04:10
SPEAKER_01: And their cloud machines are like a full-blown developer machine,
1:04:13
SPEAKER_01: which is a lot of work,
1:04:14
SPEAKER_01: a bit like cloud development environment, which is all this to say, like doing this.
1:04:18
SPEAKER_01: For one company,
1:04:19
SPEAKER_01: specifically for them it's possible,
1:04:21
SPEAKER_01: doing it as a generic solution is probably more difficult.
1:04:24
SPEAKER_01: I'm sure it will come, or I'm sure there will be attempts.
1:04:28
SPEAKER_01: But this is the difference between scratching your own itch and doing something
1:04:32
SPEAKER_01: that works for so many people.
1:04:35
SPEAKER_00: Slack launched their whole developer experience IDE thing,
1:04:38
SPEAKER_00: which they might be able to make this experience happen.
1:04:41
SPEAKER_00: Buzz from Jack Dorsey is going in this direction too.
1:04:44
SPEAKER_00: There's clearly lots of people realizing this is a problem.
1:04:47
SPEAKER_01: And you're also experimenting in this direction, clearly.
1:04:50
SPEAKER_00: Yeah, yeah, yeah.
1:04:50
SPEAKER_00: So one of the things, like,
1:04:52
SPEAKER_00: when we tried to,
1:04:52
SPEAKER_00: like, build this whole thing, it was too ambitious to do it with all the micro VMs.
1:04:55
SPEAKER_00: But now we're kind of scaling back and thinking, okay,
1:04:58
SPEAKER_00: what bits of this can we prototype in ways that would help GitHub leadership,
1:05:01
SPEAKER_00: like, have...
1:05:02
SPEAKER_00: Have a bit more direction of like where they should run next.
1:05:04
SPEAKER_00: Like, you know,
1:05:05
SPEAKER_00: one of our functions of our team is just to run out ahead of product and say, okay,
1:05:07
SPEAKER_00: you guys know what like the things you should probably do in the next six months
1:05:11
SPEAKER_00: to one year is.
1:05:12
SPEAKER_00: We are much more like what is a big crazy swing GitHub could make
1:05:15
SPEAKER_00: or what's more risky or what's like the far future.
1:05:18
SPEAKER_00: And so that's why we're trying to prototype.
1:05:20
SPEAKER_00: Okay.
1:05:20
SPEAKER_00: People are now on this multiplayer like sessions thing.
1:05:23
SPEAKER_00: What comes after that?
1:05:25
SPEAKER_00: So now we're interested in like,
1:05:26
SPEAKER_00: how do you make proactive agents that aren't annoying is one of
1:05:28
SPEAKER_00: the things we're trying to explore.
1:05:29
SPEAKER_00: Yeah.
1:05:30
SPEAKER_00: Because proactive agents,
1:05:31
SPEAKER_00: you think we have this dream of like, okay,
1:05:34
SPEAKER_00: assume intelligence is like free or so cheap that like it doesn't matter, right?
1:05:38
SPEAKER_00: That's a fun assumption to make.
1:05:40
SPEAKER_00: If you could have 100 background agents running,
1:05:43
SPEAKER_00: what would be the least overwhelming,
1:05:44
SPEAKER_00: least annoying way to consume any of their outputs is like
1:05:47
SPEAKER_00: an open research question we're trying to pick.
1:05:49
SPEAKER_00: How would you have agents working in a document with humans where it's not, again,
1:05:53
SPEAKER_00: overwhelming or annoying for them to be making changes
1:05:55
SPEAKER_00: or doing helpful background work in a way that doesn't feel like invasive
1:05:58
SPEAKER_00: or interruptive?
1:05:59
SPEAKER_00: So we try and take these kind of design questions and figure out prototypes
1:06:03
SPEAKER_00: that could solve them in the hopes that like,
1:06:05
SPEAKER_00: you know,
1:06:05
SPEAKER_00: if one day GitHub wanted to build something in this direction,
1:06:08
SPEAKER_00: we can then be like, hey, here's all the research we did.
1:06:10
SPEAKER_01: A question I wanted to ask you is...
1:06:14
SPEAKER_01: It's coming from a developer angle.
1:06:15
SPEAKER_01: It's what happens when,
1:06:17
SPEAKER_01: in your experience,
1:06:20
SPEAKER_01: observations, when we start to give more and more decisions to the agent.
1:06:24
SPEAKER_01: And I'm going to quote a developer,
1:06:27
SPEAKER_01: Jorge Manrubia, who wrote, He wrote an article called Oh My Craft.
1:06:31
SPEAKER_01: He wrote, the need to intervene on the small stuff is decaying quickly as models improve.
1:06:35
SPEAKER_01: And when models are capable enough to handle those details themselves,
1:06:39
SPEAKER_01: spending too much time on these details starts to feel like a poor use
1:06:42
SPEAKER_01: of human attention.
1:06:43
SPEAKER_01: And he writes about how he's not opening it.
1:06:45
SPEAKER_01: He rarely opens his code editor anymore.
1:06:47
SPEAKER_01: It's been months he wrote a single line itself.
1:06:50
SPEAKER_01: But the thing is, he used to live inside and see all the details.
1:06:53
SPEAKER_01: And there is a sense that craft has to do with being there with the details.
1:06:59
SPEAKER_01: And of course,
1:07:00
SPEAKER_01: you're coming from a design perspective,
1:07:03
SPEAKER_01: but what have you observed of your own craft changing,
1:07:07
SPEAKER_01: decaying, or what you're seeing?
1:07:10
SPEAKER_01: Do you think there's a danger as there's this temptation as we've talked about,
1:07:15
SPEAKER_01: like, just good, good, good, make this decision.
1:07:18
SPEAKER_01: We just did this with the slider, which didn't matter.
1:07:20
SPEAKER_00: Yeah, but it can matter.
1:07:22
SPEAKER_00: But it can matter.
1:07:23
SPEAKER_00: I think it doesn't matter.
1:07:24
SPEAKER_00: I mean, it's hard because I'm not like I would say like a true engineer in the sense
1:07:29
SPEAKER_00: of like I don't necessarily care about clean code that much.
1:07:31
SPEAKER_01: But you care about great design.
1:07:33
SPEAKER_00: But I do care about design.
1:07:34
SPEAKER_00: But the thing is, they definitely can't do design to my standards yet.
1:07:39
SPEAKER_00: I have to still be really involved to get the design to look and feel the way I would make it.
1:07:44
SPEAKER_00: So I'm annoyed almost that I'm always in there being like, no, that is a terrible transition.
1:07:49
SPEAKER_00: Here's how we should do it, you know, like this.
1:07:51
SPEAKER_01: So you're into details.
1:07:52
SPEAKER_01: You're not letting go of those details.
1:07:54
SPEAKER_00: Yeah, because I wish I could.
1:07:55
SPEAKER_00: I wish.
1:07:56
SPEAKER_00: I mean, I've written.
1:07:57
SPEAKER_01: Do you?
1:07:57
SPEAKER_00: Design.
1:07:58
SPEAKER_00: Well, I don't know.
1:07:59
SPEAKER_00: I want to push you on this.
1:08:00
SPEAKER_00: Yeah.
1:08:01
SPEAKER_00: What I keep thinking this,
1:08:03
SPEAKER_00: like if tomorrow I keep trying to write design skills that tell
1:08:05
SPEAKER_00: the agents exactly my design preferences, which won't be universal.
1:08:08
SPEAKER_00: Like I have preferences about how much.
1:08:10
SPEAKER_00: Padding I like between a border and an element, right?
1:08:13
SPEAKER_00: And that's not everyone's preference and it's not right for every product.
1:08:15
SPEAKER_00: But for me, I'm like, there are set rules.
1:08:17
SPEAKER_00: Is this much padding?
1:08:17
SPEAKER_00: Is this kind of border shadow?
1:08:20
SPEAKER_00: Like, it's pretty standard.
1:08:22
SPEAKER_00: And I've tried to write these rules,
1:08:23
SPEAKER_00: but then they don't universally apply them properly and it doesn't work.
1:08:27
SPEAKER_00: And so I always end up in there changing specific values that you would think
1:08:31
SPEAKER_00: by this point in time would have been automated away the way that we all talk about
1:08:34
SPEAKER_00: that agents are like taking over software you would think this would be like
1:08:37
SPEAKER_00: so trivial to like for them to implement like a certain kind
1:08:40
SPEAKER_00: of whatever it is opacity level on a border.
1:08:44
SPEAKER_00: So I kind of think like if I woke up tomorrow and the skill just worked or like
1:08:47
SPEAKER_00: the agents just had enough context and they just designed exactly to my specs,
1:08:51
SPEAKER_00: I wonder if I would be like, oh, all the fun bit is gone.
1:08:54
SPEAKER_00: Like, is it satisfying to have like a gorgeous inner face appear in front of me,
1:08:58
SPEAKER_00: but I didn't do anything to make it?
1:09:01
SPEAKER_00: I'm not sure.
1:09:01
SPEAKER_00: I also think about this with like...
1:09:04
SPEAKER_00: Interior design stuff.
1:09:05
SPEAKER_00: Have you been on Pinterest any time in the last two years?
1:09:08
SPEAKER_01: I'm familiar with it.
1:09:09
SPEAKER_01: And my wife is.
1:09:11
SPEAKER_01: And when we're decorating a new house, I remember collecting these things.
1:09:17
SPEAKER_01: We work with an interior designer.
1:09:20
SPEAKER_01: My wife collected a lot of things, but I also was there.
1:09:22
SPEAKER_01: And you have these kind of like, we're kind of thinking like this, but not quite.
1:09:26
SPEAKER_01: But like when you mix this,
1:09:28
SPEAKER_01: trying to explain,
1:09:29
SPEAKER_01: you know, like we didn't have the tools beyond like we just take screenshots.
1:09:31
SPEAKER_01: So yes.
1:09:32
SPEAKER_01: Yes.
1:09:32
SPEAKER_01: Yeah.
1:09:33
SPEAKER_01: Familiar in OpenPrintist, I guess.
1:09:34
SPEAKER_00: You know, you know, yes.
1:09:36
SPEAKER_00: A lot of it is AI now, if you get on there.
1:09:38
SPEAKER_01: Oh, really?
1:09:38
SPEAKER_00: There's gorgeous rooms.
1:09:40
SPEAKER_00: If you search for certain things,
1:09:41
SPEAKER_00: but you look and you can kind of start to see like, oh, that's for sure AI.
1:09:44
SPEAKER_00: That's for sure AI.
1:09:45
SPEAKER_00: Which is definitely problematic when there are images of a physical space,
1:09:48
SPEAKER_00: because then you go like,
1:09:49
SPEAKER_00: well, the light might not be accurate or this is all just fake.
1:09:53
SPEAKER_00: And they look beautiful,
1:09:54
SPEAKER_00: but you kind of think like,
1:09:55
SPEAKER_00: well, someone didn't make that room and that's not a real room that exists.
1:09:58
SPEAKER_00: So it's almost irrelevant to my interior design.
1:10:00
SPEAKER_00: Yeah.
1:10:00
SPEAKER_00: And I kind of wonder if it's like this with interfaces
1:10:02
SPEAKER_00: where like it is all objectively like looks gorgeous because it's been trained.
1:10:06
SPEAKER_00: It's like very much like everything that was on the web that we've,
1:10:08
SPEAKER_00: you know, clicked thumbs up on.
1:10:10
SPEAKER_00: But with interfaces,
1:10:11
SPEAKER_00: if they all become extremely slick,
1:10:12
SPEAKER_00: like think you take,
1:10:13
SPEAKER_00: you know, what's popular right now, the linear and the Vercel aesthetic, right?
1:10:16
SPEAKER_00: Like very minimalist, very clean, very white, a little bit of rounded corners.
1:10:20
SPEAKER_00: If all the agents can implement that, no problem, just like dead on.
1:10:24
SPEAKER_00: People will start to design in different ways because that will become a tell
1:10:27
SPEAKER_00: that you've just used an agent to do the design.
1:10:29
SPEAKER_00: And if you want to look like you're going to stand out,
1:10:31
SPEAKER_00: you need to have some human do something new and different.
1:10:33
SPEAKER_00: I think it gets into like design,
1:10:36
SPEAKER_00: although it has universal basics, like is the text big enough to read?
1:10:39
SPEAKER_00: Do you have enough space around things?
1:10:41
SPEAKER_00: A lot of it is fashion.
1:10:42
SPEAKER_00: Like it's in fashion right now to look a little bit like linear.
1:10:45
SPEAKER_00: It's not in fashion to look a bit like MySpace, but it was a while ago.
1:10:49
SPEAKER_00: And there'll be in 10 years in the future, if you look like linear, you're very out of date.
1:10:53
SPEAKER_00: You look like some old crafty piece of software and there'll be some new aesthetic
1:10:57
SPEAKER_00: that comes up.
1:10:57
SPEAKER_00: So it's like the models can't necessarily understand that design is like within
1:11:02
SPEAKER_00: a cultural context and that cultural context is always changing.
1:11:05
SPEAKER_00: And it signals different things to people.
1:11:07
SPEAKER_00: And if people read it as like,
1:11:08
SPEAKER_00: oh, they didn't care about this site because it's just like the bland basic code.
1:11:12
SPEAKER_00: Like Claude has a specific design language.
1:11:14
SPEAKER_00: It's very like noticeable, right?
1:11:16
SPEAKER_00: Cream background, like slightly red text, like this eyebrow text on things.
1:11:20
SPEAKER_00: Like you look at it and you're like, yeah, Claude generated that.
1:11:23
SPEAKER_00: No human was involved in this.
1:11:24
SPEAKER_00: Like, I don't know if I'm going to bother looking at it.
1:11:26
SPEAKER_00: So it gets into this, like,
1:11:27
SPEAKER_00: the aesthetic style you pick communicates to the person viewing the product or page.
1:11:33
SPEAKER_00: And things that clearly were made by agents, I think we might start to reject out of hand.
1:11:38
SPEAKER_01: Maybe.
1:11:38
SPEAKER_01: I have this thing where when it's text,
1:11:41
SPEAKER_01: when it's written text, I can immediately tell if it's AI written.
1:11:44
SPEAKER_01: But I can never tell if it's because I write a lot.
1:11:49
SPEAKER_01: I also read a lot.
1:11:50
SPEAKER_01: I read a lot of, you know, like fiction.
1:11:52
SPEAKER_01: And so it just feels off immediately.
1:11:54
SPEAKER_01: And it's not that it's just Claude.
1:11:56
SPEAKER_01: It's just the repetitiveness, the adjectives.
1:11:59
SPEAKER_01: But I can never tell for sure,
1:12:01
SPEAKER_01: like, Am I the one who notices because you said like, oh, you see, it's Claude.
1:12:05
SPEAKER_01: And I'm not sure if I looked at a web page that I would see that it's Claude
1:12:09
SPEAKER_01: because you're you're so into this the same way I'm so into words or
1:12:14
SPEAKER_01: or nonfiction writing.
1:12:15
SPEAKER_00: Yeah.
1:12:16
SPEAKER_01: So I do wonder if if maybe we're having to as an expert, you can always tell.
1:12:22
SPEAKER_01: But the experts are always a small percentage of population.
1:12:24
SPEAKER_01: We don't know.
1:12:25
SPEAKER_01: But but I do agree with you that it is predictable in ways
1:12:28
SPEAKER_01: that humans are are luckily not, I guess.
1:12:30
SPEAKER_00: Yeah, like I definitely noticed the writing one too because I'm the same.
1:12:33
SPEAKER_00: I'm a writer.
1:12:34
SPEAKER_00: I love writing.
1:12:35
SPEAKER_00: And the minute I hit any sentence that is like remotely like,
1:12:38
SPEAKER_00: yeah, just like close tab, never mind.
1:12:41
SPEAKER_00: And then even like going back and reading nonfiction books
1:12:45
SPEAKER_00: that like were really like not written a long time ago.
1:12:47
SPEAKER_00: You could go back and read a book written 50 years ago.
1:12:49
SPEAKER_00: And just the opening is so refreshing.
1:12:51
SPEAKER_00: You're like, oh, these are such novel words.
1:12:53
SPEAKER_00: Like there's none of this like...
1:12:55
SPEAKER_00: If this, then that, or it's not X, it's Y.
1:12:57
SPEAKER_00: There's not a single bit of it to be seen.
1:12:59
SPEAKER_00: You just like breathe in like original human writing and it's so refreshing
1:13:03
SPEAKER_00: that you go, yeah, I can't stand to read anything that's agent contrived.
1:13:07
SPEAKER_01: Speaking of original writing and human writing,
1:13:10
SPEAKER_01: you've talked about the concept of the digital garden.
1:13:12
SPEAKER_01: What is a digital garden?
1:13:14
SPEAKER_00: It's a blog, but with some extra rules attached.
1:13:17
SPEAKER_00: So it's a blog where every piece you put up does not have to be finished as long
1:13:21
SPEAKER_00: as you clearly signal to the audience it's not.
1:13:23
SPEAKER_00: So it's a blog that you grow over time.
1:13:25
SPEAKER_00: So you can put a piece up that's half done.
1:13:27
SPEAKER_00: And then update it later.
1:13:29
SPEAKER_00: You just have like an update of that date.
1:13:30
SPEAKER_00: And I usually will, I communicate it with three different stages my posts go through.
1:13:35
SPEAKER_00: So I have seedlings, budding, and evergreen, like very in the gardening metaphor.
1:13:39
SPEAKER_00: So I can put up something that is like a half finished thought and mark it as
1:13:42
SPEAKER_00: a seedling and then like come back to it later and finish it up or fix it.
1:13:47
SPEAKER_00: Writing in this way has made me,
1:13:48
SPEAKER_00: allowed me to publish much more than I ever would have otherwise.
1:13:51
SPEAKER_00: Like I have perfectionistic tendencies.
1:13:54
SPEAKER_00: This was very much a counter to that.
1:13:56
SPEAKER_00: Where, of course,
1:13:57
SPEAKER_00: I want some of my stuff to be my best work,
1:14:00
SPEAKER_00: my most polished, but it's completely unrealistic.
1:14:02
SPEAKER_00: I would never put anything up if that were the case.
1:14:04
SPEAKER_00: So working in this way really freed me.
1:14:07
SPEAKER_00: I mean, I started my garden in like 2020, I think.
1:14:11
SPEAKER_00: So it's not been that long,
1:14:12
SPEAKER_00: but I've written a fair amount,
1:14:14
SPEAKER_00: more than most people have blog posts on their website,
1:14:16
SPEAKER_00: because some of them are frankly like three paragraphs long.
1:14:21
SPEAKER_00: Or they have some of them.
1:14:22
SPEAKER_00: I have big, long essays.
1:14:24
SPEAKER_00: They've taken me a long time to write,
1:14:25
SPEAKER_00: but halfway through the SL,
1:14:26
SPEAKER_00: you will hit something that says draft in progress and everything below
1:14:29
SPEAKER_00: that point is pretty rough.
1:14:31
SPEAKER_00: One of these I started three years ago.
1:14:32
SPEAKER_00: Every time I have like some free time, I go back to it, I like finish one more paragraph.
1:14:36
SPEAKER_00: I like move the little draft notification down.
1:14:38
SPEAKER_00: It like allows me to work in public,
1:14:40
SPEAKER_00: which I also think helps like people kind of get to see behind the garage door
1:14:45
SPEAKER_00: a little bit because often the notes below the draft point are like,
1:14:49
SPEAKER_00: Somewhat related to this, again,
1:14:50
SPEAKER_00: just going away from AI,
1:14:51
SPEAKER_00: you also talked about barefoot developers and home cook software.
1:14:52
SPEAKER_00: Yeah.
1:15:18
SPEAKER_01: Who are barefoot developers?
1:15:21
SPEAKER_00: So this was, I'm not sure I remember what year this is.
1:15:23
SPEAKER_00: It might be 2024, which feels like an eternity ago.
1:15:26
SPEAKER_00: It was not that long ago.
1:15:28
SPEAKER_00: This was at the local first conference in Berlin in 2024.
1:15:31
SPEAKER_00: I did this talk, yeah.
1:15:33
SPEAKER_00: Home cooked software and barefoot developers.
1:15:34
SPEAKER_00: Home cooked software is a phrase that comes from Robin Sloan.
1:15:37
SPEAKER_00: Which is software that you make for you and your family and the way you would make
1:15:40
SPEAKER_00: a home-cooked meal.
1:15:41
SPEAKER_00: So it's about this app that he built for his family
1:15:43
SPEAKER_00: where they send each other little video notes.
1:15:45
SPEAKER_00: And it's not on anyone else's server.
1:15:47
SPEAKER_00: He doesn't have to pay for it.
1:15:48
SPEAKER_00: It's not managed by some big company.
1:15:49
SPEAKER_00: And this is the way more software should be.
1:15:51
SPEAKER_00: There's lots of little pieces of software we can build for ourselves
1:15:55
SPEAKER_00: that benefit our lives and our families.
1:15:56
SPEAKER_00: And it doesn't have to be like $2.99 on the App Store or sell my data in exchange
1:16:00
SPEAKER_00: for this software.
1:16:01
SPEAKER_00: And I made this case at this conference that...
1:16:05
SPEAKER_00: I expect to see an explosion of this with language models.
1:16:07
SPEAKER_00: And okay, the writing was already on the wall.
1:16:08
SPEAKER_00: I don't think it was that crazy of a prediction, but we were still more in the early days.
1:16:12
SPEAKER_00: We didn't have agents yet.
1:16:13
SPEAKER_00: But now, of course, it's like totally happened.
1:16:15
SPEAKER_00: There's like personal software everywhere.
1:16:16
SPEAKER_00: Everyone's building their own recipe manager and their own household app, their own gym app.
1:16:20
SPEAKER_00: And like,
1:16:21
SPEAKER_00: there's all these great use cases that people can just make the software
1:16:23
SPEAKER_00: to be what they want without having to download a bog standard version of it.
1:16:28
SPEAKER_00: But I extended this concept by saying, okay, home cook software is for you and your family.
1:16:33
SPEAKER_00: But there's another concept that I call barefoot developers that comes from in Maoist China.
1:16:37
SPEAKER_00: They had this thing called barefoot doctors where they took peasants
1:16:39
SPEAKER_00: from rural villages and trained them up with basic medical training,
1:16:42
SPEAKER_00: like giving vaccinations,
1:16:43
SPEAKER_00: you know,
1:16:44
SPEAKER_00: administering antibiotics,
1:16:46
SPEAKER_00: and then distributed them out to the villages so that you would improve health care
1:16:49
SPEAKER_00: for everyone.
1:16:49
SPEAKER_00: And it was a wildly successful program,
1:16:51
SPEAKER_00: like across the board improved health,
1:16:52
SPEAKER_00: as you would expect people who can't get to a hospital.
1:16:55
SPEAKER_00: And I think we need the same concept for developers because
1:16:57
SPEAKER_00: so far developers are extremely expensive as a profession
1:17:00
SPEAKER_00: and like extremely skilled.
1:17:02
SPEAKER_00: But there's lots of people with lots of needs
1:17:04
SPEAKER_00: who would maybe need little software built for like their allotment garden
1:17:08
SPEAKER_00: or their like street.
1:17:09
SPEAKER_00: But they have some problem they need solved.
1:17:11
SPEAKER_00: And like Google Sheets doesn't really cut it.
1:17:13
SPEAKER_00: Like it can't really just do like a Google Doc.
1:17:15
SPEAKER_00: They want something like maybe managing a schedule or like managing inventory.
1:17:19
SPEAKER_00: And like either they would have to all like raise money together and like fund it
1:17:22
SPEAKER_00: and pay some piece of software that doesn't quite fit their needs but does
1:17:26
SPEAKER_00: an okay job or maybe trades their data for this.
1:17:29
SPEAKER_00: Or if they had someone who was,
1:17:30
SPEAKER_00: let's say,
1:17:31
SPEAKER_00: this barefoot developer type who's like kind of a power user,
1:17:33
SPEAKER_00: someone who might use Notion and Airtable in an old world,
1:17:37
SPEAKER_00: now can just use an agent or language model and build them the software they need
1:17:41
SPEAKER_00: for like not many tokens, like pretty cheap.
1:17:43
SPEAKER_00: They don't have to read the code as long as it works, like as long as they verify it's fine.
1:17:47
SPEAKER_01: I mean, what reminds me a little bit of the webmasters.
1:17:50
SPEAKER_00: Yeah, webmasters.
1:17:51
SPEAKER_01: So back,
1:17:52
SPEAKER_01: it was a long time ago,
1:17:53
SPEAKER_01: but in the 90s,
1:17:55
SPEAKER_01: maybe mid-90s,
1:17:56
SPEAKER_01: late 90s,
1:17:57
SPEAKER_01: the webmasters were just power user administrators who would administer,
1:18:01
SPEAKER_01: for example, schools.
1:18:03
SPEAKER_00: Yeah.
1:18:03
SPEAKER_01: Yeah.
1:18:04
SPEAKER_01: Yeah.
1:18:04
SPEAKER_01: Yeah.
1:18:27
SPEAKER_01: They might forget to patch stuff or they might not know.
1:18:29
SPEAKER_01: And there were like forums, webmaster forums where they helped each other.
1:18:33
SPEAKER_01: And it was kind of like because what you talk about,
1:18:36
SPEAKER_01: the China example, that was, of course, a government program with the goal.
1:18:39
SPEAKER_01: But this was just a grassroots movement.
1:18:41
SPEAKER_00: Yeah.
1:18:42
SPEAKER_00: Yeah.
1:18:42
SPEAKER_00: And I think that needs to happen again.
1:18:43
SPEAKER_00: Like we definitely have these people to some degree.
1:18:45
SPEAKER_00: Like there's like a techie person in each community who kind of people go to for help.
1:18:49
SPEAKER_00: But I think these people need more like support and tools.
1:18:52
SPEAKER_00: And I mean that in the sense of building like libraries or frameworks for them.
1:18:55
SPEAKER_00: Because I think there's a whole bunch of people now coming in as vibe coders
1:18:58
SPEAKER_00: or whatever, right?
1:18:59
SPEAKER_00: Like...
1:19:00
SPEAKER_00: Building apps without looking at the code.
1:19:01
SPEAKER_00: But you know they're hitting terrible security foundations
1:19:06
SPEAKER_00: and they're probably handling the data wrong.
1:19:08
SPEAKER_00: They're going to lose their database at some point.
1:19:10
SPEAKER_00: They don't have good foundations to build on because they're just telling the agent,
1:19:13
SPEAKER_00: hey, make me an iOS app that does X.
1:19:15
SPEAKER_00: But they don't really understand engineering.
1:19:17
SPEAKER_00: I would advocate for something that is,
1:19:20
SPEAKER_00: I think it should be local first because as a philosophy,
1:19:22
SPEAKER_00: it matches up with the need.
1:19:23
SPEAKER_00: It's like data should be local.
1:19:24
SPEAKER_00: You could just sync it between machines.
1:19:25
SPEAKER_00: You don't really need a database in the cloud.
1:19:27
SPEAKER_00: There's no point to that.
1:19:29
SPEAKER_00: A little bit of data sovereignty.
1:19:30
SPEAKER_00: They own the data.
1:19:31
SPEAKER_00: No one else can buy it or trade it or get access to it.
1:19:34
SPEAKER_00: I just feel there should be much stronger local-first frameworks
1:19:37
SPEAKER_00: that give you these good,
1:19:39
SPEAKER_00: solid principles in place,
1:19:40
SPEAKER_00: good security,
1:19:40
SPEAKER_00: good data persistence,
1:19:42
SPEAKER_00: make it easy for these barefoot developers to build,
1:19:45
SPEAKER_00: maybe some good interface primitives on top.
1:19:47
SPEAKER_00: It's maybe something like the next version of Airtable would be this, is what I would hope.
1:19:51
SPEAKER_01: What advice would you have for engineers to learn from designers
1:19:56
SPEAKER_01: when they have access to working with a designer?
1:20:00
SPEAKER_01: Learn from, also collaborate better with?
1:20:02
SPEAKER_00: Yeah,
1:20:03
SPEAKER_00: I think it's in the same way that I'm advocating that designers shouldn't be afraid
1:20:08
SPEAKER_00: to get involved in engineering now, vice versa, right?
1:20:10
SPEAKER_00: Like in the same way that a designer can sit down with Codex or Claude or Copilot
1:20:16
SPEAKER_00: and be like,
1:20:16
SPEAKER_00: hey, explain the back end to me, like just not the details, the shape of it.
1:20:20
SPEAKER_00: I think in the same way a developer can sit down with an agent and be like,
1:20:23
SPEAKER_00: okay, I need to design a sidebar.
1:20:25
SPEAKER_00: What are the principles of good sidebars?
1:20:27
SPEAKER_00: Okay, I have my blog.
1:20:29
SPEAKER_00: Teach me about typography.
1:20:30
SPEAKER_00: You now have a very patient tutor who's just going to sit there and be like,
1:20:33
SPEAKER_00: okay, this is when you would vary line height.
1:20:36
SPEAKER_00: Here's exactly how many characters you should fit on a line.
1:20:38
SPEAKER_00: They can just teach you.
1:20:40
SPEAKER_00: As long as you're willing to learn and you want to expand into the design skills.
1:20:43
SPEAKER_00: I think before it was very hard to learn these things.
1:20:45
SPEAKER_00: You had to be on a design team,
1:20:47
SPEAKER_00: especially product design stuff where it's like,
1:20:49
SPEAKER_00: Going back to your anthropology roots.
1:21:10
SPEAKER_01: What are things that us engineers,
1:21:13
SPEAKER_01: people building software could take inspiration or learn from anthropology,
1:21:18
SPEAKER_01: like methods, approaches, things that have really served you well?
1:21:23
SPEAKER_00: Yeah.
1:21:23
SPEAKER_00: Yeah.
1:21:23
SPEAKER_00: Because I feel like I look at everything a little bit through the lens
1:21:26
SPEAKER_00: of anthropology or I try to.
1:21:27
SPEAKER_00: It's like a pair of glasses you can kind of put on.
1:21:30
SPEAKER_00: In the same way, you can be like, I'm going to look at this like an engineer.
1:21:33
SPEAKER_00: But if I look at things like an anthropologist,
1:21:35
SPEAKER_00: you think about like,
1:21:36
SPEAKER_00: what are the unspoken cultural rules going on in this interaction or this context
1:21:40
SPEAKER_00: or this problem?
1:21:42
SPEAKER_00: And especially in the context of software,
1:21:44
SPEAKER_00: like sure,
1:21:45
SPEAKER_00: maybe you're building a database,
1:21:47
SPEAKER_00: but you're building it for users in a cultural context, right?
1:21:49
SPEAKER_00: Like they have assumptions about how they decide something is trustworthy.
1:21:53
SPEAKER_00: They have assumptions about how they decide that something is worth their time
1:21:57
SPEAKER_00: or like how a flow should go.
1:21:59
SPEAKER_00: And it's all...
1:22:00
SPEAKER_00: Culturally contained,
1:22:02
SPEAKER_00: it depends also on how international your audience is and who you're designing for.
1:22:06
SPEAKER_00: But there are cultures where time flows from right to left, or from up to down.
1:22:12
SPEAKER_00: But we assume in the West, time flows left to right.
1:22:16
SPEAKER_00: But that's not universal everywhere.
1:22:18
SPEAKER_00: I think just reading a little bit of cultural anthropology and understanding...
1:22:21
SPEAKER_00: How varied people's worldviews can be.
1:22:24
SPEAKER_00: Like some cultures see,
1:22:26
SPEAKER_00: not see different colors,
1:22:27
SPEAKER_00: but categorize colors differently in a way that makes them see them differently.
1:22:30
SPEAKER_00: Like they might see blue and green as actually one unified color.
1:22:33
SPEAKER_00: And if you're designing an interface,
1:22:34
SPEAKER_00: like not that that necessarily directly applies,
1:22:37
SPEAKER_00: but I think it's understanding there's a broad range
1:22:39
SPEAKER_00: of ways humans can interpret something.
1:22:41
SPEAKER_00: And, you know,
1:22:43
SPEAKER_00: if you're building tools for humans,
1:22:44
SPEAKER_00: there might be some way you could like teach them to see the world differently
1:22:47
SPEAKER_00: through the thing you're building as well.
1:22:48
SPEAKER_00: Like change their perspective on like how they interpret reality.
1:22:51
SPEAKER_01: And then in anthropology,
1:22:53
SPEAKER_01: you said that,
1:22:53
SPEAKER_01: you know,
1:22:54
SPEAKER_01: anthropologists would,
1:22:55
SPEAKER_01: one thing that they would do,
1:22:56
SPEAKER_01: they would,
1:22:57
SPEAKER_01: the traditional anthropologists would live with natives and be embedded in them.
1:23:03
SPEAKER_01: Is that something that maybe now that AI is, you know, like it's making code a bit easier?
1:23:08
SPEAKER_01: As engineers,
1:23:08
SPEAKER_01: I think it's pretty clear that an engineer becomes more valuable the more they take
1:23:12
SPEAKER_01: on the other parts of the business, the more empathy, the understanding.
1:23:16
SPEAKER_01: Could it be just an idea when you have the opportunity to just embed yourself
1:23:19
SPEAKER_01: with Customers, become a customer yourself.
1:23:22
SPEAKER_01: I mean, this used to be the thing, right?
1:23:23
SPEAKER_01: Like Amazon had their customer obsession, which also starts with understanding the customer.
1:23:28
SPEAKER_01: But I guess this might be very natural as someone who's an anthropologist,
1:23:31
SPEAKER_01: which is like, go and try to be one of them, whoever you're building with.
1:23:36
SPEAKER_00: There is a good point that like engineers are about to have more free time in a certain way.
1:23:41
SPEAKER_00: I know there's an infinite number of engineering problems to solve.
1:23:43
SPEAKER_00: And then it's like, maybe we solve them better, you know, by having with that extra time.
1:23:48
SPEAKER_00: But a big part of this is, yeah,
1:23:49
SPEAKER_00: you could expand into more of the design side and past the interface is really like
1:23:54
SPEAKER_00: the user research side.
1:23:55
SPEAKER_00: And that is like,
1:23:56
SPEAKER_00: go fully understand the domain you're designing or building for and like,
1:24:00
SPEAKER_00: What real-world context are people using it in?
1:24:02
SPEAKER_00: When are they pulling out your app?
1:24:05
SPEAKER_00: In a factory?
1:24:06
SPEAKER_00: Are they in a tube?
1:24:07
SPEAKER_00: I don't know.
1:24:08
SPEAKER_00: A little bit of understanding context of use is a big thing that user researchers do.
1:24:13
SPEAKER_00: And then understanding the moment when someone reaches
1:24:15
SPEAKER_00: for your product versus reaching for a different product.
1:24:17
SPEAKER_00: When do they decide you're the right solution?
1:24:19
SPEAKER_00: That definitely gets into the user research side,
1:24:21
SPEAKER_00: but that's a great thing for engineers to expand into if that appeals.
1:24:25
SPEAKER_01: And as closing, what books would you recommend, ones that you enjoyed reading and why?
1:24:29
SPEAKER_00: One of my favourites that I give to most people is called Addiction by Design.
1:24:33
SPEAKER_00: So it's about people addicted to gambling machines in Las Vegas,
1:24:37
SPEAKER_00: but it's an anthropologist doing it.
1:24:39
SPEAKER_00: And she talks about both like living among these people and their experiences,
1:24:43
SPEAKER_00: but also the machine designers.
1:24:44
SPEAKER_00: Like how do you design a machine that is so addictive that someone sits at it
1:24:48
SPEAKER_00: for 12 hours straight?
1:24:49
SPEAKER_00: It's like a really fascinating thing.
1:24:51
SPEAKER_00: And how are like gambling casinos designed to have no windows?
1:24:54
SPEAKER_00: So there's like no time around you as you like sit at this machine.
1:24:57
SPEAKER_00: It's by Natasha Dauschal, uh, And I read it in university and I love it.
1:25:02
SPEAKER_00: It's a total mix of like cultural anthropology,
1:25:06
SPEAKER_00: participant observation,
1:25:07
SPEAKER_00: and also machine design and engineering
1:25:10
SPEAKER_00: and like how do you design addictive systems,
1:25:13
SPEAKER_00: which of course you read it and then you think about phones and Instagram
1:25:17
SPEAKER_00: and you reflect a little bit on like what are these systems we're building
1:25:20
SPEAKER_00: for people.
1:25:21
SPEAKER_01: Maggie, thanks so much.
1:25:22
SPEAKER_01: This was very interesting.
1:25:24
SPEAKER_00: Yeah, thanks for having me.
1:25:25
SPEAKER_00: Really fun.
1:25:26
SPEAKER_01: This was a very visual episode,
1:25:28
SPEAKER_01: so I hope you were able to watch it over video,
1:25:30
SPEAKER_01: especially the parts where Maggie showed her sketches in the notebook.
1:25:33
SPEAKER_01: And this was one of the most interesting parts of this episode for me.
1:25:36
SPEAKER_01: Even as Maggie spends her days prototyping the future of agentic tools,
1:25:39
SPEAKER_01: she starts her design process with pen and paper.
1:25:42
SPEAKER_01: Her reasoning is that agents cannot see, but us humans can.
1:25:46
SPEAKER_01: And I have to wonder, is it only Is it only true for visual stuff?
1:25:49
SPEAKER_01: I mean,
1:25:50
SPEAKER_01: when I think back to my most productive brainstorming sessions with engineers,
1:25:54
SPEAKER_01: it was usually in front of a whiteboard drawing stuff,
1:25:57
SPEAKER_01: where we drew out our ideas about the system and the components.
1:26:00
SPEAKER_01: Another extremely cool trick from Maggie is how she uses AI to generate prototypes
1:26:04
SPEAKER_01: with sliders to change parameters and change the design.
1:26:08
SPEAKER_01: She calls it a build your own Figma and we did a live demo with
1:26:12
SPEAKER_01: the Constellation animation that she played around with.
1:26:14
SPEAKER_01: This was so darn cool.
1:26:16
SPEAKER_01: You get Figma-like tools dynamically built for you.
1:26:19
SPEAKER_01: It's also pretty incredible that we can do this with agents in a matter of minutes.
1:26:24
SPEAKER_01: Finally, I appreciated how Maggie was honest about how she feels about the design craft.
1:26:28
SPEAKER_01: It's cool that agents make her work faster,
1:26:31
SPEAKER_01: but she said she's not sure she'd love if an agent just spit out a perfect design.
1:26:36
SPEAKER_01: And yet,
1:26:37
SPEAKER_01: agents are getting Now,
1:26:38
SPEAKER_01: I don't write the code,
1:26:39
SPEAKER_01: but it does feel more transactional, I feel a bit less confused.
1:26:56
SPEAKER_01: For design,
1:26:56
SPEAKER_01: agents are not there yet,
1:26:57
SPEAKER_01: but I didn't sense that Maggie wants to let go of that designer in her.
1:27:01
SPEAKER_01: I wonder if one thing to learn from her for us devs is that we should also have
1:27:05
SPEAKER_01: a notebook to make sketches of ideas, not just UIs, but architectures and systems.
1:27:10
SPEAKER_01: I don't know, but I might give it a go.
1:27:12
SPEAKER_01: At least it will make me feel less that I'm dependent on agents
1:27:14
SPEAKER_01: and I'll have some AI-free space, if you will.
1:27:17
SPEAKER_01: Thanks for listening and let me know how you like this less conventional episode.
1:27:21
SPEAKER_01: Thanks and I'll see you in the next one.