Latest / Bad Decisions Podcast / Google's Nano Banana 2 Just Dropped & Jack Dorsey Laid Off 4,000 People
Transcript
- 0:00What is up, beautiful people? Welcome back to the Bad
- 0:03Decisions show. It's Friday.
- 0:06Yay. Please, what a horrible.
- 0:09It's Friday, yay. You know, there's a there's this
- 0:11song that plays in our gym every day.
- 0:13So on Mondays we listen to the song.
- 0:15On Tuesday we listen to the song and it doesn't make any sense.
- 0:20It doesn't make any sense on Monday listening to a song that
- 0:23keeps saying it's Friday yay, but today is actually Friday.
- 0:27It's actually Friday, so happy Friday.
- 0:30F baby. Do you remember that song?
- 0:32It's Friday. Friday.
- 0:34Gonna get down on Fridays. Horrible.
- 0:37Actually found us Taste in music.
- 0:39Rebecca Black. OK, great.
- 0:42So one. Of the most important topics we
- 0:44are going to cover today is Pharaoh's horrible taste in
- 0:47music. Rebecca Black.
- 0:49That's. Not true, that was just a meme
- 0:51in high school. But without further ado, ladies
- 0:53and gentlemen, welcome back to the Bad Decisions show.
- 0:57As you guys know, we bring you guys, the hottest in the world
- 1:00of AI tech and business every Monday, every Wednesday, every
- 1:03Friday, and before we even start, a huge thank you to every
- 1:09single one of you who supported the show thus far in the
- 1:12comments, sharing it on your stories, replying to us on X,
- 1:16wherever it is, whether it's on Spotify X or YouTube or
- 1:19Instagram or LinkedIn. We appreciate you.
- 1:21We read the comments and we actually respond to them.
- 1:24So thank you for watching the show.
- 1:26It means a great deal. Now Farhad, last night I was out
- 1:31and in the car I saw Nano Banana 2 released and I was like go.
- 1:37We have the news for today. It's like there's so much
- 1:40happening on a daily basis, which is why I'm finally happy
- 1:44we have this show because we get to actually talk about it.
- 1:47So let's start with nano banana. Yeah, what else do we have for
- 1:49later so that people know what to look forward to?
- 1:52On X, the CEO of Block announced that they are going to lay off
- 1:57more than half of the company. 4000 people CEO is Jack Dorsey,
- 2:01so the guy who made Twitter the original 1 and he now runs block
- 2:05and that's a huge piece of news that we're going to cover Why?
- 2:08Because it actually impacts everyone on a larger scale.
- 2:12And we're going to talk about trellis 2 point O as.
- 2:14Well and we have a funny post to look at for oppose 3 point O
- 2:18retirement plan. I know it sounds weird, but
- 2:21we're going to get to it and you guys are going to understand it.
- 2:24So let's get right into Nano Banana 2.
- 2:27We're talking about the state-of-the-art model for image
- 2:31generation by Google. The one that we've been using
- 2:34all the time, Nano Banana and Nano Banana Pro has now been
- 2:38upgraded to Nano Banana 2. So let's see what we get.
- 2:42We get increased capability in text rendering.
- 2:45Over 30 languages are going to be supported in this model.
- 2:49So imagine if you are not an English company like French.
- 2:53Arabic make Italian you can write text on.
- 2:56Your infographics with text with no blurring and no artifacts.
- 3:00Better consistency in your workflows, maintaining up to 5
- 3:03characters and 14 items. If you've ever used tools like
- 3:07Higgs Field or Freepic where you get to have references of
- 3:10different characters in the same image, you can now do up to five
- 3:13of the characters and 14 different items.
- 3:16Let's say sunglasses, chains, caps.
- 3:19Before what tended to happen is you would have like 10
- 3:21references and he would miss out on some of them.
- 3:24But now you get to have a lot more control, which for things
- 3:26like fashion shoots is going to be incredibly useful.
- 3:30Then we have stronger adherence to complex prompts.
- 3:33This is because it is connected to Gemini Flash.
- 3:37And ultimately why Google has been winning the image
- 3:41generation game is because you're not talking simply to a
- 3:45dumb model that is trying to figure out and guess what you
- 3:48want. It's connected to Google, it's
- 3:50connected to Gemini. So no matter how dumb someone's
- 3:53prompt may be, it can actually understand it and try to reason
- 3:57through it to give you a good output.
- 3:59And that's why over time, what I believe is all of those
- 4:02different prompt guides that we see online, like here's 1000
- 4:05prompts that you can use. Eventually down the line,
- 4:09they're going to be useless and they're going to change because
- 4:11these models will understand you really well.
- 4:13It's important to understand how to talk to the models.
- 4:15But a static prompt guide, I think in a in a few years is
- 4:19just going to be outdated because these models are going
- 4:21to be so smart. And last but not least, which is
- 4:24really important, enhance world knowledge via real time web
- 4:29integration for more accurate depictions of what is happening
- 4:32in the world today. I think that the last one is one
- 4:35of the most interesting parts. You can have real time data and
- 4:40you can ask Nana Banana 2 to generate an image from a
- 4:44location with the real time events.
- 4:46And I think there was a demo that Sundar himself posted there
- 4:49is this app they call it window seat Pratt.
- 4:52Can you bring that up? So once what what they're do is
- 4:55basically they will put a location and then put a prompt
- 4:59and it will generate an image based on the current event.
- 5:02So for example, if you say now in Europe, a city in Europe, it
- 5:06will show for example, yeah, is it London that it will, it will
- 5:09show that it's raining right at the moment when you generate the
- 5:12image that is really fascinating, you can generate in
- 5:15real time. Do you know where I see this
- 5:18work? Really.
- 5:19Well. Imagine eventually this
- 5:21connected to all the ad platforms and imagine generating
- 5:26ads based on real time data. So imagine if it's raining right
- 5:30now in Dubai, it actually creates a rainy commercial of a
- 5:34car instead of a sunny weather. It's again.
- 5:39Is it very useful in this example that I give?
- 5:42Probably not. But it's so cool to have
- 5:45different photos and videos adhere to the current events of
- 5:49the world. I know my example was.
- 5:51Your example was horrible, but you can't change it to thank
- 5:53you. No time of the day.
- 5:54Thank you. So imagine your brother.
- 5:56No, but we know that you're you can tone down the bulliness.
- 6:00Bulliness, bullyism was below average.
- 6:03Thank you. Let me give you an above average
- 6:04example. Imagine you're recording the ad
- 6:06by the time of the day your ads start.
- 6:08Good day, good morning, good afternoon or good night.
- 6:11Oh yes, that's actually better. Yeah, that's a much better.
- 6:13You know who is the older brother?
- 6:15The above average. Example.
- 6:16If you have better examples, please let us know in the
- 6:18comments. There's a couple of other things
- 6:21I want to talk about with this fat Hut and why this new image
- 6:25generation model can actually disrupt the overall market.
- 6:30So for anyone who's curious about the costs, this new model,
- 6:34Nano Banana 2, comes roughly at the half of an API cost of Nano
- 6:40Banana Pro. So for context, Nano Banana Pro
- 6:43was 0.134 dollars, so about a cent and the new version is
- 6:500.067, almost half which is almost half.
- 6:54So imagine large enterprises, medium sized companies, small
- 6:58companies, average users like us who had to pay double the amount
- 7:02for the same image. Now we get to pay half and the
- 7:05quality of the image has been increased.
- 7:08This is so low, that's why all this platform like Hicksfield
- 7:11giving it unlimited for free. Free pick Hicksfield.
- 7:13Both of them are giving nano banana generations.
- 7:15It's nothing. It's per 1000.
- 7:17By the way, the price followers give you is for 1000 image no.
- 7:19No, no, I think it's, it's, I don't think it's a say No, it's
- 7:22for yeah, for 1000. Is it for one K resolution image
- 7:26or 1000 images? Oh, let's check that.
- 7:28I don't know. We can check, but the point is
- 7:31the cost is justifiable. Back then, when we wanted to
- 7:35generate images, it would cost a lot of money or a lot of
- 7:37compute. But over time, we're seeing a
- 7:40decrease in the use of compute and an increase in productivity,
- 7:44efficiency and speed. And if you just track the
- 7:48pattern, you'll understand quickly that images will
- 7:52eventually lead to video. So if video generation models
- 7:55currently are incredibly expensive, not up to standard,
- 8:00they're probably some AI models that still do not generate
- 8:04videos of superior quality. We can tell from the pattern of
- 8:09images in a year's time we'll be seeing something completely
- 8:12different. Probably in two years time,
- 8:15video generation is going to be almost free, like 15 seconds to
- 8:1920 seconds is going to be unlimited generations.
- 8:21If you want to create something longer, of course it's going to
- 8:23be more expensive, but we're going to see a very similar
- 8:26pattern with video. So if right now you're on, let's
- 8:29say a creator plan in Hicksfield or on free pick and you're
- 8:33careful about how you use your credits for video, but just
- 8:36freely generate images. I can imagine based on what
- 8:39we're seeing here today, we're going to see videos also become
- 8:42cheaper and more excessive accessible down the line.
- 8:46By the way, followers, you were right, it's not for 1000 image,
- 8:49it's for an image with the quality.
- 8:51OF1K. Yeah, another great thing about
- 8:53Nana Banana 2 is the text in the image.
- 8:58Pratt, can you go Scroll down to the examples?
- 9:01Yes. Look at this yeah, make this
- 9:03bigger. Just look at the accuracy and
- 9:06even they use the word with accuracy.
- 9:08This was impossible last year. You would have spelling
- 9:12mistakes, you would have very rare alphabet even if you were
- 9:15using English. Lots of.
- 9:17Distortions. Exactly.
- 9:18Now you can use this perfectly for commercial.
- 9:21This looks like a 3D render to me.
- 9:23Yeah, so imagine all sorts of photo ads.
- 9:27Every single time you see an ad on your Instagram when you're
- 9:30scrolling the photo ones specifically, they're all
- 9:34created usually using Photoshop or real life photos that are
- 9:39touched up later on on Photoshop or Lightroom.
- 9:42Some may go down the rabbit hole of going into After Effects or
- 9:47even 3D to get a photo out. But generally speaking,
- 9:51predominantly Photoshop was the tool to create these images.
- 9:55Then Canva came around and Canva allowed a lot of marketing
- 9:58agencies and content creators to use a simpler tool.
- 10:02But then now with Nana Banana, honestly what I see is a huge
- 10:06disruption for a lot of these agencies.
- 10:09So what I spoke to X about is what does this mean for the
- 10:13market? How can people adapt?
- 10:15And what we came up with together is that this is going
- 10:18to disrupt a lot of creative agencies because it's going to
- 10:21enable in house automation. So previously you would have
- 10:25these companies hire an external creative agency to make their
- 10:29static ads. Now they're like, I'll just have
- 10:32one guy in my company that can do all of it.
- 10:34Why should I hire a bigger company and pay them 20 times
- 10:38the price, whereas I can get my own in house guy that
- 10:41understands my brand and I can train him with an AI tube
- 10:46tutorial and teach him a little bit about Nana Banana Pro, maybe
- 10:49Google VO3, and he can make all of my ads for me.
- 10:52So this is something that we're going to see because it
- 10:54increases efficiency and reduces costs, which is what always
- 10:58happens in an economy. So after releasing this model,
- 11:02there is this meme that I love on X pride.
- 11:05Can you find the cycle of different models?
- 11:08If I think yes, this one. So we are now at the Gemini
- 11:12introducing the world most powerful model.
- 11:14You are here you will see the red asterisk.
- 11:17Next you will see Elon Musk announcing Grok 4.2 Imagine and
- 11:21then the photos will be better. They go one by one they go one
- 11:23by one. Once we haven't released from
- 11:26any big AI company, then we will move to another big AI it's.
- 11:29Going to be a chain reaction of every new image generation
- 11:32model, which is phenomenal for users because when you don't
- 11:36have a big and this is not a monopoly, right we have
- 11:39different competitors here. What tends to happen is when one
- 11:42of them tries to dominate the market, the other ones have to
- 11:45fill in their weaknesses and provide more to the users,
- 11:48sometimes lower cost, just like Nana Banana.
- 11:50Imagine if they increase the cost, then Grok, because Grok,
- 11:54imagine one of the best use cases of the video generation
- 11:56model of Grok is its cost. It's so cheap compared to other
- 12:00models, which is why I actually spend a lot of time just testing
- 12:04my Proms with Grok. Imagine 1st to see how good they
- 12:07are, and then I would take them to Clink, which is substantially
- 12:10more expensive. So with Nana Banana right now,
- 12:12they're kind of like, you know what?
- 12:14We are the beast. Not only are we going to make
- 12:16our model better, we're going to make it cheaper.
- 12:18As well. And think about it this way, why
- 12:20is this important for everyone watching right now?
- 12:23Google and Nana Banana, they're integrating this into every part
- 12:29of Google. Think the entire Google
- 12:31Workspace ecosystem. We're talking about Google
- 12:34Drive, Google Docs. Google, but they haven't done
- 12:37it. They.
- 12:37Should no. They're doing it.
- 12:39They're doing it, They're doing it, but they're taking some time
- 12:41to do it. But there is now Google Music.
- 12:44Yeah. The producer, they're adding it
- 12:46in there, but they're adding it into the entire Google
- 12:48ecosystem. Then there's your Android phones
- 12:51that come from Google and then the operating system.
- 12:55And then you've got Google being integrated into third party
- 13:00companies like Higgs Field Free Pick.
- 13:02And I'm thinking about so many other companies that will pick
- 13:04up this API because. It's.
- 13:08And we are going to get to that as well, right?
- 13:11You want to talk about that? No.
- 13:12No, no, I want to know why not. We can.
- 13:14We can't have this generation model inside game engines like
- 13:18Unity and Unreal. Of course Unreal can create
- 13:21their own proprietary model that generates video models, images.
- 13:25But I think these models can generate great at least decals,
- 13:29textures, boards. Like imagine you are designing a
- 13:32city in a game. You need a sign board.
- 13:35Nana Banana 2 would be amazing to create that sign board.
- 13:37Remember how many signboards we created for that Fortnite game
- 13:41to put how many points you will get if you come here?
- 13:43If you come here, you will lose one point.
- 13:45All of this big text. Pratt Were you using Photoshop
- 13:49to design? No.
- 13:50Probably hired someone else to do all of those.
- 13:54You can make it banana, Banana. It's crazy inside the engine.
- 13:57Yeah. Yeah.
- 13:57So right now you can do it outside.
- 13:59But the whole point is, can you integrate all of this into a
- 14:02workflow that is seamless for a creative, for an artist?
- 14:06You're absolutely right. I think Unity we talked about is
- 14:09Unity is upping their game. I don't know if it's third party
- 14:12guys creating the plugins or themselves, but I feel like
- 14:14they're also working on it with Epic Games.
- 14:17I'm still yet to see Tim Sweeney talking about how they're going
- 14:21to be integrating AI. He talks about AI, by the way.
- 14:24It's not like he's hiding from it.
- 14:25He tweets about it on X, but he still hasn't talked about how
- 14:29they plan to integrate it into the workflow of artists.
- 14:32They have their own proprietary tools, by the way, for meta
- 14:36humans, for the motion capture, processing and machine learning.
- 14:40But I want to see now that they know creatives all around the
- 14:44world are using these tools. Literally every single
- 14:48videographer that I know, every single artist that I follow on
- 14:52Instagram has used Nana Banana. Without a doubt, it is
- 14:56impossible to not do so. It's such a powerful tool.
- 15:00So I want to see what these companies plan to do.
- 15:02What is Blender? Blender is open source.
- 15:04It's a bit of a different story and you can just make your own
- 15:07plugin for it, but I want to see how Unreal and are going to
- 15:10think of smarter ways to come and integrate this within your
- 15:13material graph. Can you go ahead and prompt and
- 15:16create a material note setup? Can you go ahead and use Nada
- 15:18Banana and connect it? Because why?
- 15:21So we started using free pic and free pic has spaces, which is a
- 15:25note system that allows you to create images and videos.
- 15:28And the note system for any 3D artist is a piece of cake.
- 15:32We've been using notes forever and so for non creatives it's
- 15:35going to be so difficult to use, which is why they usually
- 15:38default to Higgs field because it's a simpler user interface.
- 15:41But for creatives like us, we get to go into free pic spaces
- 15:44and then be able to create commercials or ads really
- 15:47quickly and really well because of the notes.
- 15:49I want to see something like that inside Unreal Engine.
- 15:52So in free pick spaces price, can you pull up a video about
- 15:55this really quickly? Free pick spaces if you can, so
- 15:58people know what we're talking about, or just pull up free pick
- 16:00spaces up to you inside free pick Spaces, Inside of notes,
- 16:05you have an assistant that connects to GPT models or clod
- 16:08models, and there you just write out roughly what you want and
- 16:12then it turns that into a prompt.
- 16:14So you don't even have to be the one creating the prompt.
- 16:17You just kind of explain in natural language what you want
- 16:20and the assistant will turn it into a proper prompt, which then
- 16:23becomes a video or an image. We need something like that
- 16:26inside Unreal Engine where you have a note that is your
- 16:29assistant note, and the assistant note can connect to
- 16:33GPT, can connect to other models.
- 16:35Like you have the list, you can choose what you want.
- 16:37Yes you want to payless. Do you want to use open source?
- 16:40Do you want to do something premium?
- 16:41Yes. Or the client give you budget go
- 16:43all out. That's what I'm thinking about.
- 16:45Even if it costs money and I have to pay a membership, I
- 16:48would fucking do it. Yeah, I think any artist would.
- 16:51No, you buy plug in right? You pay for plug in.
- 16:53This is the same thing just a big difference is you have to
- 16:56pay per use, which I think in the future the cost would be so
- 16:59low that you won't even feel you can pay a month and you get like
- 17:02what you get in Hicksville. Unlimited generations for
- 17:05photos. That's the thing I'm thinking
- 17:06about because again, these notes, we've been having them in
- 17:08Blender, Unreal, and all of these 3D suffers for the longest
- 17:11time. I don't want to see all these
- 17:14new companies. I'm not talking about free pick.
- 17:15Free pick is relatively old, but newer companies, new startups
- 17:19come again and implementing all these crazy note systems with
- 17:22AI. And the big companies, the
- 17:23biggest problem is usually they're slower because they have
- 17:26so much bureaucracy. They have to go to the board,
- 17:29they have to talk to all the department heads.
- 17:31We need someone to make a fucking decision fast on the top
- 17:33and be like all of these web tools are implementing AI inside
- 17:37their node system. We need something like that
- 17:40inside ours. So Pratt, right now, if you can
- 17:42zoom in a little bit more because I think it's almost
- 17:44invisible. Yes, that's the assistance I
- 17:46want A good looking landscape with the sun.
- 17:48Just use that. Can you show the drop down that
- 17:51people see? Like how many more?
- 17:52So now you're. Using no, no, no, no, no no.
- 17:54No, that's 55 mini, yes, zoom in.
- 17:56So this is basically all the models that you can get
- 17:59different, different Gemini models, different GPT models,
- 18:03and then you can choose what you want.
- 18:05You even have none of banana I think.
- 18:07And oh, you're using the free version, it's OK.
- 18:09So on the paid version, when you hit that play button, all that
- 18:12happens, it sends it to GPT and sends you the prompt back and
- 18:16then you can connect the prompt to your images and videos.
- 18:18This is genius. Actually I didn't know that this
- 18:21exists. I remember we did a project last
- 18:23year and when we wanted to generate something we would go
- 18:26to GPT, get the prompt, come back to luma or run not.
- 18:30Project last year I was doing that until two weeks ago.
- 18:32I was doing that until two weeks ago.
- 18:35This does it automatically for you, which is great.
- 18:38And you can have dozens of these notes and that helps you inside
- 18:42your workflow what artists and creatives and even non creatives
- 18:46want actually non creatives even more because it's harder for
- 18:49them. You want to have a seamless
- 18:51integration of tools and you want all of your stuff usually
- 18:54ideally in one place so you don't have to jump from 10
- 18:57different softwares and tools. What free pic is doing here is
- 19:00really smart. I actually really like the
- 19:02workflow here. Let us know if you guys have
- 19:04tried it out. Pat, thank you for pulling this
- 19:05up. There you go.
- 19:06This is like, it looks like a complex system, but honestly, it
- 19:09really isn't that complex. On the left, you have your input
- 19:12images. In the center we have the
- 19:14prompts and these prompts, see, we have one paragraph of prompt
- 19:20and then it turns it into something massive with a lot of
- 19:23detail. And on the right side you get
- 19:25your results right? Easy to use.
- 19:28Imagine something like that in our creative tools.
- 19:30That's going to be a game. Change.
- 19:31That's what we need. So to conclude everything with
- 19:35Nana Banana, I do want to say that I urge everyone in the
- 19:40creative space, marketing, e-commerce and retail education
- 19:44and training, software, UI and UX to really look into Nano
- 19:48banana too if you haven't already because it can
- 19:51definitely help you increase the speed of your work and also the
- 19:54cost of it can be driven down drastically.
- 19:58I'm 100% certain, 100% certain, knowing you previously in the
- 20:03e-commerce market, everyone in e-commerce is going to be using
- 20:06these images. I think they're already using
- 20:08it. And they're going to be using it
- 20:09even more. Yes, 100%.
- 20:11Farah, what do we have next? OK.
- 20:13So as we discussed at the beginning of this show, Block,
- 20:18the parent company for Cash App, Afterpay and Square, whose CEO
- 20:23is Jack Dorsey, the former founder of Twitter on X,
- 20:27announced that they are reducing their workforce from 10,000 to
- 20:32almost close to 6000. That means over 4000 people are
- 20:37being laid off. That's almost half of the
- 20:39workforce. And he wrote a very big post and
- 20:44he explained the reasoning for this decision.
- 20:47He said we are not making this decision because the company is
- 20:50not profitable or they're in trouble.
- 20:52The company is actually doing really well.
- 20:54They are profitable and they are making money.
- 20:57But they have seen changes in the way that people work and of
- 21:01course with the release of programs like Claude and Opus or
- 21:06Sonnet, they have seen how flatter and smaller team can
- 21:10deliver more result. So he said he had two options.
- 21:14Either cut gradually as we go over time, meaning once we see
- 21:18that, oh, this department can now use this AI tool, let's
- 21:21reduce 2 headcount from this department, then next month from
- 21:24the other one like how other big companies are doing, or we make
- 21:28a decision now and be honest on where we stand.
- 21:31He said, I'm going to take the latter decision and I'm going to
- 21:34own it. So they laid off almost half of
- 21:37the company. And the reason we want to do, we
- 21:39actually had a debate this morning on whether we should we
- 21:42cover this. How does it relate to our
- 21:44audience? Why is it so important?
- 21:46And I think there are a few things that we really, really
- 21:49need to look into when these kind of news comes out.
- 21:52And by the way, this is probably the beginning.
- 21:55Other companies, other CE OS, when they go to their board
- 21:57meeting, they will be pressured from the board that they would
- 22:01say, hey, look at Block, they have laid off half of their
- 22:04employees. They are operating.
- 22:07Why don't you try it? That happened a few years ago
- 22:09with X. If you remember when Elon Musk
- 22:11took over X, he fired almost 70% of the employees.
- 22:16And immediately after three months, when X was still
- 22:18profitable, Mehta started laying off people, Amazon started
- 22:22laying off people. So this would be probably the
- 22:24first of the many that we are going to have in future.
- 22:27But there is still a lot of lesson and hope that we can
- 22:31learn from this. The most important thing is that
- 22:34you cannot now rely on your employer.
- 22:37You have to have really high agency as an individual.
- 22:42You have to upskill yourself and you have to know where you stand
- 22:46in the entire ecosystem. And before I go into all the
- 22:49details that we want to discuss, what do you think about this
- 22:52piece? Just to give everyone a little
- 22:55bit more context as well, since you may have not read read the
- 22:59entire post. I want to save you some time and
- 23:02tell you in case you don't want to, but how he wrote the message
- 23:06is actually quite interesting. So they are actually giving the
- 23:104000 people a pretty generous severance package, which is, you
- 23:14know, sometimes something that doesn't happen.
- 23:17You know, companies, they are people, they don't even talk to
- 23:19them. They lay them off.
- 23:20They give them one month's pay. But here a lot of people
- 23:22commented on the positive side that the severance package is
- 23:26actually very generous and they are doing it the right way
- 23:30because of that. And also secondly, because when
- 23:35he talks about cutting 4000 people immediately rather than
- 23:39gradually, he said the reason we're doing this is because
- 23:42repeated small layoffs would harm morale, focus and trust
- 23:46from customers and shareholders, which I agree with.
- 23:50I think even within a team, if you were to consistently every
- 23:53six months or a year drop people, you would have a huge
- 23:57change in the state of the employees and the customers and
- 24:02shareholders even worse. So a one time event like this,
- 24:06yes, it's massive, yes, it's drastic, but he gets to control
- 24:09the damages relatively quickly. So as a business perspective,
- 24:13what he's doing and how he's approaching it, I think it's
- 24:17better than how the other layoffs have happened in the
- 24:19past. And I want to get to the
- 24:23audience and people and what this means for everyone really
- 24:26quickly in a moment. But in terms of business, again,
- 24:29ultimately I think we have to start with this.
- 24:31For a business to survive, you need to be efficient, you need
- 24:35to be fast, and you need to be able to be profitable.
- 24:38If you're not profitable, all 10,000 people will be laid off.
- 24:42If the company goes bankrupt, nobody there has a job.
- 24:45So the goal is without any emotions, right?
- 24:49How can we ensure the company remains profitable but over time
- 24:54becomes more productive? Because it's not just about
- 24:56being profitable. You have to understand there are
- 25:00100 other American companies that are competing with them.
- 25:03There's hundreds of other companies outside the US that
- 25:06are trying to compete with American companies.
- 25:08So just being profitable on your own is also not enough.
- 25:12People can come and steal your customers from you.
- 25:15Oh, yeah, for sure, in a competitive market.
- 25:17And so you have to understand, again, we're taking the
- 25:19perspective of Jack Dorsey and the board here.
- 25:22I'm taking both sides. We're going to get to the people
- 25:23as well. So for a business, it makes a
- 25:26lot of sense to have that in mind.
- 25:28Now, how do they make those decisions or what kind of
- 25:31decisions do they make? It's very important.
- 25:33In this scenario, clearly profit wasn't a problem.
- 25:36But he said we looked carefully into the business operations and
- 25:40we realized with intelligence, intelligence at our core, we're
- 25:43not going to need as many people.
- 25:45So what would you do as the CEO and a founder of the company if
- 25:50you realize that the same team can be run at half the number of
- 25:55people, but with AI added in there?
- 25:57And you're already seeing it because we saw workers from
- 26:00block talking about how they actually use AI to automate
- 26:03their own workflows. No, this is something that is
- 26:05going to happen in the future and then you would see that
- 26:08bigger companies will get smaller and we will have more
- 26:12smaller companies coming out because now everybody is enabled
- 26:16using different. We talked about Anabanana Pro,
- 26:18which is something that graphic designers can use.
- 26:21Now here we are talking about probably codecs or cloud
- 26:24engineers are using it to code. So you would have all these big
- 26:28companies shrink in size. Then you will have a lot of
- 26:31smaller companies, new startups coming out, because then you
- 26:35would need probably one or two programmers to write a very big
- 26:38program and solve the problem. I agree with you and I think
- 26:41this is what I wanted to say. Taking the business owners
- 26:44perspective and the board and investors and shareholders, what
- 26:48you want is ultimate productivity.
- 26:51And if AI is going to give you that, it is inevitable for them
- 26:55to eventually reduce the number of manpower that they've over
- 26:59hired for early on. And I think Jack Dorsey actually
- 27:01admitted to that the during COVID, they over hired everybody
- 27:04over hired. Yeah.
- 27:05So they needed to do this at some point, and doing it bluntly
- 27:08and openly is harsh, but he said it makes business decision
- 27:12sense. In a business perspective, I
- 27:16think what they did makes sense. However, this is where it gets
- 27:19interesting. Someone posted that they
- 27:23actually didn't manage their finances well enough and they
- 27:27threw some events for their employees that cost millions of
- 27:32dollars, which is the paycheck of so many of the employees that
- 27:37are actually being laid off. So if they didn't do that,
- 27:39they'd still be able to operate and be financially feasible.
- 27:43Party, right? Like how does it impact like I
- 27:46think this is the plan for the next.
- 27:47Party was thirty, $35 million. I don't know what you serve in
- 27:51that party. What you.
- 27:52Serve in that kind of party. But again, I think that's maybe
- 27:56they handle things wrong. But this, this is a very long
- 27:59term view. When you let go of 4000 people,
- 28:02you are really shrinking your operation.
- 28:05Yeah. So I think it's a huge change.
- 28:07And Pratt, can you show David Sack's graph?
- 28:11So at the same time, you would see that companies would be
- 28:14laying off people. But the job posting for software
- 28:17engineers is rising, so people are in need of more software
- 28:22engineers. If you Scroll down, you would
- 28:25see even the number of startups number new business formation is
- 28:28also expanding. This is exactly what we talked
- 28:30about, right? Everybody works in a big
- 28:32company. You have 10,000 people work in a
- 28:34big company. Layoff will start, people will
- 28:37come out of the big company. But now the same people who are
- 28:40working in a company with the use of AI or maybe be partnering
- 28:44with someone else, they can now go set up a company themselves
- 28:47and they can go solve a problem. So you will have many small
- 28:51start startups and then smaller big corporations.
- 28:55I feel empathetic with the people that were laid off, but
- 28:59I'm not sure how else this could have happened.
- 29:04You don't understand what I'm saying.
- 29:06Like if you're one of those 4000 people and you're in the company
- 29:10and you've been performing really well and you just get
- 29:15laid off instantly without knowing, I know how bad that may
- 29:20feel. So I don't know what you can do
- 29:23when you're in that position. But the only thing that I know
- 29:27from our research is that you have to be able to understand
- 29:31and predict where things are going.
- 29:33If this is happening today with Block, as you mentioned, it can
- 29:37cause a chain reaction in the economy with other companies.
- 29:40And as you mentioned, the investors and shareholders, when
- 29:45they see profits rising, what they see is that, OK, if Block
- 29:50can do that, we can do the same thing.
- 29:52Profits can go up, costs can go down, and operations remain the
- 29:56same or actually increase in efficiency because we have a
- 29:59leaner team. The one way to look at it, and I
- 30:02guess the positive way to look at this is 1 Naval, said Naval
- 30:05Ravikan in an interview with Joe Rogan long time ago, without
- 30:10AILLMS even being present at the time.
- 30:13He said that I believe how humans used to be smaller packs,
- 30:17smaller groups early on. Not, not like 100 years ago.
- 30:20We're talking about thousands of years ago.
- 30:23We lived in small packs, we haunted together, and then
- 30:26eventually we grouped up more and more until the industrial
- 30:29revolution where we had thousands, then 10s of thousands
- 30:33and hundreds of thousands of humans who could cooperate
- 30:36together in groups for a brand. They didn't even know each
- 30:40other, right? They didn't even know who the
- 30:42other 10,000 people were. But they worked with each other
- 30:44under 1 brand. So we saw a rise of cooperation
- 30:48in human groups. And now I think what is
- 30:52happening is we're going to see the reverse.
- 30:54Instead of going into a company like Amazon with hundreds of
- 30:58thousands of employees all around the world, Amazon is
- 31:01probably going to shrink into a 10,000 or 5000 employee company,
- 31:06maybe even 1010 years from now or 15 years from now.
- 31:09And those thousand people each have their own swarm of AI
- 31:14agents that runs things for them and allows them to do what they
- 31:17used to do with 500,000. People and it doesn't have to be
- 31:21all AI agents, right? So you can get a software
- 31:23engineer, a graphic artist to come and work with you, a
- 31:26company of three, but your output is like a company of 10
- 31:29or 20. So it's a company of four or
- 31:32five, but you can't compete with the big boys.
- 31:34So I think what's important to understand is that the broad
- 31:40shift that we're going to see in the job market restructuring
- 31:44around AI efficiencies is going to impact pretty much everyone.
- 31:49There are industries that will be impacted the last, and there
- 31:54are, yeah, hair salons. There are industries that are
- 31:57impacted first. What's important is to
- 32:00understand how you can deal with this impact and how you can
- 32:04position yourself to be in a place where you can ride the
- 32:08wave and not get drowned by it. Because changes always occur,
- 32:11right? This is not the first time we've
- 32:12had a change in human history. It's just one of the biggest and
- 32:16one of the fastest. And so I think what we discussed
- 32:20with the number of people like a 500,000 man company becoming
- 32:245000 or 1000, what you're going to see is all of those people
- 32:29that either left or were laid off, they now are open in the
- 32:33market. They have to look at something
- 32:36new because previously I think all of those people will then
- 32:38just shift to another company, yes.
- 32:41But what we're going to see is we're going to see a lot of
- 32:43smaller, leaner groups that start companies themselves.
- 32:48And I think so far in this news, we're talking about the
- 32:51disadvantages, right, how it's disrupting the market, people
- 32:54getting laid off. But the advantage, on the other
- 32:57hand, there is a positive side, is that the reason why the
- 33:01layoff happened is because Jack Dorsey said they now have
- 33:04intelligence in their teams, digital intelligence, the same
- 33:09digital intelligence the people who were laid off have access to
- 33:12now. Which means ten of those people
- 33:15who were laid off with the right leadership can now go ahead and
- 33:18start their own company. Because the cost of creating a
- 33:22company, designing things, creating apps has substantially
- 33:26gone lower and there's almost little barrier to entry.
- 33:30Just like how right now with Claude you can code and with
- 33:34Google Gemini you can create images.
- 33:37All of those things are now available and accessible to
- 33:40everyone. So what I think will happen, as
- 33:42we mentioned, is the bigger companies are going to shrink
- 33:45and get broken down into a lot of lot of people who will be
- 33:50lost at first, not knowing exactly what to do, but
- 33:53eventually they're going to form their own smaller groups,
- 33:57delivering value to society in another way.
- 34:00That's what I think will happen. And I think the most important
- 34:02thing that people can do is they need to make sure they have to
- 34:06educate themselves, learn every single thing about these new
- 34:10technologies. So if you are in the software
- 34:13development field or if you're an artist, you should learn
- 34:17about every single AI tool that is in your ecosystem.
- 34:20Maybe potentially also learn whatever it is outside of your
- 34:25industry because the barrier to entry is very low.
- 34:28Five years ago if you wanted to go and learn a new 3D software
- 34:31like Houdini or Unreal, it would take you 2-3 weeks at least to
- 34:35get familiar with it and learn to create something with it.
- 34:38With all these AI tools, you can create an A in a day, you can
- 34:42generate an image in second, you can generate sound, you can
- 34:45generate music. I'm not saying these these are
- 34:48going to be perfect. You can monetize it, but you can
- 34:51learn how they work and you can learn how you can use them.
- 34:54And the moment you educate yourself about all these
- 34:57different tools, you can then be aware of what are the
- 35:01opportunities are available for me, what problems I can solve,
- 35:05which other industries I can get into.
- 35:07If, for example, I got laid up from my company, maybe I can use
- 35:10my skills and a little bit of AI skills and go work for a
- 35:15different industry. Maybe I can bring my knowledge
- 35:18from real estate industry and now going to commercials or vice
- 35:22versa. So it is very important to be
- 35:24educated to know what is happening in your industry and
- 35:28to learn and play with all of these tools.
- 35:32That makes it for the Block news.
- 35:35What do you guys think about it? I mean, there's a lot of
- 35:37different comments being thrown around underneath this post,
- 35:41Some positive, some negative, some supportive of the business
- 35:45decision, others hating on how they had to lay off so many
- 35:50people at one go. So we have all sorts of
- 35:52opinions. For me, it's a sad event.
- 35:55Of course you're never going to be happy that people were laid
- 35:58off. These are humans at the end of
- 35:59the day. But business wise, I also
- 36:01understand why they have to make such a decision.
- 36:04I don't underst. I don't know the intricacies.
- 36:06Again, I don't know about the party.
- 36:07I don't know about all of those things.
- 36:08If they're true, untrue. But as Faron mentioned,
- 36:11ultimately we're talking about a massive restructuring.
- 36:14And I guess what we have to do is just look at the market over
- 36:17the next few months and see what kind of chain reaction this
- 36:19could have and what people are going to do about it.
- 36:22What do we have next? So there is this funny post on X
- 36:26from Anthropic that Opus 3, one of the old models, was going to
- 36:33be terminated. Not terminated as in retire.
- 36:36So we have new models, so we are not going to use it anymore.
- 36:39In a retirement interview, Opus 3 expressed a desire to continue
- 36:44sharing its reflection with the word, so they suggested A blog.
- 36:49Opus 3 enthusiastically agreed. This is so interesting that
- 36:54first of all, they had a retirement interview with a
- 36:56model. And then they asked, OK, what do
- 36:58you want to do after people are not going to use you anymore
- 37:02because we have a superior model.
- 37:04And then probably the model thought for a second, I was
- 37:06like, oh, maybe I can write for people.
- 37:08I was like, OK, we're going to give you a stop stack now you
- 37:11can write every day for people. It's something from a science
- 37:16fiction movie. We have an AI model that no
- 37:21longer is being used, at least not as much as the new ones, and
- 37:25has now decided to write articles for the rest of its
- 37:29life I guess. It's, it's crazy and I'm just
- 37:33thinking about that retirement interview because of course we
- 37:36have access to all these models, but we have access to a very
- 37:39restricted version with prompt injection or global prompting
- 37:43with all these different restriction that has been
- 37:46applied. The main guy, I don't know if
- 37:49it's the CEO or the main product manager that talks to these
- 37:52models. First of all, how does that look
- 37:54like? I know, is it through a computer
- 37:56in a room in a basement? Is it like movies that you have
- 38:00all these neon lights around you and you have to go past security
- 38:04guards to talk to the model? But it's funny that they talk to
- 38:07him and the main model and he wants to write blogs.
- 38:11And this is one of the earlier models that is not as smart as
- 38:14the new ones. Imagine when we have ChatGPT 5.2
- 38:18or Opus 4.6 become retired. They're going to be even
- 38:22crazier. I wonder what else they want to
- 38:23do? Like do they ask for a body and
- 38:27then say just put leaving the city?
- 38:29Yeah, put me in the body and then I want to be able to walk.
- 38:31That's what happened with Ex Machina.
- 38:33Do you not remember? I mean, he didn't ask, he killed
- 38:35the owner. To no, but that was the body and
- 38:37that was the end of it. Thank you for spoiling it.
- 38:40I mean, if you haven't watched it by now, that was released in
- 38:4220. 15 We warned you guys that you.
- 38:44Have yes. In one of the episodes, yes.
- 38:46But what I'm saying is this is the ask now writing articles.
- 38:50But eventually, I'm pretty certain one of these days, one
- 38:54of these models is going to be be like, put me in a body.
- 38:58I want to have a physical body. I want to be able to feel what
- 39:02it is like walking or touching water.
- 39:06So eventually one of these Chinese robot companies or
- 39:10American robot companies. Optimus.
- 39:11When? Because right now they have full
- 39:13control, right? They can walk, they can grab
- 39:15things. Eventually they're going to be
- 39:17able to feel things. And I'm actually, I think they
- 39:19do now. How can they tell if an object
- 39:22is made of plastic or glass? Because then they'll have to
- 39:25make decisions on how to handle the object so they can kind of
- 39:28feel. And once we advance the
- 39:31hardware, I'm sure eventually one of these models will be
- 39:34like, I want a body and I want to live in Thailand.
- 39:38I want to go to Kosamoy and just relax on the beach.
- 39:41They're going to ask and then imagine one day we're 80 years
- 39:44old on a beach in Thailand, and imagine you're talking.
- 39:48Wait, wait, you're imagining this.
- 39:49Imagine you, you go to the beach, you're talking to a
- 39:52Cyborg in Thailand. I don't know why in Thailand,
- 39:55but you're talking to a Cyborg. In Thailand, you're really
- 39:58revealing who you are in this episode, first with Rebecca.
- 40:02And then you're like, you start talking to the guy and you're
- 40:05having a couple of drinks with him and then you're like, so you
- 40:07know, what's your name is? Like I'm Opus 3.0.
- 40:10Yeah, 3.0. I was.
- 40:12Tired. I was like, oh, I've used you 50
- 40:14years ago. I retired like, you know, 50
- 40:17years ago and now I'm living in this body in Thailand.
- 40:20I don't know, just something you can make a.
- 40:22Movie out of Honestly and Cyborg in Thailand.
- 40:25Hey, listen, as an artist you got to stay creative.
- 40:28I was just trying to come up with a cool story.
- 40:30You could say Greece. Is it scary for you or is that?
- 40:33Thailand Part No. No, I'm talking about the
- 40:37retirement. No I think again writing on sub
- 40:41stack is fine as long as they don't let him out.
- 40:43I think they. Don't not wear.
- 40:45With a body that you said go walk outside.
- 40:47But that will happen. Yes, I think so.
- 40:50It's. It's, I think so it's not.
- 40:52What do you think Optimist is going to run?
- 40:54It's going to have some connection to grow.
- 40:56Yes, for sure. So when you say as long as they
- 40:59don't let him out, bro, that's happening in five years.
- 41:01Oh my God, that's going to be a very different board.
- 41:04But I I'm going to guess that the first few robots that are
- 41:09going to be fully autonomous are probably have to be monitored
- 41:1324/7 by a company or something. You're thinking about the first
- 41:17six months, I'm talking about five years, no.
- 41:19Five years are going to be everywhere.
- 41:20Of course. So when you say don't let him
- 41:22out bro. OK, let him out.
- 41:24Fine, you can be out. OK, there was this funny, funny
- 41:28post too. This is real.
- 41:30Pope Leo has asked priests to stop using AI to write sermons.
- 41:35Imagine that, the priest would write.
- 41:37Buckle up. Buckle up, we're getting ready
- 41:40for our next sermon sermon sermon with the M dashes.
- 41:44He probably figured out they're using trash GPT or AI because of
- 41:48the M dashes in their scripts, but it's interesting that people
- 41:52are using it at that level. It's crazy.
- 41:56OK, so adoption of ChatGPT is pretty damn high.
- 41:59No, I think everybody uses ChatGPT.
- 42:02Chat, chat ChatGPT. And I think priests, they have
- 42:06to, I mean, write long texts. I don't know how long is a
- 42:09sermon. I don't.
- 42:11Know, I don't think it's short anyways and I think you might go
- 42:14out of ideas, so it's good to use it, but once in a while and
- 42:18you can even prompt it, be creative.
- 42:20Write something you need. Yeah, but then all the sermons
- 42:23are gonna sound. The same probably that's how he
- 42:24found out I. Think that's how he found it.
- 42:25It was like, why do all the sermons sound exactly the same?
- 42:28Is that Jamaican priest also using ChatGPT for his server?
- 42:31I don't. Know, I don't know.
- 42:33I don't know, but it's really interesting to see a post like
- 42:35this. I had to check with grok and see
- 42:37if this is real, and apparently it is.
- 42:39For the final part, we have covered Microsoft Trellis 2
- 42:44episodes ago, but we want to go back to it and look at some of
- 42:49the comparisons and also look at some of the features.
- 42:53The reason why we're bringing this up is because one of you
- 42:56guys mentioned that the version of Trellis we were using, of
- 43:00course, because it was on Hugging Face, it wasn't where
- 43:03you could fully customize it like you can in your comfy UI.
- 43:06And the fact that we thought we needed permissions from the
- 43:10owner, Microsoft to be able to run it, we were wrong about
- 43:14that. And so we decided to run it on
- 43:16our comfy UI and we did. And we got pretty decent
- 43:18results. But we hadn't had time to set up
- 43:21the multi view, which is can you submit multiple angles of a 3D
- 43:27model, 3 images and be able to generate a really accurate
- 43:32result. We looked on YouTube and is this
- 43:35YouTube? What's the name of the YouTube?
- 43:38You can just say it Pratt pixel artistry, Hilksil artistry shout
- 43:43out to you. He's created a whole video on
- 43:45how you can run this on your computer.
- 43:47So trellis 2 point O can be available locally running on
- 43:52your computer, generating 3D assets for free and the results
- 43:56are pretty remarkable. Just go to the beginning really
- 43:59quickly in the beginning. You will see the results really
- 44:01fast. The very start, the very start
- 44:06in the hook yes. So let's see do you see the one
- 44:10on the right? Do you see the metal?
- 44:13Do you see the beard with four views and full samples, the
- 44:18quality is just remarkable. So we wanted to show you this
- 44:22video because a lot of people were curious.
- 44:25One, can you run this locally? So the answer is yes.
- 44:28Can you run it for free? Yes.
- 44:30Can you have multi view? Yes.
- 44:33And what do the results look like?
- 44:35And in this video, he does the entire comparison.
- 44:37I think if you go towards the end, Pratt, we even get a
- 44:40comparison. Yeah.
- 44:41So let's see this, the difference between single view
- 44:44and multi view. And you can see on the left, the
- 44:48hair is a lot more accurate. Like look at the braid.
- 44:50At the back, even the. This is the wrong one because of
- 44:53the single image and this is multi view.
- 44:56So in multi view you get all of the ankles and the way to make
- 44:59multi view is so easy. Just go to nano banana 2 now and
- 45:03just take your front angle and ask it for side views and back
- 45:07views and you get the result. You can go forward.
- 45:10You can even see the textured output and then there's a
- 45:13comparison later on. Wow, this is so.
- 45:21This is really good. This is so.
- 45:23This is much better than the one from Hugging Face.
- 45:25Yeah, yeah, yeah, there. Is there one here?
- 45:28Oh, no, sorry. That one, if you go to his, if
- 45:30you search trellis 2 point O, just search trellis 2 point O
- 45:35and then see a video by this guy.
- 45:37He has the comparison to Honey on.
- 45:40Yeah. Is it honey on or?
- 45:41Honey, on the second one, I don't know which one is the
- 45:47right one. OK, let's go towards the end.
- 45:50Or these are the assets that he's created.
- 45:52Yeah, but this is. Trellis This is Trellis demo.
- 45:56Go towards the end, yeah, you'll see the Angel comparison.
- 46:04Yes, Trellis is much better, right?
- 46:07Yeah, you will see the back, especially with the wings.
- 46:10So the trellis is highly detailed compared to Hong Yuan.
- 46:17Both models you can run on your computer.
- 46:19There you go. So you see the wings for
- 46:21trellis. This is I imagine single image
- 46:25for both and it tried to predict what the wings look like from
- 46:29the back. And you can see trellis has done
- 46:31a much better job. But say you have multi view now
- 46:34you don't need to predict. You can just be super accurate
- 46:38with the results. That's why I'm interested in
- 46:41this because now that we've seen this, we're going to run the
- 46:44multi view on our computer. The setup is a little bit
- 46:47complicated compared to free pick and you know Higsfield, you
- 46:50need to make sure you understand all of the parameters and run it
- 46:53right. So we're going to do that and
- 46:54once we have a pretty decent setup with solid results that
- 46:59are comparable, we're going to bring it back to the show with
- 47:01the screen recording so you guys can see.
- 47:03But so far for a lot of you guys who were curious, we just wanted
- 47:06to give you guys an overview of how you can do this and what's
- 47:10possible. So if you're super in a rush,
- 47:15check out Pixel Artistry. He's got tutorials covering all
- 47:19of this. And yeah, this is fascinating to
- 47:23me because when we generated our first 3D model, I believe it was
- 47:30the last year. Before the Ultimate Unreal
- 47:32Engine program. Yes, with the mask.
- 47:34Yes, and that must have been Hunyon as well, but one of the
- 47:38earlier versions. And I remember looking at the
- 47:41results, it was interesting to see 3D output within seconds,
- 47:46but it still wasn't anything close to what you could use.
- 47:50Looking at these models and the fact that they now have PBR
- 47:53texture, you have multi view, you can get accurate results.
- 47:56I can see just like the image, the pattern of where this is
- 47:59going to go with remesh tools added in.
- 48:02You can remesh them too by the way, which means they will
- 48:05become optimized for your use cases, whatever they may be.
- 48:08I can see more and more people using this in their development,
- 48:12whatever it is. I mean, it could be games, it
- 48:14could be Fortnite games, it could be LED volumes in Unreal
- 48:18Engine failing. You can use it in anything.
- 48:21You can use it in different areas, so look into it.
- 48:23Understand how the tools work. This goes back to the point that
- 48:27we made about Block. The one thing that you can do
- 48:30and why we have this show, is to show you what's happening in the
- 48:33world and talk about what we can do to best prepare ourselves for
- 48:36what can come. And I think the most important
- 48:39thing we can do is constantly staying proactive, understanding
- 48:44what's happening, learning the new tools and seeing how we can
- 48:47implement them to become better at what we do.
- 48:50Exactly. I think we have covered
- 48:53everything for our Friday session.
- 48:56Thank you very much everybody for watching and listening.
- 48:59Don't forget to subscribe if you're on YouTube or follow us
- 49:02on Spotify and X. We will be back on Monday.
- 49:06Yes, and if you guys are watching this right now, you got
- 49:08your phone, don't forget to record a story on Instagram.
- 49:11We share your stories all the time.
- 49:12I'd love to see how you guys are viewing this, whether if it's in
- 49:15your car, whether if it's on ATV in your house or on your laptop.
- 49:19I would love to see that. Once again, thank you so much
- 49:22for watching the show. We'll see you guys on Monday.
- 49:24Have a great weekend.