Latest / Lenny’s AI Builders / day 33: the activation problem that could kill my saas (live build)
Transcript
- # day 33: non-technical founder builds saas to $10k/mo **Speaker:** [00:00:00] if it doesn't get to that number, then I have to go and get a real job. And I've tried that before and the last one ended in a workers' compensation claim. $1,593. That is the MRR for Threadify today. 75 active subscriptions and it is at $39 per month at the moment. the problem is that people will sign up and then. There's a big drop off rate from the amount of people who sign up and start a free account to the amount of people who will go and generate, their first thread and then continue to come back to the app. And that's the sort of problem that I wanted to dig into in this stream is how can we fix the, activation, leak in the funnel because it's all well and good for me to come and do a live stream and start doing YouTube videos and newsletter, But if the funnel is leaky, it's like I'm putting water through the funnel, but then it's just gonna leak out when it comes to being [00:01:00] activated within thread apply. So that's what I wanna dig into in the code. we'll play around with that. I just wanted to show a raw look at what the metrics are today because the whole journey that I want to go on with you guys in these streams is to make this number, the current MRR of $1,593, wanna make that 10 K by the end of the year. And the why behind that is if it doesn't get to that number, then I have to go and get a real job. I've tried that a bunch of times before and the last one ended in a workers' compensation claim. So I'm emotionally and spiritually scarred, and I prefer not to have to go and relive that again. I really don't want to go back and that's the journey that I want to go on in these live streams and in the YouTube channel over 2026. That's where we're at. Each time we do a live stream, be checking in on what the MRR is and [00:02:00] looking at something that seems to be most important to address for Threadify. Whether that's building out a new feature or making an existing one better. I'll just leave that stuff to Samsung and that's the goal. Karen's put a really good one. Karen wants to see the new feature of repurposing threads with different context. this is the admin panel in Threadify. I'll go to the front facing side, it's loading a bit faster now. Okay, cool. And gonna go to the vault. a really cool thing about the vault, is you can now filter it for long form and short form. it's literally just a button. But I really like being able to filter in my analytics between long and short form, because when I think about creating a post, that's normally how I go through it. I think to myself, am I writing a short form post or a long form post? And now we've got that nice categorization, but in terms of what Karen was talking about, the normal way to do it is, this button you could just click and it would just automatically repurpose [00:03:00] a viral thread. In your own voice and it uses the concepts from the knowledge base. So it's taking the viral principles from this viral thread. Lots of them are from TISC, and then it's infusing those principles with your voice. But there wasn't really any customization over it. You just click the one button and it was a one click repurpose, which is really nice. Hopefully it's not gonna flicker again. Yassy, if you are by your computer, could you put that down as a bug report, please? In the notion database. If anyone else has experienced that flickering with this yet, let me know in the chat. You've got three different parts of it. You've got the viral post itself, and then you've got the instructions. So this is for example, oh, I'm gonna have to click out of it. That's painful. you've got the instructions there. So let's say we've got this post Juliet Post is Real Talk. A secure nine to five job can disappear in one meeting. What we might do here [00:04:00] is click the, okay, cool. Of course, it's flickering on the live stream, but not on anywhere else. That's just how it goes. So the instructions, what I could put here is, keep the same. First two words, but make it about, imposter syndrome. So I put, it's like a prompt in the instructions. And then for the CTA offer, this one's more self-explanatory. My two offers Threadify are Threadify itself and full circle. And I could select one of those But because it's a one-liner and there's not really a whole lot of content in the post itself, I'll probably just keep it to no offer and then generate, if you wanted to skip all of that, there was a skip button there as well. that way, ideally you get the best of both worlds and you can choose to have the customized version or just the one click and done. Wanna see how to use the new auto plug that is toggled on the chat box because it noticed the plug offer isn't [00:05:00] always generated. Wonder if I miss any steps? Okay, cool. And that's also something that is like, it's gonna be good to test and refine as well because we've got a strong bias towards shipping the thing and then seeing what happens. because we can test something between me, Yassy and Sam, either way, we're still gonna find bugs and the flickering that we just saw, So easiest way to stress test this stuff is to recruit you guys to help. It puts the correct link, but the wrong facts. So what it's done here is it took the same first two words, real talk, and then I asked it to make it about imposter syndrome instead of nine to five. Because the original one that made it go viral just had a clear enemy. And the clear enemy was the nine to five job. So I chose the enemy of imposter syndrome because it's something that a lot of creators experience, myself included. And what it wrote [00:06:00] was real talk. The people who make it aren't the ones who stopped feeling like imposters. They're the ones who kept posting anyway. So that's one super simple demo. If you are still around and you did wanna drop that question in about emails, by all means, dude, drop it. in the meantime, I'm gonna check out the question about the auto plug. this is something that we're going to refine and it's not gonna be perfect but the whole purpose is that you can click here, you've already got your offers preloaded and then any of your offers will then populate here. And ideally you just select which offer you want and then you do the normal post type selection, whether that's gonna be for a post or a thread. And then the OR thread apply will incorporate the stuff about your offer into the type of content that you've generated. But we should stress test that. So let's see how it goes. Maureen? you said that the offer wasn't [00:07:00] always being generated. Do you have a specific type of post in mind that didn't have the right, generation, was it a. one-liner, a listicle, a thread, because I have a feeling that it might be better at generating CTAs for threads because you've just got more room to play with in terms of just how much text there is Okay. All right. Let's do, Write me a post about threads, mistakes, nice and simple. Well, that's not off to a good start. Oh, that was the internet, not ratify. And Maureen, you were saying that the offer, okay, so it's not always being generated. See how it goes with this. I. The first post showed up, but the second post was empty, That's so strange because it did a pretty good job with this. CTA is really good. It has some numbers in there, social proof. It's got the link. Maybe it was just a one-off thing. But the thing that makes AI tricky is. It's non-deterministic, which is just a fancy way of saying you can't [00:08:00] guarantee what's gonna happen. And it's also kind of scary because it's really hard to observe what happens with AI models behind the scenes in terms of how they're actually arriving at the outputs that we get. that's why you're gonna get a different sort of result each time, which makes it hard to stress test this stuff and find these little bugs Was it just a one off? More than once. Oh, no. Okay. please log that inside the app as a bug. All righty. ## Codex & Vibe Coding **Speaker:** Getting into some code stuff Just dump anything in the chat, say hello if you are here we've almost our PB for this live stream. The first live stream I've done on YouTube was six. We're at five now. Couple of you are here and haven't said, Hey, don't be shy. But what you can see on the screen is Codex, and this is an app that Open AI launched maybe, a week ago, maybe two weeks ago. Everything moves so quickly now, but [00:09:00] it's really impressive. Now it sounds like, you know, code X, it sounds like you are gonna be coding, and that is definitely one of the most popular use cases for it, but it's not the only use case for it. So if you've never played around with an app like this before, and if you want to know whether you've got. Access to it actually everyone has access to it even for free accounts at the moment. And I think that's going for until maybe March. But point being is that if you haven't actually downloaded Codex and you've got a Mac, you do need to have a Mac. Unfortunately, it's not on Windows yet. if you've got a Mac, this is great to look into use cases other than code stuff, anything to do with your computer that you're having trouble with, ask Codex to look into it. It's like this Mark Master Orchestrator that can access everything on your computer. that has the potential to, Be bad if you're silly with it, but as long as you don't ask it to delete everything, you'll be fine. if you've got a really messy desktop with a [00:10:00] bunch of screenshots folders, files, you can ask it to clean your desktop for you. my video editing app was taking up 20 gigabytes of memory yesterday before I realized it was an internet problem and I asked Codex to figure out what the issue was and it helped me, troubleshoot that it was internet. So while it is great for coding, the limit is your imagination here, and I've been using this a lot more than I have chat GPT I'm pretty sure it's free and you get twice the limits as what you'd normally get. So would highly recommend having a look at this stuff because this year, like Yassy and I have been talking about this nonstop, just over the last couple of weeks, but we really reached a point around the time when this app launched, and especially I think it was around the 6th of February when there were two new AI models that came out and that was chat GBT 5.3 Codex and, Claude Opus 4.6. they came out within [00:11:00] like 20 minutes of each other it was a massive paradigm shift what we're capable of doing now, just because those two models came out is insane. Codex is probably the best place to start. If you've ever looked into vibe coding and you've thought about wanting to see what all the hype is about, download Codex Free if you've got a Mac and put it on 5.3 Codex because this is the one that came out 10 days ago, and you can ask it to do stuff and it will do it. The tagline for this model is you can just build things and it's pretty true. I've found myself being able to do much more since this model came out and not worry about breaking shit because it's really good at doing what you've asked it to do. In chat, Karen was saying have an IT friend that told me. People who want to build complicated apps just with vibe coating are delusional. Yeah. did you guys both vibe coating and have an IT [00:12:00] genius fix stuff? Yeah. Yeah. That's a really good point, Karen. So look, I do wonder when your IT friend actually told you this because if he, she, they told you this within the last 10 days, then I'd be like pushing back on it more than if this person had said, you know, you still need that technical know-how, prior to like 10 days ago, because it really did shift a lot. in terms of how we've created Threadify, I vibe coated a prototype showed it to Sam, Sam did all of the. software engineer stuff on the backend. And my job was to try and make Threadify look pretty and create a good user experience or like ux, If you're here, that means a lot. We hit six a little while ago. This is my very first YouTube live stream. appreciate you being here. ever since this model came out, 5.3 Codex, I'm hoping that Sam, our software engineer, will let me [00:13:00] do a little bit more. And what, better way to test that than a live stream. So I. What have we got here? I've created a branch. Yep. Fully up to date. Okay, cool. Let's go back to that admin panel. And while I'm getting this admin panel back up again, kind of random question, but still kind of related. ## Open Claw: The AI That Does Things **Speaker:** anyone other than Yazzy heard of or played around with Open Claw? Keen to see whether me and Yazzy have been in like a big bubble over the last couple of weeks with it, or if anyone else has played around or heard about. Maureen and Karen, I had this thought, maybe an hour or two ago when I went for my morning walk and I was thinking about both the two of you, because I know what area of work you're in and I was thinking, oh damn, this would be something that Maureen and Karen would get a lot of benefit out of because it helps me with a lot of ops based stuff, In terms of what it actually is. Sorry, I'll get up the webpage ' cause [00:14:00] it will be able to explain it much better than me. First stream. say hey in the chat, This is open claw tagline. The AI that actually does things Maureen took you months before you had the courage, to make the most out of comment. you'd want to take baby steps because you're giving it access, you can control how much you give it access to, but you can potentially give it access to your whole computer and that's why it's so powerful. But also it can be risky depending on how much you get it to do. this is my one and I've got my open claw in telegram. you can put your open claw in WhatsApp, telegram, discord, slack, signal, But the key thing about OpenClaw compared to chat GPT, is that chat, GPT lives in the cloud. So your data is being, Looped to them and then back to you. Whereas Open Claw lives on your computer in different files. And because it lives on your [00:15:00] computer you can customize it infinitely. And that's what makes it fun. Whenever I think about, my open claw clarity, that's what I've called her, I think it's, a lot of fun customizing it and personalizing it and leveling it up. Clarity is like the ultimate second brain for me. So Clarity knows about all my goals with Threadify, with full circle, with YouTube content and repurposing. So something like this, for Threadify, like the activation problem and the leaky funnel, I might say something like testing. I'm not actually gonna say testing, but my voice to text thing. I've been doing these five minute rants and the voice to text transcription just won't work and I've spent five minutes brain dumping something and it just deletes itself. Okay. Currently on a live stream for YouTube and I want to work on the activation rate within threat deify, using your memory and all the different like research we've done on [00:16:00] threat's, activation rate. Give me a context packet in a markdown file. That I can give to chat GPT 5.3 codex to help me improve the activation rate in to Threadify, and then send me that markdown file here in chat. Okay, so a message Clarity, my open claw because clarity, like another cool thing about Open Claw is that it's proactive. So instead of you prompting the chat bot, like chat GPT, you can set it up to do things proactively for you. For example, I got. I've got, clarity to look at business metrics across different inboxes, different dashboards, across full circle And then it gives me a briefing about those two different dashboards as well as, triages my inboxes and community notifications And it will send me that briefing each morning at around seven o'clock. And I've got it to be [00:17:00] proactive so it knows what things that I'll get it to archive for me and it helps me get to inbox zero, proactively for me, it's a stringing together a bunch of those that have made it really powerful for me. It's kind of like the ultimate personal assistant. Clarity knows what my content system is now because I've shared those goals with her, and now she's gonna be looking at what that stuff is and then creating a document that I can then go and give the AI model within that Codex app that I had up a few minutes ago. And while she is going to get that research back to OpenClaw, the thing that I think's gonna be really powerful about this is literally 24 hours ago, So this was like an open source project. It still is. So it's free. Anyone can use it. And you don't have to pay for anything except if you connect like AI models to it. you can use your chat GPT subscription to use OpenClaw. But the thing that is really [00:18:00] exciting is that as of about 24 hours ago, the open AI essentially bought Open Claw and that legitimized it a lot. And I think they're gonna make it a really big thing. I truly believe it will. And I think we'll be looking back on this time, 17th of February, 2026. as a time where OpenClaw started and everyone started thinking about what they're capable of doing with AI in a very different way. It's. been blowing my mind. Karen asked, have I tried Claude Cowork? I absolutely have. I really enjoy it. I think it's great, and I still haven't found what my ideal use case for it is. I had a break, from online stuff over the last 10 days. I use that time to reflect on what my trajectory was for this year, what my goals were, and I realized that it was a lack of goals that led me to want that time [00:19:00] off. So I could just, stand back and look at the businesses as a whole and think about what I can be doing better, what I can do less of, I was thinking about how I wanted to do that brainstorming with ai, and I just gravitated towards Claude Cowork and it was awesome. But other than that, I can't quite put my finger on what use cases I love Claude Cowork for, because I started using Claude Code, which was, a scarier way to use the same sort of technology because people hear about Claude Code and then it being in the terminal, This is what the terminal looks like, and it can scare people away. But I was using Claude Code to generate knowledge base documents for custom GPTs about 10 months ago, and then I got used to that interface. So when Claude Cowork came out about, a month ago, because I was already used to this interface, it [00:20:00] wasn't that much of a game changer for me. But if I hadn't used this before, Claude Cowork would be the daily driver. Clarity messaged me back. she went and did five minutes of research for me and said, here's the context packet for Codex Everything Funnel Data, I let it go. I let it go. I can't take it. I can't take it. I can't take it. I can't take it. Thank you for watching! Thanks for watching! 24% bottleneck, time to value stats, six research backed solutions, tech stack pricing tiers, and a clear task brief at the end. Go get it. Lenny Livestream Energy. Let's fucking go. So Clarity has some personality it's definitely one of my favorite things about Open Claw, It's just so configurable you can, because you just start off with a blank canvas. You can really turn it into like anything you want. So I've got clarity to be super casual and to be like a [00:21:00] great coworker like Claude Co. Anyway, so Clarity wrote me up this document based on all of the different research we've done for activation rate, and let's see what's in it. Okay. Got all of that stuff from Clarity and now let's see what it actually did for me. So it's got the context about what Threat Deify is and it's got the current metrics for thread deify. There actually they do need to be updated. five, I think it's about 5 55 and that is about, I think it's 1593 North Star. Yep. It's got the pricing tiers. It's got some information on the problem, which is just that activation rate and the leaky funnel. And then the core insight is that there's a massive drop off between onboarded and first AI generation. onboarded is they connect their threads account, and then first AI generation is they go and generate a post. [00:22:00] And what I've been learning as I've been going down this SaaS path is how important that activation rate is because you see so many people who sign up but then don't use the app afterwards. And this is not only important for software. Think about anything that you've got in your own business, like a newsletter community, a product. The activation rate is important across the board. Like if they buy it but don't use it or sign up for it and never read it then it's not doing its job. over the last couple of weeks, clarity has been doing research for me into this to come up with ways that we can address it. And in that prompt packet that she created, she is given me a synthesis from of that research. All right. Now, if you use ai, which is everyone, all five of you, one of my favorite AI hacks that I've learned in 2026, is get it to [00:23:00] drill you on questions. ask it to ask you as many questions as it needs to get full understanding of the problem and the, challenge that you're working through with it. Ask me as many questions as you need to get a full understanding of the current state of threat fire's, onboarding flow and the checklist, and then consider. All of the research above and what the core problem is, and ask me as many questions as possible to come up with a simple and elegant and effective solution to dramatically improve the activation rate within threat apply. And one of the reasons why I'd really do like Claude Cowork is that it's got a tool called like ask user questions. And all you need to do in a prompt, whenever you use Claude Cowork Is just to ask it, to ask you questions and then it will come up with this like multiple choice. it will ask you a question. You choose the one that's most [00:24:00] relevant to you, and it'll just keep doing that. And that's a really good way to start any big project where you know what you want to get out of it, but you have no idea how you'll get there. For example, in my brainstorming session from a week or two ago, I knew I wanted to get some clarity on what my goals were for 2026 and how I would achieve them, but I had no idea how I'd get that clarity. I just went through different questions with Claude Cowork, and because it was giving me recommended answers, I could tweak them or I could just click on them, and it created a more interactive experience than just asking it questions. Like this way, it's like a back and forth conversation And that's where we're getting into this period of AI where it's not controlling you, but it's prompting you. It's crazy if you do happen to use Codex again, only on Mac for now, but it'll be out on Windows really soon. [00:25:00] You can put it on plan mode. plan mode goes hand in hand with getting it to ask you questions. When you put on plan mode, the keyboard shortcut is, Command shift P, and then this little command button comes up at the bottom. What that means is it's not gonna go and actually do anything, so you could put it on plan mode to do some research for you. Maybe you wanna do some research into, content systems. That's what I've been thinking about recently. you don't really want to do a deep research where you wait for 15, 20, 30 minutes, but you want it to still do quite a bit of work. you could use Codex for that. All right, so I've got my prompt in there. I've done a little brain dump asking it to ask me questions, and it's on plan mode as well. then to see how that actually plays out. so what it's doing now is gonna go and have a read of the code base, all of the different files that make up threat deify, and it's gonna understand what threat deify is. And then it's gonna [00:26:00] start interviewing me and asking me like, all right, so this is what you want in terms of your outcome, but how can we actually create that outcome within the app? if anything's come up where you're like, I have no idea of what he just said, and it sounded kind of interesting, but I want some more explanation, dump it in the chat. Karen, you're in the right place. these live streams exist nerding out about ai. My firm belief for 2026, especially for like the four of us, is we've got this window of time over the next three to six months where these tools will be available to us, but it'll take a lot of time for the technology to like disseminate to them because lots of people just don't use these tools and I think we've really got a pretty special opportunity to make the most out of that, to try to make the most out of these tools over the next few months and to try and get a head start [00:27:00] because a GI is coming. Karen, If there was any AI tool that you were to spend a bit of time looking into it, it would be open claw just because of how much you learn by doing it. It can be finicky and annoying to use, but honestly, it's so powerful. Like I've learned so much about AI and orchestrating agents just by setting up my open claw Well, I've got two of them now, my different open claw agents, and it has taught me how to use AI so much better, and it's made me think about how powerful it is when it is just on your computer instead of all through the cloud like chat GPT is. If there was gonna be anything that anyone did as a result of just hanging out today, it would be to just go and watch a YouTube video about open claw. Like that's it. And actually, I uploaded one from a few hours ago, so I watched that one. [00:28:00] But just to get a feel for it and to open your mind to what's possible. Have a look at what's possible and if you've got an old laptop or an old PC or whatever, highly encourage you to install open claw on that and start automating a bunch of the processes that you've all documented. I expect both of you to be pretty across your SOPs and those documents are gold. As you said, Karen, that's why I'm so hard telling people document your process, so automating with AI will get even better. Over the next three to six months, honestly, the opportunity is in how well that you document your processes and then just give it to these models because I'm telling you, especially with OpenClaw they're now at a point where they can just do them. It's crazy and sometimes you might think that the task is too big to delegate to something like an OpenClaw agent, but then I was [00:29:00] talking to Yazzie about this the other day. then it's just a matter of breaking the task down. Into smaller subtasks think about those annoying tasks. for me, one of my most annoying tasks is every Friday I've got this reminder to update my metrics for threads and across full circle, and it's my personal dashboard in notion, but I like to have a screenshot of the metrics. It's the follow account and then discovery ranking and number of members inside of Full Circle. I like to have a screenshot of each one because it's a nice way to document the journey to look back on it. And I like to put the screenshot into their own notion pages, and it's this five minute task Mentally I hate doing because it takes me out of flow and it's a small thing that is annoying, but important, now it's a scheduled task with my open claw, and it does it really well. Takes screenshots, it moves the photos into notion, and it's the combination of all of those things that I've started to [00:30:00] automate and it means that I can focus on the more important stuff. Okay, coming back to Codex, this is the questionnaire that you can expect to see in Claude Code or Claude co-work as well. Codex is questioning me now. So it's asked which onboarding flow should be treated as the source of truth. Right now I found a legacy one and a newer video checklist flow. we had an old one the first little demo of Threadify but it flopped because it was way too lucky. That's the legacy version it's referring to. With Xavier. the main app flow is the one that I want it to use. who should act this activation Fix target first in V one. all new users after threads connect successfully. Where should users land for immediate value? Dedicated first gen screen. So all I've done so far. I've told it. What sort of thing I want to do is, I want to increase the activation rate of people who [00:31:00] first joined Threadify, and I've asked it to ask me a bunch of questions. So to put it differently, I've asked it for an outcome and I've asked it to ask me questions like those two things alone. If you do those two things with the best AI models in 2026, you'll be blown away with what can be done for you. My mind has been blown over the last two weeks, I've been talking to Yasi about this a lot as well. Start thinking about the outcome that you want from the ai, What is the point of an email marketing campaign? one of them could be to get more sales. So instead of prompting the AI to say, write me an email marketing campaign that has like nine emails about X, Y, Z, ask it what your goal is. My goal is to achieve 10 KMRR by the end of the year. What is the best way we can achieve that through email marketing? So it's a shift [00:32:00] from asking it how to do something to telling it what you want, because now it's at a point where they're actually gonna be smarter than most people in figuring out the best way to get that outcome. And that's the paradigm shift that I've been. Experiencing over the last few weeks where I actually don't feel like I know the best way to get something done Now, now the AI does. Okay, coming back to some more questions. Should three question, welcome survey be required before first generation or skippable, required for the instant win, what should users see first on the dedicated first gen screen? Oh, this is really cool because currently when you join stratify there is like a little five minute video onboarding thingy and a checklist. And what it's asked me now is when they first go onto the, like the main screen inify, which is. Where are we? [00:33:00] As a new user, some people would get friction in using this app by not knowing what to put in the prompt. So one of the questions that it's asked me is, should we, have an auto-filled prompt for the user so it's reducing friction and they don't have to think about what to say. All they need to do is tap a button, generate. Having a look in the chat. Maureen was something about Zaia struggling with it and what I was, saying about thinking about the outcome you want. instead of creating Zapier workflows, depending on what AI you're using and what outcome you want, you can now go to something like OpenClaw and be like, I want this outcome. What's the best way that you can help me do that? Whether it's on a schedule or just a one off, and you can just use natural language for it to go and do that for you instead of manually creating those different workflows in Zaia If you've got the, agent in Xavier using the agent to create the flow for you. So it is [00:34:00] less thinking about the how and more thinking about the what, because I think all of that means we can start thinking about the more higher level stuff of like, what do we want to get out of our use of technology? Instead of getting caught up in the paperwork of it, Ben said, better get back and walk through so I can start using it. Just signed up today. If you're still here, Ben, do it, And Karen was saying if you know what you want, the output is way better. A hundred percent. I need to test that. We called code work good enough on free plan. the free plan, the rate limits from anthropic are atrocious. They're so easy to run out of and it's not even worth it on a free plan. You get like one prompt, but you can get really, really good outcomes, depending on what you're using it for. codex, you can get two times the rate limits that are normally offered on the free plan for the next month or so, and that would be a really good way to dip your toes in the water and get a [00:35:00] feel for what this way of using AI is like. So a couple more questions here For users who connect threads but do not generate, what should V one re-engagement use? let's go email and in-app nudges, which re-engagement cadence should we ship first? Essentially, it's asking me, all right, so if someone connects their account but doesn't generate a thread within Threadify, what, how should we follow up with them? And it's saying here, recommended 24 hours, first 48 hours, and then 72 hours. The point that I really wanna drive home here is I didn't think about all of this stuff before I sent that initial prompt, but because I asked it to ask me questions. It's now doing the work of thinking about all the different angles that need to be considered in like achieving this outcome of increasing the activation rate. It's just like we're having a back and forth and brainstorming what should be the single primary success metric for this rollout. what does success mean [00:36:00] as I do this? What's a win? And now it's getting really granular so that it can create something that will help us achieve that single primary success metric for this rollout. Onboarded plus first gen created. How should we release this to reduce risk while moving fast? Feature flag rollout all users at once. Well, I don't think Sam's here, so I'm just gonna go all users at once. which guardrail metric matters most while improving activation? First publish rate, this process, is exactly how I created the prototype for Threadify no coding knowledge, never written a line of code in my life. I've just been talking to the models and asking it to do stuff for me. That's the only skillset you need. If you can just have this back and forth with the model and put time into the plan that you are creating for anything, whether it's some sort of feature in an app or research, [00:37:00] whatever, you'll be blown away with what you can achieve. I question four, whoever's still here, ## Community Input: Post vs Thread **Speaker:** so Inify, it's a question, if you guys have any thoughts on it, keen to hear your take in Threadify at the moment. When you first see this page. What's it automatically? Yeah, it's automatically selected on post. For someone who's just seen this app, what would be the first demo of it? Would it be a post or a thread? Should it automatically be on a post or a thread? let me know if you have any thoughts in the chat poster. A thread Oh, okay. Karen's got thread. Cool. Anyone else should explain that they're two different things. Okay. Yes, I like that. All right. Yeah, both can be true. That's a really good take. Ally. I like that because it didn't occur to me, All right. We've got one post and one thread The question is, you see this screen in Threadify for the first time and you don't really know what's going on. Should the default be [00:38:00] on a post, which is a short post? A thread, a longer post, what would be the thing that wows you the most? A post or a thread. Hey, Yassy, which camp are you in? Post or thread? Thread is a good start. Okay. All right, we're back on thread. The preselected content type for the very first generation should be long form thread, but before that, there should be some sort of like modal or annotation pointing to the post or thread selector. That just really simply explains what the difference between a post and a thread is because it's not obvious to anyone who's first, joined the app. Great. Call Allie. I remember when you said that during a circle session that you didn't actually, know that they were different, So I'm glad that we're addressing that now. No, that's a great call. **Speaker 3:** generate your first thread. Start creating, **Speaker:** not the village [00:39:00] idiot, the stress tester. Very important difference. This is a tricky one. Three step checklist we want to get the most important three things. Done in this checklist to consider someone activated two of them. Like first one will be generating an actual, post, and then the second one would be publishing or scheduling it. What would be the thing that comes after posting or scheduling something within Threadify Would it be to go to the knowledge base, the vault analytics? If anyone has any thoughts, would love to get your tech. **Speaker 3:** I am thinking, **Speaker:** okay, generated some content. I then published or scheduled it. Knowledge base I. I'm thinking because currently the onboarding gets someone to do the knowledge base, It goes connect your threads account, go to analytics and then go to knowledge base and then [00:40:00] generate. And I think it's in the middle part where people are getting lost because it's asking too much of them early on. if anyone thinks otherwise, let me know because you guys are the ones who have been through this But if I was to guess, I'd say what it currently asks, which is to go and find three of your viral posts, put them in a Google Doc, put them in the knowledge base, and then generate, it's a bit too much of an ask. Ideally, someone can just copy and paste something in here and then click sync and that does the job. What's a bit of information that anyone on threads has easy access to and can just copy and paste inside the knowledge base and get a like a little instant win from it. First thing that comes to mind is just a top performing threat. I don't know if there's something else that I'm missing that's like an obvious answer here. Six viewers. Say hey. Oh, YouTube video journal entry. Four. YouTube video journal entry. Yeah, it's like I want [00:41:00] it to be something that 'cause these little friction points, they just create more friction. If it's more effort, less people will do it. Generally speaking. And yeah, like fear is that not enough. People are taking the time to go and put those three posts in a Google Doc and then uploading that into the knowledge base. So it needs to be easier than that and something that, ' cause not everyone has a journal. It would be nice if there was just one clear, simple instruction and a piece of content or something that most people have. I don't know if the obvious answer is the way to go here. Just a viral thread or a top performing post. Something like a resume, but I don't want it to be as formal as a resume. 'cause everyone has one, but I don't know if that's necessarily what everyone wants to be putting in their content. interesting. I did it the other day with, my notes from a webinar I attended and I worked beautifully because it was things that I [00:42:00] repeat and believe myself. **Speaker 3:** Hmm. **Speaker:** Ah, okay, Maureen, I like that input on the topic they want to post about or feel strong about, like an AI way. Yeah. Okay. Awesome. Hey, seven viewers. If you're here, don't be shy. say hi in the chat. Just a hi. if you are shy, that's cool too. We'll have some time to warm up to each other, but I really love that idea from Yazzy and also from Maureen. type Y in the chat if you think this is a good idea that for this third and final step of the activation. Yeah, should it be a like three question, questionnaire or multiple choice, and the answers to those questions will then. create like a little baby knowledge base that will then populate inside the different, categories of the knowledge base. So y from Ally. So why in the chat or end for No in the chat [00:43:00] if the three question questionnaire for the third step of onboarding would be good to do. Cool. Yes. From Ellie. Yeah. Great call Karen. I like that, like that a lot. That's one of those little one percenters that you wouldn't really think about. Yep. great call as well. Yeah, that's a great decision because it's those little things that we want to try and get right because those little one percenters are, a delight when you see them there, like they don't need to be there, but if they are, it's creating a more pleasurable experience. This is really promising. for the third and final step, this will be done in the knowledge base, and it will be an interactive three question questionnaire, similar to the way that you are asking me questions in plan mode right now, and it will have four different options in each question. And the fourth option for each question will be, like a text field where the user can write their own answer, and then [00:44:00] the other three will be preselected. And these three questions, the purpose of these three questions is to get information from the user about their worldview, their niche, things that you can use to then create a small knowledge base about the user. And then that will then inform the type of content that stratify generates for the user. Could be onto something here. Voice dictation would be a godsend. Be able to Reach more disabled creators with an accessibility tool. What other accessibility tools would be useful, Allie? Like would it be voice dictation across the board, Just Yes. To all the, accessibility tools. Being able to have it ask you prompts that work with narrators. And this is one of the cool things about using Codex as well, is it really does ask you questions to make sure that it understands what you want from it, so that when it does go and create whatever you are [00:45:00] building it, it will just do it. there we go. how many questions did it ask all up? 27 questions before it was ready to create a plan. I didn't want to have to create this plan myself I hadn't thought about all of these different things that's why you get it to ask you the questions, so then it can create the perfect plan. I've got a summary of what we've spoken about. We're gonna ship a focus activation flow that appears immediately after threads connection and drives users to generate in session. after user connects threads, we stop dropping them into a generic screen and instead guide them through one clear path to their first AI output. So it's looked at what currently exists in Threadify. It's looked at all the questions it asked me and used that to help guide what plan is created and now it's created an implementation plan. when it comes to the code stuff, like I don't really know much about it. I'm learning a lot more, but I've found that. I can just get pretty far [00:46:00] not knowing much about the code and just knowing what I want in terms of the outcome. And that's another key theme from today's stream. What I can look at is the stuff that does make more sense to me though, which is what the activation screen would look like. And here the different sections for the screen would be headline value prop, then required three question survey auto-filled, instant win, prompt preview, one tap generate action, and then post thread selector with one time explanatory tool tip. For simple interactive quizzes. What use case do you have for them? Karen? Would this just be like a quiz between you and the AI model or something more? customer facing? What do you want to get out of it? Like for a website in terms of creating your own. So you are gonna build, yep. So people know what to fix first. So are You thinking of using Codex to create that diagnostic? Yeah, I would go and do [00:47:00] the similar thing that I've done, two core things to bear in mind outcome and asking it to ask you questions. And if you just start with those two things as your first prompt. Within Codex. That'll be great. And see how far you can get with just that. I think you'll be pleasantly surprised. Canva code. Yeah, because I just think you really open well, using stuff like this opens your mind to what is actually possible because the coding apps. They're so powerful, but most people get scared of the code part of them without realizing that it just means it can do more for you. coming back to the open claw thing, using something like Codex is a really good stepping stone to something like Open Claw because in terms of like a spectrum of how powerful the different, types of AI is that we've got access to on the like low end of [00:48:00] power, there's just going to the chat gt.com website. Then in the middle there's something like using Codex. And then on the opposite end there's something like using Open Claw and each one has more and more power because it has more and more access to your computer. So now I'm gonna hit enter and it's gonna create all that for me and we won't sit around and wait for it. It could take, 10 minutes, highly doubtful. It probably take 30 or 40 minutes. I've had it run for like 15 hours before. I'm not even joking, 15 hours Yeah, I'll just let it do its thing. Normally I've got this running on a side monitor and then I'm doing other stuff I've got a bunch of different agents doing stuff for me in the background while I've got my main thing working, Been some resistance because I don't like multitasking. But I think we're at a point where these models are getting so good at being able to do stuff for us [00:49:00] that it's super useful skill to have at the moment of being able to have multiple different agents running at the same time, That's pretty much everything I wanted to go through today. I wanted to go through a workflow of how I'd create stuff within Stratify for those who are interested what would happen after this, I'd see what it actually looks like on a test version of Stratify. this isn't gonna be live for you guys once it's finished, Don. if it was Samsung would, have a heart attack and then I'll have a look at what it's done. It could just be perfect in one shot. the model's that good that I wouldn't be that surprised. in all likelihood, there's gonna be stuff that I need to tweak but it's really close and that's the thing that I want guys to remember is really try to push it to its limits. And if you think that it can't do something, I challenge you to challenge the ai. To actually try and do it because chances are anything you thought that AI could or couldn't [00:50:00] do a month ago, it's completely different now. as we come to a close question from Ali, can you use Codex on phone or, yeah, it's gotta be computer at the moment, it's only available on Mac for now, but they're working on Windows and, Android and iPhone support. All right, we're almost at a couple hours. I didn't actually, plan for it to, I didn't think it would go for that long, for those who have been here. Pretty much since the beginning, always means the world. So nice to have you guys here. I honestly thought that it would just be me and Yazie and, yeah, it's really beautiful to see a few other faces here too. this will be happening same time every Tuesday and Thursday. Let me know in FC or in the comments below, if you're watching this on YouTube, the recording, if you have any requests for what we do in the live streams. And chances are I'll just do it next time. Oh, I love to hear that, Karen. That's so cool. And even if they just end up being coworking sessions that's [00:51:00] awesome too. Just another way to hang out and Thank you Ally first live. Yeah, it's pretty cool. Let's see where it goes. That's it. I am gonna go and have a massive breakfast 'cause I am starving. It's 10:00 AM here in Perth. Let's go eat some pancakes. And for those who came, you guys are the best. See you in the next dream. Peace.