Claude Cowork Is Claude Code for Everyone Else

0:00Today on The AI Daily Brief, Anthropics' new co-work tool is Clawed Code for Everybody Else, and today we're talking about why it's potentially a very big deal. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. Firstly, and of course, thank you to today's sponsors, Optimize Lee Zencoder, Assembly and Super Intelligent. If you are one of those folks who would love to pay for the privilege of not hearing the ads and has an extra $3 bucks a month to spend, you can find the ad-free version at patreon.com-aideally-brief, or if you're an Apple listener, you can subscribe directly from the Apple Podcast app to learn about sponsoring the show. Send us a note at sponsors at aideally-brief.ai. And speaking of aideally-brief.ai, one more quick thing while you're there, press the number five or navigate to aideally-brief.ai slash compass. If you are interested in the new strategy compass product we are launching from Super Intelligent, it's basically our agent readiness assessments plus an information subscription

1:02all bundled into a self-serve package. And we are going to start onboarding people over the next couple of weeks. Now, last note before we dive in, this is one of those meaty main episodes that really deserves its own show. And so that's what we've done. We will be back with our normal format tomorrow. But for now, let's dig in and talk about Claude Kowork. One of the universally recognized truths among people in the AI industry is that the single worst-named product is not somehow chat GPT or GPT-52 Pro QXYZ, or they totally out of left field, Nana Banana Pro, but is in fact Claude code. And the reason that Claude code is such a bad name in some ways is that the code part of Claude code in fact distracts from the full possibilities that the tool actually represents. Now, all it takes is a quick look at any especially non-technical person who is really dug into Claude code to see why it's so much more powerful than it seems just from that name alone. Just a couple of days ago on Sunday, Nikhil Krishnan wrote, I've spent the last 48 hours in Claude code. As a non-technical person, it's basically unlocked three very big things for me.

2:05One, the ability to interact with APIs generally. Again, as a non-technical person, one of the big barriers to running the business has been touching APIs. For example, what you can do in Stripe and the Non-Developer Portal versus through the APIs is night and day. Two, the ability to thread things together. Another issue has been threading several different projects we work with together to do cohesive tasks. Zapier gets you part of the way for triggers, but Claude code lets me do way more complex things that touches multiple things simultaneously. Three, run something regularly. Being able to set a script and run it regularly with this level of ease is a game changer. In about an hour, I set up a daily email to myself that tells me the top three emails I need to respond to, based on a priority scoring system we made together that pulls data from a few different places. I know I'm late to this and I'm probably doing things poorly, so be nice to me. But it's been really awesome to dive into this. Back in October, Lenny Richitsky of Lenny's newsletter and Lenny's podcast wrote a piece called Everyone Should Be Using Claude code more. Where he wrote, and I quote, a few weeks ago, I finally started playing around with Claude code and holy crap, we've all been sleeping on it. The key is to forget that it's Claude code and instead think of it as Claude local or Claude agent.

3:06It's essentially a super intelligent AI running locally able to do stuff directly on your computer. From organizing your files and folders to enhancing image quality, brainstorming domain name, summarizing customer calls, creating linear tickets, and as you'll see below, so much more. Now from there, he curated 50 of the most creative ways that non-technical people in his audience were using Claude code. Point being that the non-technical folks who have spent the time to actually go figure out this technology have come away, blown away, and feeling like they are in a fundamentally different era of AI assistants where they really have a capable agent at their back in call. But poor Nikhil, man, 691 likes 111,000 views, only to have all that work just two days later. Be all for not. He followed up on Monday. Why did I learn this all the day before co-work came out? Yesterday, Anthropic announced co-work, as they put it Claude code for the rest of your work. In the announcement post, they put it simply. When we released Claude code, we expected developers to use it for coding. They did, and then quickly began using it for almost everything else.

4:06This prompted us to build co-work, a simpler way for anyone, not just developers, to work with Claude in the very same way. And it's very clear from the testimony of Anthropic team members on Twitter slash x, that this was a very user-behavior-inspired product evolution. Felix Reesberg from Anthropic wrote, Claude code doesn't just resonate with developers anymore. Non-technical people are using it to build things. Technical people are using it for non-technical work. The line is blurring. I'm by far not the first to think about this. Multiple teams at Anthropic have been working on agentic experiences for months. Claude not just as a chat partner, but is something that helps you do real work. Boris Chirney, who editors note, was the creator of Claude, nudged me. Can we take what we've built internally and ship an early scope down version in a few days? So we took a small team, set an aggressive deadline, Monday sound good, and got to work. Boris himself also talked about the inspiration. He posted, Since we launched Claude code, we saw people using it for all sorts of non-coding work, doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions,

5:07recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven. These use cases are diverse and surprising. The reason is that the underlying Claude agent is the best agent in Opus 4.5 is the best model. Today then, Boris continues, We're so excited to introduce co-work. Our first step towards making Claude code work for all your non-coding work. So what are the features? With co-work as opposed to just regular Claude, you're actually giving the system access to local files on your computer. Claude can then interact with those files, including reading, editing, or even creating new ones. This obviously creates a whole different set of use cases. Anyone who's ever dragged more than 10 documents into a chat conversation, only to be frustrated than they told it there are limits, and yes, I know there are different work grounds for that. But being able to give access to everything in a folder that has potentially hundreds of files, think about my massive screenshots folders, for example, on this computer, and you start to see how different things could open up. As the Anthropic team points out, when you're using co-work, Claude also completes the work that you give it, with as they put it more agency than you'd see in a regular conversation. They write, once you've set it a task,

6:08Claude will make a plan instead of complete it, while looping you in on what it's up to. From there, you can give co-work access to connectors, which are Anthropics ways of connecting to data sources, like Google Drive. If you pair it with Claude and Chrome, co-work can then also complete tasks that require browser access, too. And of course, it's also powered by Anthropic's skill system. Now, the more agency that co-work has really is part of the value proposition, they write, you don't need to keep manually providing context or converting Claude's outputs into the right format. Nor do you have to wait for Claude to finish before offering further ideas or feedback. You can queue up tasks and let Claude work through them in parallel. It feels much less like a back and forth and more like leaving messages for a co-worker. So let's talk about how people reacted to it and what they started to do with it. I will say first that there were a lot of especially developed person technical folks that were to use a word underwhelmed. Claude co-work didn't really add anything new, but they were of course not the actual target. Olivia Moore for A16z gets it when she writes, watershed moment for Claude and Anthropic. Claude has always been a fantastic product

7:08with arguably the best-agentic models, but is struggled to make these advantages usable for the mainstream consumer. This feels like the first concentrated effort to change that. Greg Eisenberg says, said this to every normal person you know because normal people don't want to touch a terminal and this will make them 100x more powerful and productive at whatever it is they do. Indeed speaking to how much more inviting this user interface is, Avthar shared a screenshot of the new Claude co-work on his desktop saying, let's knock something off your list with a highly action oriented UI that suggests creating a file crunching data, organizing files making a prototype prepping for the day or sending a message. He wrote, even as a Claude code die heart, I'm excited to play around with it more. So what have people been doing? While Kat Woo who works on Claude code and Anthropic wrote, I've been delegating a lot of my tasks to co-work. Today, co-work filed a workplace ticket for a coffee spill, prepped me for upcoming meetings, looked up Salesforce data via Chrome tool and helped me buy a new down jacket. Lenny Richitsky has been trying out a number of different use cases. In one, he said, I asked it to go through every Lenny's podcast episode and pull out the 10 most important themes and lessons for product builders. Then the 10 most counterintuitive truths.

8:09I gave it access to a folder with 320 transcripts. First, it said, this is a substantial task, 320 podcast transcripts to analyze. 15 minutes later, he had his list. So I think it's worth pausing and double clicking on this. This is, of course, a type of use case that lots of people use existing tools for. In other words, taking the context of one set of content and turning it into other content. The difference here is, of course, the scale. Lenny's episodes are dense, they're long. 320 transcripts is going to be between 450 and 600 hours of content, if not more, which is not something that would have been easily possible with other types of tools. In many ways, a lot of the core and differentiating use cases can be summed up as stuff that needs access to local files on your computer. As off menus Hunter Hammond's wrote, irony is every productivity app pouring billions into mediocre AI implementations just for local first files to become the chosen one. So what else worked well? Joseph on X writes, co-work is magical, seeing the context that on the artifacts on the right side feels unreal. I got a flight canceled for no reason and Claude is filling out all my claims. There are lots of compliments about the user experience.

9:11Paper clippers writes, just started testing it now, but I'm really liking the Claude coworker you eye so far. Really nice for doing research and quickly turning it into files for later reference. And yet, it didn't take long for people to remember that this was a research preview. Brian Levin writes, one of the featured workflows for Claude cowork is to help you plan your day and week. This is the default experience. Feels like it didn't fully bake. And he shared a screenshot of cowork just not quite being able to actually use the connectors, like Gmail, Google Calendar, and GitHub, even though technically they were connected. Maho Mohan writes, just tried Claude coworker and it's basically unusable for me. Claude did the work in two minutes, coworker is stuck on the screen for multiple minutes. As Ed from Stride pointed out, however, considering they said it was a research preview, I don't think anything should be considered fully baked. And indeed, the team behind it made that point clearly. Back to that post from Felix, he writes, we're releasing coworker early. It has rough edges, but figuring out what to build is increasingly the hardest part of software engineering. And we think getting feedback early and hearing what users actually need is how we build something truly good.

10:12Now, outside of just things breaking, there are some concerns that people have. By and highlight jokingly summed up, Anthropic is constantly pushing the frontiers of giving an LLM route access to your computer. And indeed, the risks here were part of Anthropic's announcement. In their announcement post, they write, in cowork, you can choose which folders and connectors Claude can see. Claude can't read or edit anything you don't give it explicit access to. Claude will also ask before taking any significant action so you can steer or course-correct it as you need. That said, there are still things to be aware of before you give Claude control. By default, the main thing to know is that Claude can take potentially destructive actions such as deleting local files if it's instructed to. Since there's always some chance that Claude might misinterpret your instructions, you should give Claude very clear guidance around things like this. Now, when folks went and looked how this was executed, some of their concerns were alleviated. Ben Hylac actually reposted himself and said, spoke too soon it's running in a VM. To be clear, there are many, many risks, especially when you're talking about giving access to normies, but strictly speaking safer than Claude code. Simon Willisson and his blog post about it said, andthropics say that cowork can only access files

11:13you granted access to. It looks to me like they're mounting those files into a containerized environment, which should mean we can trust cowork not to be able to access anything outside of that sandbox. However, even with that, there's another concern. In a section called the ever-present threat of prompt injection Simon writes, with features like this, my first thought always jumps straight to security. How big is the risk that someone using this might be hit by hidden malicious instruction somewhere that break their computer or steal their data? He points out that anthropic actually discuss this in that announcement as well. Writing, you should be aware of the risk of prompt injections, attempts by attackers to alter Claude's plans through content it might encounter on the internet. We've built sophisticated defenses against prompt injections, but agent safety, that is the task of securing Claude's real world actions, is still an active area of development in the industry. One of their suggestions to minimize risks is to quote monitor Claude for suspicious actions that may indicate prompt injection, but as Simon points out, I don't think it's fair to tell regular non-programmer users to watch out for suspicious actions that may indicate prompt injection. Now, ultimately, he concludes that anthropic are being honest with their warnings, and that they can of course provide total guarantees.

12:15One of the other concerns for some is the cost. Right now, cowork is only available in research preview to people with max accounts which started a hundred bucks a month. Nick Torres writes, all right, coworker looks amazing and extremely helpful, but how does the company of 150 plus actually implement it without immediately paying $100 to $150 a user out of the gate? How have others started training users on it? Now, I think this is a totally reasonable concern, but I also think that the max plan gating is very much about the rollout of this product rather than an ultimate intention. In other words, I think they needed to have a relatively small set of people be able to test it, and the max plan subscription provides as good of a filter as any, especially considering that those are people who are going to be a little bit more adept with or at least engaged with Claude in general. One of the best summaries of cowork overall came from Claire Vaux, the host of the How I AI podcast. She wrote an article on Twitter called Who Is Co-Work For Exactly? She talks about what exactly it is, providing a slightly more technical look at all the different pieces of the system, and reaffirms that this is an action-oriented tool, saying every chat is now a task focused on doing a thing. She writes, you ask cowork to do a variety of tasks

13:15and it takes a very Claude code approach to executing your task. For her, she tried task one, prepped me for my day, which was one of the starter prompts in cowork, and unfortunately had some issues. Claire writes, I didn't have the Google Calendar connector set up at the time, but no big deal, right? There's a connector button right there. Unfortunately, Co-Work couldn't recognize my Google Calendar connector once authed, and refreshed, and restarted, and so it tried to use Claude Chrome to open my browser and view my calendar. I banged my head against this problem a few times but ultimately gave up on this task. Co-Work seemed to know that it could call tools, but for some reason couldn't connect. Womp, Womp. Task two was to make a competitive research brief and she said this one went a little bit better. She writes, after clarifying my needs, it made a basic plan to do some web research and use the DocX skill to make a doc. It took probably five minutes to run the whole thing. Task three was to make a presentation from a document, which it did proficiently, although she thought that the UI suffered from showing her too much about what was going on in the background or as she put it, how the sausage was made. Ultimately, she thought the UI was a nice upgrade, but a lot was broken. On the main question, though, of was this meaningfully better than just doing a straight chat,

14:17she said despite its flaws, it did create better outputs than straight chat. Still, she ended with an interesting question. Who is this for? And her answer was that she's not sure. Claire writes, I can't imagine a Claude Max user on OSX who knows WTF2 do with an agent and also would prefer a limited desktop app experience versus loading up the terminal. I understand that it's just in research preview right now, but the overlap between Claude code and cowork users right now is probably a circle. The challenge with this sort of thin wrapper on Claude code UX is that it's not quite optimized for the non-technical and too kneecapped for the TUI build. Co-work right now sits in the fuzzy middle and the team is going to need to optimize for one of the other to win over a new audience. Most marketing teams aren't short on ideas, but what they are short on is time. And that's exactly what optimizely Opal gives you back, with AI agents that handle real marketing workflows. You know, like creating content and checking compliance, generating experiment variations, personalizing user experiences, analyzing pages for GEO,

15:18even tasks like approvals and reporting. It's your AI agent orchestration platform for marketing and digital teams, plugging seamlessly into the tools you already use, handling the boring busy work, and keeping everything on brand. That leaves you marketers with more time to do your actual job. See what Opal can automate for your team by signing up for a free enterprise agent at AI Workshop with Optimizely. Learn more at optimizely.com slash the AI Daily Brief. If you're using AI to code, ask yourself, are you building software or are you just playing Prompt roulette? We know that unstructured prompting works at first, but eventually it leads to AI slop and technical debt. Enter Zenflow. Zenflow takes you from vibe coding to AI first engineering. It's the first AI orchestration layer that brings discipline to the chaos. It transforms freeform prompting into spec-driven workflows and multi-agent verification, where agents actually cross-check each other to prevent drift. You can even command a fleet of parallel agents to implement features and fix bugs simultaneously. We've seen teams accelerate delivery to x to 10x.

16:20Stop gambling with prompts. Start orchestrating your AI. Turn raw speed into reliable production-grade output at Zenflow.free. If you're building anything with voice AI, you need to know about assembly AI. They've built the best speech to text and speech understanding models in the industry, the quiet infrastructure behind products like granola, dovetail, ashbee, and cluelie. Now, as I've said before, voice is one of the most important modalities of AI. It's the most natural human interface, and I think it's a key part of where the next wave of innovation is going to happen. Assembly AI's models lead the field in accuracy and quality so you can actually trust the data your product is built on. And their speech understanding models help you go beyond transcription, uncovering insights, identifying speakers, and surfacing key moments automatically. Its developer first, no contracts, pay only for what you use and scales effortlessly. Go to assembly AI.com, slash brief, grab $50 in free credits, and start building your voice AI product today. Today's episode is brought to you by Super Intelligent.

17:21Super Intelligent is a platform that very simply put is all about helping your company figure out how to use AI better. We deploy voice agents to interview people across your company, combine that with proprietary intelligence about what's working for other companies, and give you a set of recommendations around use cases, change management initiatives, that add up to an AI roadmap that can help you get value out of AI for your company. But now we want to empower the folks inside your team who are responsible for that transformation with an even more direct platform. Our forthcoming AI strategy Compass tool is ready to start to be tested. This is a power tool for anyone who is responsible for AI adoption or AI transformation inside their companies. It's going to allow you to do a lot of the things that we do at Super Intelligent, but in a much more automated, self-managed way, and with a totally different cost structure. If you are interested in checking it out, go to AIdailybrief.ai slash compass, fill out the form and we will be in touch soon. I think this is a super on point critique, and perhaps super to be expected for this stage of the release. I think Claire's right,

18:21that they're going to have to lean even harder probably away from the cloud code interface and into some real serious regular value use cases to attract people who aren't just going to upgrade to cloud code right away anyways. For me myself, I do everything in the cloud, and so I'll try to have this trying for my own personal interest. A lot of the use cases that people are getting value from don't super resonate right out of the gate, although certainly seeing the types of things that Lenny is doing with all of his transcripts could inspire some new activities. But I certainly think we're at the very beginning stages of figuring out exactly what this particular slice of the AI assisted experience is useful for. Riley Brown referenced this saying, not convinced the best agent tools for work will run locally on your desktop. I think they will run in virtual computers that you can share with your team. Now here's really important context for all of this though. As people give their reviews, they figure out what's broken, and they share where they think things need to go for this to find its right home with its expected audience. This product was built in the last week and a half. When Thursday, AI's Alex Volkhove wrote, I wonder how much of this was cloud code.

19:21Boris Charney confirmed that basically all of it was written by cloud code. In other words, this thing that we're excitedly talking about and having a bunch of ideas for what it could do in the future was a week and a half long project entirely written by AI as today, Jipora put it, insane velocity. So what are the implications? First of all, whenever we get a new product from the big labs, everyone jumps to how many startups it kills right away. And Sean Wang aka Swix does point out that Anthropic took a different approach than many startups in a way that's currently looking pretty good. Swix writes, so many ambitious startups making the LLMOS tried all these fancy UXs and failed. So many ambitious startups making the AI browser tried to book your flights for you and failed. Meanwhile, cloud code started unpretentiously as a CLI and can now run your browser and operate your system. Classic, disruption theory. And yet, of course, Sean is not implying that that means that all those startups are dead. No, for that, we have to leave it to other people. However, at this stage, this nature of reaction has become so commonplace that actually the first thing we get isn't the people saying,

20:22look at how many startups it killed, but instead first explaining why it didn't kill all those startups. Conor bread and Burk writes, lots of posts saying cloud coerced just killed hundreds of startups. By the way, editors know that I actually couldn't find any, which is why I'm saying that we've now shifted and this becomes the average response. But for the sake of the argument we continue with Conor, they're wrong and here's why. We've seen this narrative before. It makes a clean, simple story, but it's not how markets actually work. Open AI DevDay 2023, text-to-speech launch, people said it would kill 11 labs in similar startups. In reality, demand for voice exploded and the category expanded into more specialized tools. 11 labs is doing better than ever. Chad GPT Enterprise, people said it would kill B2B AI startups. Reality at unlocked budgets made AI procurement safe and created demand for everything it didn't cover. Big labs ship a default experience. That gives massive free education and proves the behavior actually works. Users try the flagship that immediately ask for more. Vertical focus, deployment control, privacy, multimodal, deeper integrations. My prediction, 10x more workplace agent startups next year because of cloud co-work. So at least for now we have a bet that the startups aren't going to at least immediately die

21:23and in fact even some, who think that this leads to an expansion. Now another implication of this is around the productization phase of AI. Austin all read represented lots of people when he tweeted, I'm a little confused at the hype around this. Isn't this exactly what cloud does already and has done for a long time? Is this just marketing to let non-technical people know they can use it too? I must be missing something. And the point of course is that he's not exactly but UI shifts aren't just UI shifts if they open up entirely new categories of use cases for entire new sets of users because they wouldn't interact with the previous UI for some perceived real technical barrier. Robin Ferris writes, finally, cloud co-work is out and it's basically nothing new. I've been using cloud code for anything on my file system for a while now, but cloud code makes this more user friendly now and more accessible to the broad. Simon Willison included an entire section in his post called, isn't this just cloud code? And the answer is basically, yeah, but as Simon writes, cloud code has an enormous amount of value that hasn't yet been unlocked for a general audience and this seems like a pragmatic approach. Prince writes, as to cloud co-work, I personally am very excited

22:23about its ability to search my local files, which seems to come with a nice UI. Yes, I know that cloud code can do that too, but the percentage of lawyers who will be comfortable with using a command line interface is probably less than 1%. And so nice UI is critical. Nice UI's will eventually win over us non-technical people. Now, another implication that follows from all this is a continuation of the pattern that we've been talking about a lot as we transition from 25 to 26, where a lot of 26 for non-technical folks is going to be speed running through a process that developers went through last year. Leanne Zabretzky writes, welcome the non-developers to using cloud code as your coworker. We've been doing it since the beginning of December. Feel so distant now. Nicholas Cole writes, whenever I see stuff like this, I think about when my wife worked at this huge nonprofit and their training instructions for how to use the internet included, make sure both feet are planted firmly on the ground, sit upright in your desk chair to not hurt your back, hold the mouse tight with your hand, don't let it fall. Anyone on this side of the internet thinks we're living in the future, and we are, but everybody else is still in the Stone Age. Most companies don't have a single no-code Zapier automation running, and most everyday consumers don't know

23:23how to set up a basic filter inside Gmail. All of this new tech is super exciting, but don't forget 99.9% of companies and people still need help solving way, way, way simpler problems. Claire Vogue, and I think, takes this in an important direction. When she writes effectively that UI, as much as it is a part of this, isn't enough, that this is in fact an entire product design process, which is why productization is so important. Claire writes, the problem with AI products for normies is that too many technical people have seen the AGI God and think everyone will want this, and it must be a UI issue. But guess what? No, they don't, and no, it isn't. We shouldn't be asking how to make agents more accessible or the terminal less scary. If you want to increase the tam of an AI product, you need to either A, discover existing pain points, or B, inspire the art of the possible. Remember, existing pain points have to be real, not just better, but OMFG. How would I do it any other way? This looks like, I have to do this for work and you've made it 10x faster and less painful. My computer is slow and broken, and now it works. I've been working around this thing for ages, and you made it go away.

24:24I'm too tired or busy to learn this, and you taught it to me in two minutes. Once you find these problems, don't give users a general platform. Give them a button to a golden path. Your mom is not going to spontaneously vibe code a family photo sharing app. Your uncle is not going to reach for a coding agent to debug the washing machine. Your kids aren't allowed to write with AI at school. You need to hit users over and over and over with good ideas and use cases, plus your brand, so when they finally have a novel problem, you're top of mind as a solution. 99% of people do not want agents or models or primitives or skills or artifacts or file access or tools or connectors or MCPs or APIs. They want to not get fired, save time, make money, not be annoyed, entertain themselves, express themselves, and feel good. Start there, then build UI. Now another interesting implication is around the competition between the labs. This isn't something I'm seeing a ton of discussion on, but I think even Mollock made a really salient point when he wrote, Claude's gaps in image creation and voice became less important when cloud coaches use his APIs to go to Nana Banana or chat GPT image gen whenever it needs a picture made, or 11 labs API whenever it needs a voice. The anonymous I rule the world account responded,

25:24who'd have thought just making the AGI would be the thing to do? Now even if we are in early days when it comes to shifting a new set of consumers onto these types of tools, I think people do have a sense that we're opening up a new era for normal, non-technical users. As Dan Shipper did, Async AI for the rest of us, or as John McBernie intimated, the Jarvis area for the Broad public, referencing a post from Dan Shipper, John writes, Claude Coden, a trench code, I think perfectly describes the most interesting opportunity that I felt in the last six months. Claude represents a paradigm shift, the move towards a personal AI assistant like Jarvis from Ironman, and the packaging of AI and tools in this way is a key product insight that I think will define the next big thing. Now some are betting that all the other labs are going to have to follow suit. Lord of a few writes, this is a moment in time. Open AI will be close behind, followed by Google. Microsoft will do something similar and distribute to billions instantaneously. Apple will incorporate a similar technique. And like that, the entire world economy will shift over 12 months. The promise of what has been told to you for the past three years is here, and it's this ergonomics, agents on your Mac and PC,

26:24operating on files like how you work. What a time to be alive. Stop scrolling and start building the future. By the way, when it comes to building, one of my strongest arguments is that something pretty fundamental has shifted over the last few months. I talked about it a few times on this show, where all the developers and super-enfranchised AI users went back over the holiday, actually took the time to vibe code a bunch of side projects and realized that AI and agent decoding had gotten so good that they could do way more than they thought was possible before. Now, even though there are still some holdouts right now, many holdouts, in fact, who still try to draw a thick line between vibe coding things that you can use for not production and prototypes and non-vibe coded things that are ready for production, I am firmly in the camp that that line is basically already gone. And if not, it's going to explode in the next several months. The Didas Menlo Venture seems to agree. After Boris shared that all of this had been written by Cloud Code, Dede writes, today was proof that fully polished new product can be built entirely with vibe coding tools like Cloud Code. Many people, especially in big tech, who still don't believe it.

27:25You haven't tried the latest models. You don't have the right setup. Your code base is still too large. Your code style is not well represented in the training data. All four of these are entirely self-imposed problems. You will be outrun on product velocity if you aren't adopting these tools. This is something that I'm going to explore in much more depth in and around this show throughout this year, so keep your eyes peeled for that. And in the meantime, if you do happen to have a max account, go check out Cloud Codework. Once you do if you have ideas for how to make it work, share them with the team. One thing Anthropic is doing really well right now. And in fact, I think you have to give credit to all the major labs for is they are all floating all around X all the time actively engaging with and listening to the suggestions of users. Felix Reesberg of Anthropic even requests to be tagged in that feedback, saying we want to iterate very quickly and make it a little bit better every day. Ultimately for all of its faults and challenges, I think this is a huge step forward. Goodness knows if you're listening to this show, you probably have felt like people have been beating you over the head with how good Cloud Code is for months now. And with Codework, there is increasingly going to be a UI

28:25for everybody who's not a developer to try to get all those benefits. I will certainly continue to be trying it out and sharing it as I figure use cases out. For now though, that's going to do it for today's AI Daily Brief. Appreciate you listening as always. And until next time, peace.