Vibe Coders, Microsoft has a Message For You (Worth Listening to)
May 6, 2025
Today’s daily update blends a no-fluff take on a Microsoft productivity lesson with real-world moves in the AI tooling space, then lays out practical paths for learning deeply and practically as a developer.
Key takeaway from the Microsoft advice#
- Go one layer deeper in whatever you build with AI. If you’re a frontend, dive into the underlying concepts; if backend, push into the next layer of understanding. Ask AI to “dig deeper” and explain what’s not immediately known.
- Keep learning actively and consistently. Curiosity plus a structured path beats passive vibing through your day.
OpenAI deal and ecosystem shifts#
- OpenAI bought Wind Surf for about $3B; Cursor raised $900M at a $9B valuation in the same window.
- Likely impact: developers get better access or discounts on tools as the ecosystem consolidates; potential changes in model usage terms to protect market share.
- Other ecosystem signals: Anthropic + Apple collaboration and a new toolbar concept, plus model explorer ideas. These point to faster, more integrated tooling for developers.
Learning by digging: monorepos, polyglots, and pragmatic exploration#
- History and forms: Google and Meta pioneered large monorepos; Piper-like systems spread to Microsoft, Uber, Airbnb, Twitter; TurboRepo, NX, PNPM workspaces are popular tools to manage multi-app codebases.
- A practical learning method:
- Define your context (e.g., a TypeScript frontend with Next.js, a thin backend, and an interest in FastAPI for agents).
- Use independent research tools (examples: Grok, Gemini Deep Research) to learn history, benefits, and trade-offs.
- Find real-world codebases to study (starter projects or portfolios that use monorepos with multiple apps and services).
- Consume creators who break down concepts clearly (e.g., TanStack Router vs Next.js patterns) to build intuition.
- Embrace “learning in public”: follow peers who share their learning journey (e.g., someone iterating on LeetCode or a learning path) to stay motivated and get feedback.
- GitHub and starter projects are your labs. Pay the idiot tax—tackle friction and gaps head-on to move faster.
The self-healing AI-assisted development concept (practical brainstorming)#
- A concrete, agent-driven loop:
- Bug finder → Trace reader → Fix agent → Test runner → Deploy agent
- Tech stack in view: Deno for backend, Python agents, Prometheus for telemetry, Grafana for observability, TanStack Router and Vit for frontend.
- Outcome: a self-healing, AI-assisted scheduling app with rich telemetry and an agent-driven backend that can detect, diagnose, and patch issues during development.
Learning in public and practical examples#
- Learning in public works: peers sharing the process helps others ramp up faster; examples include constant practice like daily coding/LeetCode videos.
- Use real codebases and GitHub projects to practice and reveal gaps. The faster you confront gaps, the quicker you grow.
Quick prompts and community notes#
- If you want prompts for product management or a tailored prompt set, ask me—I can generate them.
- The community offer: half off now until the new platform launch—worth noting if you want to dive in.
Actionable takeaways#
- Today: pick a single layer you work in and commit to going one layer deeper with AI help. Ask for explanations, edge cases, and deeper diagrams.
- Start a learning log: document what you learn, what questions you still have, and what you’ll explore next.
- Explore monorepos with a real-world lens: pick a project, map its repo structure, and compare tools (TurboRepo, NX, PNPM workspaces) to see what fits your stack.
- Sketch a small, auto-debug loop concept for your stack using Deno/Python, Prometheus, and Grafana to practice observability and incremental improvement.
Links#
- OpenAI Wind Surf acquisition and the Cursor funding round
- Anthropic + Apple collaboration (tooling signals and model explorer ideas)
- Monorepo history and tooling references: Google Piper, TurboRepo, NX, PNPM workspaces
- Independent research tools (examples referenced): Grok, Gemini Deep Research
- Deno (backend), TanStack Router, Vite (frontend tooling)
- Prometheus and Grafana (telemetry/observability)
If you want prompts tailored to product management or a concrete learning plan based on your stack, tell me your stack and goals and I’ll generate a focused prompt list and a 2-week learning path.
Transcript
What would your advice be to juniors, senior developers, and even enterprises on this whole AI thing if you were to leave them with one thought to brew on or one actionable? I've said this before, I will keep doubling down on this. Take whatever layer that you operate at as a computer scientist. Yep. Could be React, it could be a front end, you could be a back end. And go one layer below your comfort zone. Pop the hood, change a tire, drive stick shift for a day. you will be a more valuable and a more helpful and a more educated developer if you just maybe ask your AI to dig a little deeper and help me understand that as opposed to just vibing your way through your job. So you're saying take a little bit of time out of your day to pursue what's not immediately known right to you as a as someone. Yeah. Just a couple of feet deeper in the pool. A little bit where you can't quite touch your the bottom of the pool, but you're not going to drown. Never stop learning then. That's what I'm hearing. Yeah. So, I absolutely love that message and it is from the productivity summit where a bunch of people at Microsoft spoke, but that is so important and it's also fun. Like, we have the ability to just dig so much deeper than ever before. Don't have to go to a library. If we went back like 20-ish years, you'd have to do that. and you don't have to dig around on the blue links or Stack Overflow. So, we're gonna talk about some examples of that. And this is a daily upload. So, of course, we have some news where OpenAI purchases Windurf and what does that mean? What do I think is going to happen there? We're going to go through some comments if there are any. And then I'm going to talk about that advice from the Microsoft team and then what that looks like in practice. So, first of all, the Open AI purchase. I came across this last night and I was making some posts about it to the squad and basically they bought Winerf for three billion and then Cursor Quinc or at the same time raised 900 million at a$9 billion valuation. I guess that's just bananas. What does it mean for us? And if you're watching this in the future, then comment if my prediction is correct in the future. But I am predicting and this is not a wild prediction. It just seems accurate or seems based on the clues we have that all the developers using these tools are going to get deep discounts on usage because they have billions of dollars to spend on market share. So, I think Windinsurf's going to do some sort of collaboration with OpenAI Plus users. I think that they'll get discounts, maybe free for certain models, and I think that Cursor will have to change their model. I think that there's just going to be something big there. I also think that if Windurf has restrictions on model usage, then that's going to be huge. I don't know if they'd go full on and say you can't use these others because that would affect their market share. But things are heating up in this space. We already saw that Enthropic and Apple are working together which I'm really excited about because that means iOS development is going to get better. That paired with the fact that they are reducing the fees is a big one. They also have this new toolbar. If you follow one of the designers, you'll be able to see early stuff like this. And this mimics the same pattern as the Versel toolbar or the Tanstack dev tools bar. I guess the button. And there are some new explorations going on with the model explorer. This is just a concept. I don't think they would have these tags, but pretty cool. And yes, it does look very similar to the Versell toolbar. I think this will probably ship where you can then click to screenshot which then removes the need for an browser MCP and maybe the ability to select a particular component in the tree. So very exciting stuff there. Now let's talk about the thing that the guy at Microsoft said. What does this look like in practice when I'm learning something and I want to dig a layer deeper? I have never done a I've never shipped a production polyglot monor repo. And so polyglot just means multiple languages. Monor repo means multiple apps in one. And there are a lot of advantages to having a monor repo. And so what I do is I just go on an exploration because I want to learn. I want to get more competent. So, I'll just start typing in monorreo benefits and I'll watch a bunch of videos and then I'll see is the person qualified that's that I'm listening to. So, in my case, I found someone from Microsoft that was talking about it and so I watched a bunch of his videos, but then I could also go and use an independent research tool. So, in this case, Grock and I want to say what's the history of monor repos? What's the benefit of using monor repos versus the harm? What are some of the build tools to manage monor repos? And I like to give the context of where I'm coming from. So I might mention something along the lines of I'm a Typescript dev primarily working on front-end applications with thin backend using Nex.js. So thin meaning what it sounds like. I don't really think I don't consider Next.js JS it is a full stack web set like a web meta framework but it's thin in that it's not truly a backend like I want to I want to learn how to set up fast API because I believe that's the best choice for agentic stuff and I don't want to rely on the TypeScript ecosystem. So this is the path that I'd go down. And that would be the case for monor repos. But then in another case, I talked about selfhealing code bases the other day because one of our members who's a total G, he was talking about this in our webdev channel and it got me to think where if I just scroll up a little bit. Was it in general or webdev? Yeah, he posted or he sent me this video on Graphana and he is a 25-y year IT consultant for like CIS CIS admin stuff and I wanted to learn more about that. We had used Grafana at delivery dudes but at the time I was doing zero development absolutely zero was just product design and product management. But it just led me down this rabbit hole where like I see a piece of information. I see a few articles about open telemetry, how to run these things. And then that led me down this learning path of okay, this is water that I am not touching the ground on. Like I am so out of my element, but it led me to follow that interest and then ask a lot of questions on what that would look like. So, I used chat GPT in this case and I basically speced out what it would look like. Now, I'm not building this because I don't have the bandwidth yet, but this will be something I do in the future and it uncovered a lot of different things that I could do. So, this simple thread, I highly recommend stuff like this where it's like, I'm bullish on the web and using coding agents, but realize the that the loop of getting a bug then pasting in the errors is manual. I want to work with something faster. I'm evaluating Dino for web since it has telemetry built in and then I pass in their docs. I'm not familiar with using Dino as I'm coming from next and was starting to use tan stack router and beat. The idea would be that we have Prometheus attached to monitor things as they go wrong. Then we can automatically have agents go and fix bugs. Then the agents that run the code again. Then the agents run the code again. Can you help me brainstorm this? So, I'm describing an auto debugging loop. That's a better way of putting it. It gives me this system overview, some brainstorming on it, a bug finder, a trace reader, a fix agent, a test runner, a deploy agent. And we just follow that curiosity path. And now I say, I'm confused. Help me out. Blah blah blah. But this conversation goes on and on and I have a lot of pointed questions. And then by the end of it, I have this nice doc that explains what this could be. So you're building a self-healing AI assisted scheduling app with a rich front-end guey, full telemetry, and an agent-driven backend that can detect, diagnose, and patch bugs during development. The system uses Dino for the back end, Python agents, Superbase for and billing state, tanstack, router, and vit for the front end. And it's awesome because then you get this file tree comparison. you get the behavioral patterns of how it would work, the telemetry and observability and I could continue to go on and make this better. But that's an example of following that path and it comes from interest. And then what I love about the community obviously is just there's people way smarter than me. I always say that like I am not your guru. I'm just someone that is curious and follows it. If you want to stay in Typescript world, I'm pretty good there. If you want to talk product management, I'm really good at that. But yeah, I just follow it. And I think that's the beauty of what we can do nowaday you can get to competency pretty quick. So history of monor repos, I can see that let's see Google and Meta were the ones that went after it early. Their repository grew to over 80 terabytes and handled tens of thousands of daily contributions. Then Microsoft, Uber, Airbnb, and Twitter all came on board with it using Google's Piper. I can see some of the build tools, Turbo Repo and NX. So, I was evaluating Turbo Repo, but I think that I'm going to need to use in my case probably just PNPM workspaces. And yeah, you just keep following this path. You keep asking questions. I use independent research tools like or Gemini deep research for something like this because that's what it is. We're doing the R&D and you just keep going and going and then I'll find examples of code bases. So in my case, let's see. Not used to having wind surf open, but I wanted to prep it. I wanted to get my extensions and everything lined up just in case they do free OpenAI. So I will make a video after this directly after this on my setup of what are all the extensions that I use and whatnot. And within here what I did was I found some code bases that could be helpful for examples of monor repos. So I found this one and this is someone's portfolio but their portfolio is a monor repo and it has API and it's using fast API I believe it has different apps. So a web app as well as an expo app, the mobile it has some other interesting stuff like these package setups and more importantly just the way that the codebase is set up is something that I can learn from. So I can see that he has Google package with things in here around Gmail and different services that it can do. So it has the routes in here and then I didn't know what SQL alchemy is because again I'm coming from a different language. But what I can do then is yes yet again ask lots of questions. So you can see I wanted to know what the setup was. This is where I got the term polyglot. Then I go ask questions about that. You know you're basically just scaling up faster than you could before. It's amazing. So that's on the example within that project and again like I'm trying to fully soak my brain in all the stuff around the topic. So even things like watching a video on Dino 2 with Ryan the creator. He made Node. Now I know all the history of Node. I know a bunch of the stuff about Dino, what they're doing, why they're doing it, how to use it. And this is something you can just listen to while you're doing a different task. Same thing with watching this. I didn't know that he's coming after Oracle for the JavaScript trademark. Then on another case with my Tanstack start journey, I find a creator named Dev Leonardo and he's making really good content and it makes it a lot easier for me to understand what the differences are between Next.js and Tanstack Start. And to be just super honest, like I'm loving it so far. It was confusing at first, but just seeing the way that the patterns are laid out, it's really awesome. I really like it. And I'm paying down an idiot tax at the moment or an ignorance task t ignorance tax. But the faster that you can pay that down, the faster you get to the other side of it and then you can start enjoying using it. So, I'm just looking at this and I'm like, "Wow, this is nice syntax and I'm vibing with it. I understand it." and I'll just keep following that and then maybe I'll watch a talk on why he even made it to begin with. So then I can make my own judgment on the stack and what will be a good decision or a bad decision in terms of longevity of the stack. I look at his tools. I see that they're very popular. I see the rhyme and the reason for why he's doing it now. I have more context. I watch a few of these starter videos just to prime. So this is where I started and yeah. So, one other example, a really good example of this of learning in public is actually my friend John. He's a member, one of the early members of Vive with AI. Shout out John. But I don't know anything about leak code and he's trying to get better at talking on camera. So, what he's doing is he's following that curiosity path by recording himself doing leak code every day. And he's already amassed 46 people. Now, that doesn't seem like a lot of people would look at that if they don't do YouTube and they're like, "6 isn't a lot." Imagine your room full of 46 people and you're talking to them. It's super cool. I can't say enough like how people take YouTube stuff out of perspective, but he's following his path by going this way. And then I'll occasionally tune in because I'm just like, "What is this stuff? I'm such a noob." And then lastly, just GitHub is such a great place to to learn all this stuff. So what I'll be doing is I'll be going and and doing a bunch of starter projects and manipulating them and changing them and of course asking lots of questions. So I won't do anything agentic. I won't use taskmaster. I won't use any of those things. I just want to actually pay and feel the not pain but like actually feel the friction of where my gaps are. So I highly recommend you do that. I encourage you to do that especially if you're more junior in any topic because you can be junior in backend but great in front end and that's where I feel I'm at. So I'm just honest about it and I find it. If you think this is helpful, make sure you like the video. If you want any of the stuff that I talked about in here, I think a not sure if I mentioned any prompts, but if you want any prompts, I can give you prompts for product management. If you want to check out the community, it's half off now until we launch the new platform and it's going to be very exciting. You should like the video because more people will see it. Hopefully, it'll help more people. And subscribe to the channel so that you see me again. Have a great Tuesday. See you.