01. building
Last week, I shipped a first look at Replit’s new Design Mode, using Gemini 3.
Gemini is an excellent model, especially for frontend web dev. The designs have been outstanding, though still require a human touch. Here are a few other examples of demo projects I’ve been working on [ 1 ] [ 2 ]
I also had the pleasure of speaking at Swyx & Latent.Space’s AI Engineer event in SF. You can find the video here and the slides here

02. learning
The Deep Tech Founder’s Guide to Twitter is an excellent resource for B2B and B2C founders looking to up their social media game
The Visual Display of Quantitative Information is a classic text on how to communicate charts & visuals. Maybe one day big labs will stop committing chart crimes
Finer points of strength & hypertrophy: as one of my latest projects, I’m turning the gross spreadsheet I’ve been using for my strength training into an app. As such, I need context for different progressions and periodization schemes. I’m relying heavily on my favorite resources Stronger by Science and RP Strength
03. thinking
Communicating is about doing the work. There are many people that might say yapping is easy or that anyone can yap.
But the point of yapping is not to yap for a yap's sake, but rather to communicate complicated ideas simply.
That's the goal of good writing. That's the goal of good speaking. That's the goal of good media.
And so education & marketing—or the process of communicating—is the job of doing the work to make your ideas stick.
Lazy people do not do the work.
There are very smart people who are very lazy, and that's probably the reason we can't understand a goddamn thing they're saying.
This is true of engineers. This is true of geniuses. And this is even true of some excellent leaders.
Doing the work means sitting down and refining your ideas. It means putting a pen to paper and ruthlessly writing and rewriting your thoughts until they make sense. Until anybody can read them. Until my mom could understand exactly what I'm trying to say about vibe coding.
This might sound easy, but it's not—especially if we're talking about things like multi-agents.
Now, the secret is that everyone wants you to think it’s all more complicated than it is, especially people on the inside. Most people on the inside feel there's some sort of job security or ego that comes with doing complicated things.
The truth is, it’s not that complicated… And anybody can start.
Getting started, like communicating, is about doing the work.
It means getting started and figuring out what you should even be doing in the first place. And that's hard. That's a blank page problem.
Many people ask me, "How do I get started?"
I can't answer that for you. But it's never been easier to solve blank page problems because we have amazing statistical machines that can write a first draft, that can help us structure our thoughts, that can help us do the work.
But they won't do the work for us.
There's nothing lazier than expecting someone or something else to do the work. And that's why there's nothing lazier than sending your co-workers AI-generated output, or even worse, sending your users AI-generated output that hasn't been properly mediated.
The technology may change how the work is done, but there is not, and there never will be, a substitute for doing the work.
SO if you'd like job security or if you'd like to be the best at what you do or you'd simply like to be a little better than you were yesterday, the only way to do so is to do the work.
