Ken

🧵 I've made a list of goals every year. Most of them never made it past January. So I'm making it public. I'm pinning this post for the entirety of 2026. Everything I build, learn, and fail at will be documented in this thread. See you in 5 months.

We choose to ignore it, but we achieved AGI long ago with this guy.

@daniel_mac8 @Attack Ok so, frontier models are coming out every week now. That's alarming.

Opus 5 is coming next week and it’s better than Fable according to @attack.

I can see this being a new job description. AI interpreter Engineer

I badly need a $2,000 Macbook to create more unfinished projects.

@Tim_Dettmers I feel like the AI community creates these new productivity hacks in the hopes of getting better outputs. But it never really works.

Do loops actually work for anyone? I tried them for months, and even with the best specs or a well-defined goal, there is always some drift that leads to lots of wasted work that needs to be unwound. I have had much more success with single-agent sessions running overnight

@keean_edward Yes but power tools can't work on their own, they need workers to use them. This analogy can't be applied to AI.

a good engineer friend of mine explained the effect of AI on software engineering with the perfect analogy. woodworkers used to do everything with hand tools. then came power tools. power tools didn’t replace woodworkers, but they made them significantly more productive. woodworkers can still choose to build everything with hand tools, but doing so has become more of a niche choice. AI is doing the same thing for engineers - you can still write code by hand, but it won't be the default for much longer. this will likely apply to most industries, and it doesn't mean that your job is 100% safe. but at least somebody will still have a job.

Daily reminder, "You are not the User".

I doubt it.

@arhamamiin @3blue1brown That's a nice view

Day 6 of learning AI bs from scratch > deep dive into back propagation > understanding gradient decent and local maxima’s of functions > @3blue1brown’s neural networks series continue (goated series for developing intuition, link below) > understanding cost functions and math around them > went through the drawbacks of these simple layer neural networks Had to lock in a bit towards the end of the night, was busy during the day recording a launch video, then watching world cup finals. Thinking of building another project here first as a stepping stone. Lets see. Gn.

Will I make it?

Read 10 Books (For my sanity)

🧵 I've made a list of goals every year. Most of them never made it past January. So I'm making it public. I'm pinning this post for the entirety of 2026. Everything I build, learn, and fail at will be documented in this thread. See you in 5 months.

@aneo0389 same! I want my google one pro subscription to be put to good use

@kenrt_ agreed, my google one pro subscription is waning out

Google should just ship Gemini 3.5 Pro. It doesn't have to be the best. Just make it fast, cheap, and really good. Then put everything else into Gemini 4.

@TheTotallyRed I agree. Making it fast and cheap would make people use it.

@kenrt_ just make it fast ngl

Based on what I've seen, the success formula is: ​success = (preparation × volume × time) × (luck + 1) ​If prep, volume, or time is zero, the whole baseline collapses. Luck is the multiplier, but enough volume over time makes you statistically impossible to ignore.

THE NEXT BIG LEAP ON THE INTERNET MIGHT NOT BE AI. For the last 30 years, the web has basically done the same thing: pull data from an API and slap it into cards, tables, or charts. The user reads it, and that’s it. But something much more interesting is quietly taking shape. Instead of just showing data, developers are now trying to recreate its actual behavior in real time. One of the most striking examples is happening with Three.js. Live ocean data — wave height, period, wind speed — gets pulled from an API and fed into shaders and physics models inside the browser. What you see isn’t a pre-rendered animation. It’s a living simulation of the sea that’s constantly reacting to fresh data.The real challenge isn’t making water move. It’s making the waves break correctly, knowing exactly where the foam should appear, calculating how light scatters across the surface — all while keeping it running smoothly at high performance. Modern front-end development is no longer just about interfaces. It’s becoming math, physics, GPU programming, and real-time simulation. This is where I think the internet’s next evolution is heading: from websites that display information to digital environments that simulate reality. Today we’re recreating the ocean. Tomorrow it could be entire cities, factories, or complex systems running live.

@thsottiaux It helped me from having 3 unfinished projects to now having 20 unfinished projects.

I was so inspired reading all the DMs on how folks here use ChatGPT Work. Let’s try something else that will make the team smile. What’s a time you saw ChatGPT have a deeply positive impact on your or someone’s life?