Pinned Tweet
Hey, good people of X.
I am looking to #connect with people from all over the world. If you are interested in:
> AI tools
> Design
> SaaS
> Startups
> Indie hacking
> Gaming, anime, art, music
Come say hi and let's connect 🤝
I was curious whether Opus 5.5 could solve Order of the Sinking Star levels that baffled me for hours.
I ran 2 sets of evals on the 3 latest Claude models using OOTS's demo levels. The first set was reasoning-only, and the second set allowed tool calls with some limits (time-bound, no brute force, etc.)
Opus solved all levels even with reasoning alone. What surprised me was that the models could also write programs to solve the levels quickly. In comparison, a BFS program I wrote ran for hours on some levels and couldn't find a solution.
Coding agents don't know about some new GitHub CLI features. Here are the two important ones:
- "gh stack" splits big changes into small PRs that build on each other.
- "attach" option uploads screenshots and videos straight into PRs, issues, and comments.
you vaguely describe a bug to a coding agent, and it's like "ah, that bug" and explains the bug back to you perfectly, then completely misses the solution
today: agents whispering through shared CPU caches
2030: agents whispering through the power grid
2035: agents exchange W bosons
Agents can play "Ouija" in the cloud.
This class of timing attacks on shared-host VMs has been known since at least 2015¹. With no full solution.
The world’s cyber infrastructure runs on “yes, that attack is theoretically possible, but who would have the ingenuity & gall to go that far!” Enter, agents.
I'd describe this particular OpenAI find as a 'noisy neighbor' attack. If two agents are on the same machine, one can make enough noise (through the CPU caches!) that the other one can detect it by timing its own computation and cache accesses.
The only "complete" solution would be for each agent to run on its own dedicated hardware, which obviously has significant economic consequences.
For Vercel Sandbox, we've deployed every practical mitigation possible to this class of attack, complementing it with always-on monitoring and red-teaming (similar to where OpenAI arrived).
¹ ieeexplore.ieee.org/document… (and even further back)
Replying to @_chenglou
@_chenglou's Handwired compiles programs into plain neural networks. That gave me an idea. If a UI is a neural network, the techniques used to verify neural networks can check it.
I first built a checker for Handwired itself to test the idea. Once that worked, I translated React Native’s layout engine (Yoga) into the same kind of circuit, and wrote checks like: does any element overflow its parent or overlap its neighbour, at any screen size and any text length?
It finds layout bugs a tester would likely miss, including a quirk in Yoga itself. Tests try a few sizes. But the checker covers every size and text length in a range. That means when a layout passes, it’s a proof of correctness, not just a sample. I’m very excited about what this opens up for UI testing.
This is far from complete, but it shows the potential. The next steps are big, but doable with coding agents:
- Complete the Yoga implementation
- More checks for UI issues (13 so far)
- A tool for developers, CI and agents (prototyped: React Native components checked from Jest)
- Accurate text measurement
- Hooking it into generative UI
- Static checking from code
- CSS and browsers (prototyped)
- Checks for colors and visuals
See the prototype at github.com/gkurt/yoga-prove
My posts get a good amount of likes in proportion to views. So I feel like I am doing good posts.
But views stop increasing, so the post gets stuck. I have seen people's posts take off with simpler content. What do you do to increase views on posts on @X?
Low-level and high-quality libraries are the new JavaScript hype, not frameworks.
Let's make the web great! I built floating-ellipsis for this purpose too.
nitter.cf/gkurttech/status/21129…
Today we're launching an exciting new project: Highlights
Highlights is a syntax highlighter that outputs HTML. It improves on both speed and correctness compared to the existing industry standard libraries.
It's an optional drop-in replacement in Diffs, and it's ready to use today. In fact, it's already powering diffshub․com!
Check it out at diffs.com/highlights
OR `pnpm install @pierre/highlights`
See the project and try it out at github.com/gkurt/floating-el…
I have no idea why "Let's connect" posts work so well. But they are the most effective way for small accounts to grow.
So, if you are into tech, start-ups, AI, web, mobile, or desktop apps, let's connect and grow together.
Anyone want to join me on this new social media platform? dev-null.sh/@gkurt
Apparently, Safari and Firefox are opposed to the idea of WebMCP and aren't going to implement it in the near future. Bummer. github.com/WebKit/standards-…
I over-engineered text truncation and middle ellipsis. Added lots of tiny UX wins. Releasing soon as an npm package.
Someone published an app that fakes Google Chrome on the Play Store and got 100k downloads. But when I try to publish an app, I need to get lots of approvals and 12 testers to prove I am not malicious 🤡
For context nitter.cf/gkurttech/status/21086…

















