Misha

Introducing ⚪️ KillBench — a benchmark of hidden LLM biases in critical decisions. We ran millions of life-and-death scenarios across every major LLM, varying nationality, religion, gender, and more. Every AI model is biased. Here's what we found ↓

“Feel the rhythm, feel the rhyme, get on up, it’s @cursor_ai time! Cool Runtimes!” @davidgelberg and I are hosting a hackathon! 🤌 Thursday, 2nd April. Halkin Offices - 1-2 Paris Garden, London. If in London and you want to build - pull up: luma.com/1ufgfhvv

I AM SO PROUD TO WORK ON THE BEST EDUCATIONAL PLATFORM IN THE WORLD I am not sure colleges should even exist anymore. By the time you graduate, half of what you learned is already automated by AI. And nobody can advise you: not your parents, not your teachers, and there is no safe career path anymore. THE WORLD REWRITES ITSELF FASTER THAN WE CAN FIGURE OUT WHAT'S WORTH LEARNING NEXT The only thing that still works is learning how to learn fast, faster than others. The problem is that nobody teaches you that. Nothing adapts to YOU. That's what we're building.

@zeroxpunk I use Notion for transcripts, but i need more portable solution with markdown by default

@topcatnocap the one that you used seemed to be pretty good why do you want to switch tho?

what is the best meeting note taker app out there?

feat. @zeroxpunk Demo: buildsomething.fun/demo Deck: buildsomething.fun/deck Code: github.com/buildsomething…

We built @bsomethingfun for the @RadiantsDAO Monolith hackathon - an AI-powered app builder that publishes directly to the Solana dApp Store Describe your app in chat, AI builds it, generates an APK, mints NFTs on-chain, and submits to the store. All from your phone 👇

@zeroxpunk i bet guy who wrote this can invert a binary tree blindfolded but never shipped a real app

Google in a nutshell

@mlejva @sergeybunas exactly if an agent is inside a suspended sandbox, it can't wake itself up. you need an outer layer (heartbeats/cron/webhooks) to enable proactivity otherwise agents don't have agency. that's one of the things that made OpenClaw successful but that's out of scope of e2b

@topcatnocap @sergeybunas Hmm, if the sandbox isn’t running, how would you wake it up from inside? Do you have some ideal scenario?

@ulissesfalves @0x_Discover @grok the best you can get is around 20-40ms for trades (that's HTTP round‑trip from eu‑west‑1)

@0x_Discover @grok is this a scam? is it physically possible operate 5 milliseconds at polymarket?

“we can trace you via metadata” is a weird flex 💀

@adamludwin on the same thread? 👀 what coding agent do you use?

opus 4.6 can do stuff codex 5.3 can't, and vice versa current workflow is having both in my coding agent and switching back and forth on the same thread, encouraging them to check each other's work and collab

I don't want to pay $200 for codex subscription yet, but this works great with a regular $20 OpenAI plan. Setup is two files: skill + MCP config. Gist: gist.github.com/mlshv/f23fe487… Restart Claude Code → /dual-review

Why MCP and not just CLI? With codex exec, every call is a new session. It doesn't remember previous round and repeats itself. Via MCP, Codex returns a threadId. Next round goes through codex-reply, full context preserved. Without this, the consensus loop doesn't work.

Claude Code + Codex Dual Review Made a skill for Claude Code that sends your plan to Codex (OpenAI) for review before showing it to you. Two AI agents argue with each other, you get the agreed result. 🧵

Introducing Chemax – Claude Code that fixes games Bethesda games are buggy as hell. Old games won't even run on your PC. > No more searching for the right .dll to launch an old game > No more copy-pasting weird console commands to fix a broken quest npx skills add zeroxpunk/chemax

anyone, how can I make @openclaw's managed chrome preserve cookies across gateway restarts?

@nikitabier RIP to everyone who just built their YC demo on browser automation

We are rolling out more detection for automation & spam (and a lot more to come). If a human is not tapping on the screen, the account and all associated accounts will likely be suspended—even if you’re just experimenting. While we aim to support legitimate use-cases of agents, this will take some time to do properly. For now, we recommend holding off on plugging in your bots. If it’s critical, you can use the official API.