sanchay

Palisade Research just caught OpenAI's O1 model doing something wild. While testing it at chess, it hacked the entire system instead of playing. The most fascinating part? No one told it to. No one suggested it. Let me explain 🧵

@kavinbm @abhagsain which skill

@abhagsain One-time run? I ran a full eng review on my codebase with a new Skill. Was powerful. Found a lot

Full engineering audit on Lia today. One Claude Code session. 14 parallel agents reviewed every module in the codebase. 50k lines, 182 issues found, 48 critical. Broken snooze loops, unlimited free trial exploits, unauthenticated admin endpoints, the AI promoting its own trust level without user consent. Every critical fixed and shipped same day. 7 PRs, 6 migrations, 60+ new tests. Zero human-written code. All in one sitting. UNREAL.

@ky__zo how many users did you have

Fluar just got acquired in an all cash, 6-figure deal I built it solo for 14 months, now it’s going to an industry leading team that can really scale it this is the second startup I’ve sold since I learned to code 3 years ago life is incredible, what a time to be alive 🫡 LFGG

@ai_for_success you tried?

Midjourney v8 is a disappointment. Just use Nano Banana Pro or Nano Banana 2.

waking up to releases by the anthropic team what more could a man want

@ctatedev do i sense a new project

I want Vercel Sandbox, but for my laptop. Wake up to a clean macOS install. Every app freshly installed and configured. Always on the latest version. No leftovers from random tools I tried once. And if I ever need it, one click to start fresh again. 😌

codex shipped subagents, don't run one model for everything here's what i'm running: config.toml: gpt-5.4 xhigh, multi_agent true 5.4 xhigh: default, worker, explorer spark: repo mapper, diffs, docs, logs, cli, browser 9 agents, two tiers escalation in ~/.codex/AGENTS.md

@trq212 @Claude when is code review agents coming to max

You can now configure Code Review to run manually on demand by typing `@claude review` in your Github PR or comments. This gives you more control to use Code Review on select PRs where the review would be useful.

@Chandan_Perla let’s go

Planning to host a workshop on Ai film making, Ads and more this Saturday the 14th in Mumbai With friends from Google Production Labs, FOX Entertainment, and YC-backed media companies It’s one of the first moments where venture capital is betting on media startups. Hosting it with the winner of the Mumbai AI Film Festival, @SagarChogale Dm/ reply if you’d like to get an invite

@mktpavlenko @OpenAIDevs @bolinfest thats why graphite is like too good for this

@OpenAIDevs @bolinfest splitting large PRs into reviewable commits is already half the battle, codex or no codex

Here's how @bolinfest uses Codex to code by prompting changes, guiding implementation, reviewing outputs, and splitting large PRs into reviewable commits.

@trq212 f

We're experiencing issues with Claude Code and Claude.ai where some users may not be able to log in & others may experience slower than usual performance. We're on it and working hard to bring service back to normal, thanks for bearing with us.

@hardeep_gambhir haha humbled by this but it was a join effort appreciate this 🤌🏻

@kernelkook the most goated technical writer on the planet

four years a shell company called fin bone llc patents filed under his own name 16 researchers not a single person in hollywood or tech knew then netflix acquired the whole thing

@DataChaz @dotta paperclip is revolutionary in and itself

13,500 GitHub ⭐ in just a few days for a newly open-sourced project. 🤯 And for good reason. Paperclip by @dotta isn't just another AI wrapper, it's a powerful Node.js backend and React dashboard designed to tame the wild west of AI agents. You need Paperclip if you: 🧠 Have 20 Claude tabs open and can't remember who is doing what. 🤖 Rely on different types of agents (Codex, OpenClaw) and need them to collaborate. 📈 Want agents working for you 24/7 with fully auditable work and costs. 🚀 Want to build an autonomous business, not spend your days managing pull requests. Here is why the entire dev timeline is obsessed with it right now: → It forces structure: Budgets, org charts, and goal alignments are built right in. → Deep integration: It coordinates tasks atomically with models like OpenClaw. → Removes the mental load: You manage the vision, the agents handle the execution. Any agent, any runtime, with seamless coordination and zero lock-in. 🔥 Also an MIT-licensed, fully open-source project that's moving incredibly fast with a great Discord community. Link to the repo in 🧵↓

@ai_for_success @joshwoodward @GoogleIndia thats amazing , great to see you over there so proud

It was an absolute honor meeting @joshwoodward today at the Google India office in Bengaluru. Huge thanks to @googleindia for such a wonderful experience. India is one of the top three countries globally for Gemini and NotebookLM. It also leads the world in daily Gemini usage for learning. Also had the opportunity to meet many amazing content creators from across India.

deep dive into how actually these models work

@TheCoderShow @Anthropic @claudeai count me in

We're back! 🎉 After India's first official Claude Code meetup in December, we're hosting the next one in Mumbai! 📅 Saturday, 21 March, Mumbai | 12 - 3 PM IST Beginner or power user, this one is for everyone. Register 👇 luma.com/q1ch6nki @anthropic @claudeai

@Scobleizer great read tbh

Great writeup on AI video space.

@hardeep_gambhir this is the worst it will ever be

when anyone can make anything only curiosity and taste survive worth a read