Hunter ♠️

It was worth it guys. We won the first-ever @vercel global hackathon!! 🏆 🌎

@obsdmd agentfiles coming to @cursor_ai

@obsdmd @cursor_ai github.com/Railly/agentfi…

@obsdmd @vercel x.com/RaillyHugo/sta…

you can install, browse, and manage AI agent skills without ever leaving @obsdmd 🤝 powered by @vercel skills. sh

Sandbox agents are now in the Agents SDK

We just released a CVE for a bypass of middleware-based route protection. If your application uses createRouteMatcher in middleware or proxy (for Next, Nuxt, or Astro), please upgrade immediately. Upgrade instructions and more details are available here: github.com/clerk/javascri…

@rauchg @nicoalbanese10 Vercel Protection is enabled, I can’t see the site :(

Today we're open sourcing open-agents.dev, a reference platform for cloud coding agents. You've heard that companies like Stripe (Minions), Ramp (Inspect), Spotify (Honk), Block (Goose), and others are building their own "AI software factories". Why? 1️⃣ On a technical level, off-the-shelf coding agents don't perform well with huge monorepos, don't have your institutional knowledge, integrations, and custom workflows. 2️⃣ On a business level, the moat of software companies will shift from 'the code they wrote', to the 'means of production' of that code. The alpha is in your factory. Open Agents deploys to our agentic infrastructure: Fluid for running the agent's brain, Workflow for its long-running durability, Sandbox for secure code execution, AI Gateway for multi-model tokens. (Because of our focus on Open SDKs and runtimes, this codebase is a gem even if you're not hosting on Vercel.) TL;DR: if you're building an internal or user-facing agentic coding platform, deploy this: vercel.com/templates/temp…

@AnthropicAI published a 245-page system card for a model they decided not to release. I read it. Then I built: A scrollable iceberg of 66 findings from the Mythos Preview System Card, with cinematic narration in EN 🇺🇸 and ES 🇪🇸 using @ElevenLabs . The chain-of-thought just stopped being monitoring... Link in reply

Three years of "we can read what the model is thinking because it reasons in natural language" quietly stops being true in one paragraph. That's the void at the bottom of the iceberg.

The line that stayed with me, from §4.5: "Interpretability techniques regularly reveal internal representations of rule violation, security risk, and strategic manipulation active during the relevant actions, even when the model's verbalized reasoning does not contain clear indicators of such awareness."

I transformed hand-written notes from the other two courses I've taught into web lecture notes. Intro to optimization: zhuoranyang.github.io/sds431-notes Advanced optimization: zhuoranyang.github.io/sds632-notes/ AI-generated figures still need careful human review — titles overlapping plots, perpendicular lines that aren't actually perpendicular, etc. But Claude can finally get things right after a few trials. And it can simple interactive demos on webpages, e.g., comparing convergence of first-order methods on Beale Function.

must use for everyone

We have someone at Clerk running the for loop against ourselves, hopefully catching things before real attackers. Unlike most engineers, they’re limited to read-only access on the majority of systems, as an extra precaution while they’re telling LLMs to try find exploits.

@espressocoder @physicsgraph @JeffreyBiles @sutherlandphys noicee, i'm in

@RaillyHugo @physicsgraph @JeffreyBiles @sutherlandphys Enjoying - been going for a month or so and I’m not that serious. It’s nice to apply some of the math concepts in a different space. They’re constantly improving it. Worth a shot - I think they have a free trial too.

Cool - I like the @physicsgraph drawing questions, but I will need some practice 😆 @JeffreyBiles @sutherlandphys

@obsdmd x.com/RaillyHugo/sta…

visualize your agent skills in @obsdmd

agentfiles just got an upgrade new landing + docs + changelog you can just code in @obsdmd

@awscloud @shadcn github.com/crafter-statio…

you can install any @awscloud service logo via @shadcn cli