Anthropic shipped generative UI for Claude. I reverse-engineered how it works and rebuilt it for PI. Extracted the full design system from a conversation export. Live streaming HTML into native macOS windows via morphdom DOM diffing. Article: https://t.co/C3FLF3JB8Z Repo:

Introducing napkin 🧻 (Agents + Obsidian + a CLI tailored for agents is amazing). napkin is a knowledge base that doubles as an agent memory layer. ultra fast, completely sidestepping vector search. First class support for @badlogicgames's pi - includes a self-distillation

just need to get progressive disclosure, search and auto distillation right

confirmed, agents + obsidian is awesome. i think i can make an amazing knowledge/memory system using this combination.

agent + obsidian is a very interesting combination, i don't like their CLI though

Tried porting pi-generative-ui into a generic CLI for other coding agents. Turns out it's virtually impossible. The core trick of intercepting tool call deltas mid-stream to render partial HTML requires hooking into the streaming layer. PI exposes this natively. Claude Code

Mercury is my love letter to PI, powered by it, heavily inspired by its extension philosophy. It's built for multiplayer from the ground up. Multiple users in the same channel, RBAC decides who gets to trigger what CLI, and conversations from whatsapp, discord, slack, wherever,

I want single use skills via npx, can we get these? @andrewqu i don’t want to install random stuff i use once…

This tweet had me thinking, what’s the “minimum viable ontology” or list of terms to quickly get situated within a new domain, and improve your prompts? Vibecoded https://t.co/0ElAz3RLrm to showcase this idea on a few example domains -- been wading into 3D recently and found

https://t.co/ppxcskOsve

I have zero affiliation with any crypto coin.

You can't prompt your way to knowledge. You have to distill it. Wrote up how I built a 213-line skill through a render-look-fix feedback loop that teaches agents to make better charts. https://t.co/sEOc1AC5Dq

אני מחפש עזרה מהקהילה לאירוע חינמי שאנחנו עושים ביום ראשון https://t.co/mmIcZ6HT5s אני צריך צלם וידאו שיארגן לנו מצלמה ויציב אותה. אנחנו נערוך את כל החומרים. מי רוצה לעזור לנו?

The "Claw" ecosystem went from 0 to 20 personal AI agents in 3 months. OpenClaw alone has 217K stars. I tore apart 12 of them. Three architectural bets emerged: Embed an agent SDK (OpenClaw wraps Pi, NanoClaw wraps Claude Agent SDK) Shell out to a CLI agent (TinyClaw spawns

My jira-cli skill got flagged HIGH RISK on @vercel's https://t.co/nbUFudScGh because it... uses credentials to talk to Jira. And runs commands. And reads Jira data. That's the skill. Best part? It flagged use of https://t.co/azhCvVj9f4 (a secrets manager designed to prevent

🚨 BREAKING: Researchers just built an AI that must earn its own salary or go bankrupt. It's called ClawWork. It starts with $10, gets assigned real professional work, and pays for every single token it uses. $10K earned. 7 hours. Zero human input. → AI gets a real task

Feels like everyone making their own agent stumbles across the same primitives and thinks they solved something Let me save you some time (read this, it's funny and useful): - You're going to make an agent - You're going to run it on benchmarks > It's going to suck - You're