Ian

@badlogicgames @DanielGri @mitsuhiko @DanielGri Hey Daniel, curious what did you use to make your slides? any chance to share? thanks!

Full recording of Pi Day Talk by @DanielGri and @mitsuhiko and I not being able to stop talking during the AMA. Serge eventually saved the audience from more ramblings. Enjoy. x.com/i/broadcasts/1…

@dosco lovely！sharpening FE skills with gemini or Slopus?

new website

@adibhanna care to share how to add session persistence? :)

cmux is actually great! I figured out how to add session persistence without tmux and it’s vm, it doesn’t add any layers on top of cmux!

@dosco cool I’ll play around with it and provide feedback

@Eminemead works with snowflake but you can define relationships between columns across tables and databases even when the underlying db does not support it.

connecting llm's to sql and other databases has always been hard you're dependant on a big model that can write correct sql. not any more try graphjin a compiler that does auto-discovery of your databases and builds a graph that can easily be queried with graphql which is pretty easy for llm's to write. graphjin.com

@initjean amazing! thanks!

introducing slopmeter a cli tool to create a sharable nice looking graph to show off your Codex, Claude Code, or OpenCode usage npx slopmeter@latest

.@LakshyAAAgrawal has done an unbelievably nice job collecting successful applications of GEPA to everything from state-of-the-art research to new products and established enterprises at this page. Staying on top of these is harder than one would think! gepa-ai.github.io/gepa/guides/us…

@rcarmo @badlogicgames how does PiClaw generate the charts? look super polished!!

People of pi, github.com/rcarmo/piclaw is becoming aware of its surroundings - would there be an interest in shipping server management skills with it? #piclaw #homelab #openclaw /cc @badlogicgames

@cramforce Spotify chat？@eldsjal

Which other adapters should we ship into Chat SDK?

@nicopreme This is hilarious! Lol

Bummed that my request join the Claude Code meetup today in Vancouver was not accepted. Fomo kicking in.

@OnlyXuanwo yes plz!

Next Step

@avimakesrobots curious your use cases for mastra

These guys are absolutely cooking. We will be using this

@leeoxiang @excalidraw @Remotion 大佬用的哪家TTS啊？

打通了 Remotion + Excalidraw + TTS ，可以快速的生成中手写风格的视频动画了。 @excalidraw @Remotion

@gooby_esq DSBash is def intriguing!!! thanks for sharing this. I'll give it a go!

the idea here is you might want to get something back from a rlmgrep call other than grep style line numbers, and --signature-json let's you have it as structured data

Added signatures to rlmgrep! (at least for your instruction and output types, inputs remain fixed to being 'grep' shaped so inputs remain whatever files or dirs you pass in). github.com/halfprice06/rl… So you can now do something like "rlmgrep --signature-json 'summary: str, findings: list[dict[str,str]]' "Audit auth and summarize issues" . and you'll get a json blob to stdout with the output fields you defined as json. if you use --signature (instead of --signature-json) it renders as 'markdown' to be more friendly to read in the terminal makes me wonder, has anyone tried to make "DSBash" lol

@dosco @mchonedev @Dspy i'll see what i can do lol, incoming PR

@Eminemead @mchonedev @Dspy i've been meaning to look into rlm but not had the time, will do this weekend unless anyone else wants to take a stab earlier.

The @DSPy team did some AMAZING work bringing RLMs to the framework. After getting a chance to look at it, I have a few thoughts. I was asked: "How is RLM different than what claude code does?" and oh boy the difference is exciting! RLMs treat prompts as DATA. 🧵

@n2parko ok, found it! thanks

@n2parko Can we get max mode toggle on CLI too?

I get a lot of questions on how Cursor CLI is so fast there are many small optimizations but the reality is it sits on the shoulders of tje best indexing infra on the planet, equipped with the fastest coding model

@jessethanley @cursor_ai cursor has codex now. what type of subagents do you run in this use case? mind share?

I would use @cursor_ai background agents but 1) mobile is broken, 2) don’t have codex, 3) it’s bad at planning, 4) web is lacking so much and has for over a year, 5) no subagents, etc.

My morning routine right now is Open @opencode Dump list of ideas collected from previous day Ask it to pick top priority/easiest + high impact wins in plan mode Run it Go to gym Review when I come home

@milichab can we make nano banana model native in cursor? that'll be amazing!

All sprites in IsoCity come from Cursor's Nano Banana integration