Ordered a 3D printer, designed a case in Blender, received the printer, and shipped the model without having done any of this before. And it somehow worked on the first try?

Right now, a refactoring task can take more than 10 minutes because an AI agent has no recourse but to run findAndReplace on possibly 100+ files. Some agents use sed but it doesn't split imports when necessary so that can just result in 100+ new import problems.

Native LSP tools in coding agents would make them so much faster: • Get types of expressions on a specific line (instead running a build step) • Rename symbol (instead of tens of tool calls or fragile sed replaces) • Semantic import/export replace (eg when moving files)

Second edition of @AugmentedTools. 23 makers, demo’ing work and exchanging ideas. It felt electric ✨

⚡ Any amazing front end react devs that follow me available for a freelance project? Drop your site / portfolio / github in the thread

just saw the future of vibe coding with @CompuIves look out next year

wavy

Coffee & code break at #NeurIPS2025

At #NeurIPS — DM if you’re also here!

We will be at #NeurIPS2025 from Tuesday–Saturday this week. Reach out in DMs if you want to meet!

🌁🌉

https://t.co/ANmWkAYo9i

The team is assembled

Hello SF!

I’ll be in SF for the next two weeks! Anyone want to meet up (or go to a meetup)?

Why not have a tree growing in a coffee place

📸

First one was a blast! Demos included: 🧑🏼‍🎨AI building & testing @FlorisWeers 🌚Darker mode emails @MatteoGauthier_ 🔮Visualizing AI logic @CompuIves 🍬Icon studio @robertvklinken ⚙️MCP optimization w/ GEPA @_laurynas 👥AI collab iOS apps @sirian_m 🥽VR + Rust scripting @rikarends

For context, I sometimes ask Codex to write plans for other Codex instances in addition to doing the work, so I can parallelize more, so I need to find its instructions among all the noise.

And a lovely bonus would be shortcuts that work better on a MacBook keyboard (instead of Home/End), as well as shortcuts for "previous/next message" to quickly find the start of the AI output etc.

Codex quickly became my favorite CLI AI, but navigating its output can be a drag. I hope that the Transcript view can be enhanced to have verbosity levels (1: show everything; 2: collapse tool call output; 3: collapse everything except user/codex messages) cc @thsottiaux