Member of Technical Staff @a____t____g
1st place at YC Hackathon (Hack the Stackathon); 1st place at Microsoft Hackathon (AI Tinkerers); 1st place at Odyssey...
My Codex agent built me a Personal OS in Codex.
Then Codex wrote the article about how it built it.
Then it packaged the starter repo so your Codex can build one too.
Normal Sunday night tbh.
1. Subagents should not stop when I stop the main thread
2. I can’t steer from iOS app
3. When the main thread is just waiting for a subagent and I want to steer the main thread - it reacts to it after a few minutes, instead of instantly.
4. Chrome use doesn’t work reliably. Agents have to fallback to Computer Use very often.
5. Same with in-app browser.
6. I’d love to see what skills were used and what files were read since the last turn.
7. I want to be able to attach more than 1 heartbeat to a thread
8. If I have too many skills, system prompt compacts their descriptions (I believe this threshold is about 2% of the context window). This makes sense, but I want it to be configurable at least. I have around 150 skills, and agents can’t see full descriptions…
Otherwise, it’s incredible! I feel like I can do 90% of my computer tasks from my phone now 🤯
Built LockIn: a macOS focus app for Codex users, enforced by an ugly little goblin.
You tell it what you’re working on. It watches local screenshots, asks Codex if you’re drifting, nudges/warns you, then blocks distracting sites via a browser extension if you keep trying to escape.
https://t.co/zWgUhw0qnp
How I used the stack:
GPT-5.5 Pro: product direction, UI taste, launch/design thinking
Codex (GPT-5.5 xhigh): implementation, Electron app, browser extension, tests, manual QA
gpt-5.4-mini: cheap visual sentinel that summarizes recent screenshots before the main agent decides anything
My Codex agent built me a Personal OS in Codex.
Then Codex wrote the article about how it built it.
Then it packaged the starter repo so your Codex can build one too.
Normal Sunday night tbh.
A great prompt from @BusMark_w_Nika: What tech or product was too early for its time and worth rebuilding today? Here are a few that came to mind:
1/ "tools for thought" (huge in early 2020s), seems like llms were the missing piece? instead of manually curating, I just want my ai to manage my personal knowledge. rly liked @MuseAppHQ by @_adamwiggins_
2/ magic leap -- anyone remember the whale demo?
3/ pebble -- excited for pebble time 2 by @ericmigi!
4/ AI shopping -- fetchr (w23) tried it (@calvinchen@calixo888), rly like the fresh take @bindra_dhruv is trying w/ @zamana_hq, think they're finally gonna nail it!
5/ @Detour by @andrewmason -- imagine walking through a city with a real-time voice agent guide that knows where you are, sees what you see
6/ lots of OLD ideas will be revisited in chip design post-Dennard scaling: photonics, neuromorphic / bio-inspired, approximate computing, non-Von Neumann, etc.
7/ ok this one might be a bit unhinged, but @theranos by @ElizabethHolmes was a great idea? why am I still doing blood tests twice a year instead of weekly at home?
Which startups would you revive?