David Hooton

David Hooton - Nerd Wrangler, Builder of Networks, Automation, Orchestration, AI, ML, Lover of bacon.. opinions mostly my own.

[David Hooton @Dave_Hooton]

I moved all my commit and pr content generation to batch mode a while ago. Moving ci troubleshooting and bug reports to it shortly too.

You’ve just got to pick the right workloads and have the workflow setup so humans aren’t sweating while it works. Even then it’s still not that slow anyway.

[David Hooton @Dave_Hooton]

@yacineMTB Laz0r mosquito zapper?

[David Hooton @Dave_Hooton]

@nathanclark_ @thsottiaux Yeah I want to be able to send multi-choice questions vis mcp and/or the app server to codex users

[David Hooton @Dave_Hooton]

@thsottiaux Expose Codex App widgets like multi-choice questions (and more please) via the app server.

[David Hooton @Dave_Hooton]

@thsottiaux @thsottiaux It would be REALLY great if widgets like multi-choice questions could be surfaced via the app server for plugins and other automations to use.

[David Hooton @Dave_Hooton]

@LaFrogman @YatimaOfKonishi @macjshiggins I played with an idea like this before and came up with the idea of making three or more light sources which when they converge turn into enough energy to nuke him.

[David Hooton @Dave_Hooton]

@LaFrogman These would make the most outrageously great business cards ever. I’d happily pay $500

[David Hooton @Dave_Hooton]

@thsottiaux @itsrubenbtw Oh man.. you’re in it now… codex tramp stamps incoming.

[David Hooton @Dave_Hooton]

@thsottiaux 100% this is totally a thing.

[David Hooton @Dave_Hooton]

@thsottiaux @reyneill_ Changing config.toml also can lose all chat history too. This kills me.

[David Hooton @Dave_Hooton]

@yacineMTB What are you using for stt?

[David Hooton @Dave_Hooton]

Orchestration of multiple tiers of agents. Plus it still has a bunch of weird stalls when you’re running more than 5 or 6 things at once.

Also… losing chats/sessions when you move between providers/proxies is terrible. If I’ve worked in a directory with codex ever it should show me ALL my previous chats/sessions unless I’ve archived them.

[Bill Clerico @billclerico]

Claude Code is Farmville for 40 year old former software engineers

[David Hooton @Dave_Hooton]

@thsottiaux any chance we could get a “Fly My Pretties” mode for codex that makes codex hard remember that it is the orchestrator and it should be using subagents that use the right model and thinking level for each task?

It still regularly forgets that it’s meant to be delegating work and starts doing everything itself.

[David Hooton @Dave_Hooton]

I want a “Fly My Pretties” mode that makes codex hard remember that it is the orchestrator and it should be using subagents that use the right model and thinking level for each task.

It still regularly forgets that it’s meant to be delegating work and starts doing everything itself.

[David Hooton @Dave_Hooton]

@Peter_J_Beck Yeah those shock diamonds tho…

[David Hooton @Dave_Hooton]

@realchillben @Yuchenj_UW

[ani @anirudhbv_ce]

We finally know why LLMs hallucinate. It's not the model. It's the geometry.

@OpenAI text-embedding-3-large: 91/3072 dimensions do real work.

@GeminiApp gemini-embedding-001: 80/3072 dimensions do real work.

~97% of your vector database is mathematically empty. Your RAG system is retrieving from noise.

@ashwingop and I present "The Geometry of Consolidation" - a proof that RAG compression has a hard floor no algorithm can beat, set by a single spectral number your embedding model cannot escape.

Every hallucination your RAG pipeline produces? This is why.

Paper + results: https://t.co/zut8pdoPbH

[David Hooton @Dave_Hooton]

@pigeonisperfect Dubbo

[David Hooton @Dave_Hooton]

@thsottiaux @hopes_revenge Possibly the widest open door there is. Way more likely to jump on the band wagon at that age.