kanchan

[kanchan @Stepsome]

@antirez @mitsuhiko Yet another: How would someone collaborate on ur work? What would the "shared understanding" of your work look like? Please do explain (may be in a longer format, and may be with videos 🤞)

[kanchan @Stepsome]

@antirez @mitsuhiko Another way to put it: How should I "read" ur repos/work? What are the specific things I should be looking for? What prompts should I be giving the llm to fetch these?

[kanchan @Stepsome]

@badlogicgames @mitsuhiko Any way you would do a writeup. I am sure I get to see the finished product , but the why and alternatives around each would be valuable learning guide. Even if you do a vibe coded blog :)

[kanchan @Stepsome]

@_xjdr @Kimi_Moonshot Wanted to know how you use you ncode with 1m context models?At what point do you use compaction and do you actually reach the compaction point for 1m (~970k tokens)? How frequently? Should I simply use defaults and set the model type to k3[1m]?

[Dejan Marjanović @djnmrjnvc]

@mitsuhiko There will be lots of providers, hopefully through GitHub Copilot too.

[kanchan @Stepsome]

@antirez How do you actually you use these model for such optimisation tasks? do you simply point them towards the repo and they oneshot it? Or do you work like @badlogicgames. Asking because I have a simple machine learning pipeline I would like to optimise for FPS.

[kanchan @Stepsome]

@btibor91 We needed the model not the livestream :(

[kanchan @Stepsome]

@OpenAI i dont see codex 5.6 yet, In india. Its already past 10am PT (10..30 PM ist)

[kanchan @Stepsome]

Does codex started like ` codex -c developer_instructions="$(cat ./instrns.md)` preserves the file's content verbatim post compaction? I don`t want these instruction sets to be summarised. Codex itself refuses to share the contents after compaction. any guarantees?

[kanchan @Stepsome]

@_xjdr Actual error: "API Error: OpenAI compat inference request failed: 403
Forbidden" not 423.

[kanchan @Stepsome]

@_xjdr If it ended yesterday, why was I meeting 423 error earlier. Sad the i couldn't use it for past 4 days.

[kanchan @Stepsome]

@_xjdr Also waiting for your posts about your workflows. You promised that in the launch post itself. And the promised inference -at-scale. Would love to dig into them.🤞

[kanchan @Stepsome]

@Hi_Mrinal The blog page for this is blank. Did I miss somethin? Here's the link I tried: https://t.co/v1iSSa5YNe

[kanchan @Stepsome]

@lisnkl @_xjdr And i dont understand how to update billing or pay for usage. I am now getting "API Error: OpenAI compat inference request failed: 403 Forbidden" and figured i need to pay for usage now. But where and how?

[kanchan @Stepsome]

@_xjdr hitting rate limit probably. I am gettig "API Error: OpenAI compat unary response aborted before assistant content" on glm 5.2. :(

[kanchan @Stepsome]

@antirez can I request you to write about how you are using codex with gpt5.5. how a programmer of your calibre uses llms, what he thinks. A kind request.

[kanchan @Stepsome]

@dhh @OpenAI @dhh @mario Is there any way I could get a look the the agent.md you are using with pi. I am thinking of switching to pi, but what's the system instructions to use.

[kanchan @Stepsome]

@GenAI_is_real A slight odd request . Please share ur methods and process and system of note taking . How often u do it.? How do you pair/plan note taking with actual readings and learning sessions, and working. Other things? Regards.

[kanchan @Stepsome]

@mattshumer_ Would love to try