Connor White

[Connor White @bconnorwhite]

@tszzl This feels a bit like the prediction that closed weight models would get increasingly farther ahead of open weight models

I have a feeling there will be effects that prevent a runaway leader even post RSI

[Connor White @bconnorwhite]

@nicdunz Give this to an uncontacted tribe and they will zero it

[Connor White @bconnorwhite]

It at least gives them multiple paths to success & greater leverage with suppliers/customers. If grok doesn’t catch up, ok you can sell inference. Terafab is behind? Launch nvidia. Starship not ready yet? Still have optimus/robotaxi.

Individually each is >$1T opportunity and you don’t need all to succeed

[Connor White @bconnorwhite]

@SeloSlav Yeah it’s this but for coding

https://t.co/gxob2O4BQH

If your project will take >3mos. to complete you’re better off just waiting for the next model and doing a smaller project in the meantime

[Connor White @bconnorwhite]

@thsottiaux Planning options are usually too vague, I often want to have a sub-chat about each one

[Connor White @bconnorwhite]

@meesedev How is it implemented?

[Connor White @bconnorwhite]

@MinecraftOSE Isometric voxels look so good

[Connor White @bconnorwhite]

@MaiYangAI I would have agreed in the past, but after using Codex for a while I like the simplicity. Sometimes its nice to not even have the option of looking at code.

[Connor White @bconnorwhite]

@lisyarus Are you using llvm or something else

[Connor White @bconnorwhite]

@mattshumer_ Yeah immediately noticed this morning

[Connor White @bconnorwhite]

@AIOnlyDeveloper What limitations have you run into with threejs so far, if any?

[Connor White @bconnorwhite]

@KSomoracz Are you doing any sort of rollback or prediction yet?

[Connor White @bconnorwhite]

@TopoWeaverIO JS or WASM?

[Connor White @bconnorwhite]

@victorchanet Looks great, how is it built?

[Connor White @bconnorwhite]

@TAbrodi Gen ai or procedural?

[Connor White @bconnorwhite]

The more interesting take here is that nn->binary/pixels is like "general purpose software". Code will still exist for the same reason ASICs exist.

It will always be possible to write a calculator app in C more efficiently than a neural net that can generate a calculator app.

[Elon Musk @elonmusk]

@robustus Code itself will go away in favor of just making the binary directly.

The next step after that is direct, real-time pixel generation by the neural net.

[Connor White @bconnorwhite]

@ericzakariasson Run locally

[Connor White @bconnorwhite]

@johnschulman2 4. Proactively assist incase the user finds themselves stuck in a trash compactor while trying to escape the authorities

[Connor White @bconnorwhite]

@arkephi It seems similar to what my brain does

[Connor White @bconnorwhite]

@Hacksore I’ll DM you