andrewchambers

I think you misunderstood what I said or I didn't say it clearly. I was speaking in general terms of trends across our industry, not this specific project.

I didn't say or intend to say 'they' did it. I said I bet it is being done in general. Maybe that is the core misunderstanding here.

I fully believe the BSD people didn't use it - I just think in general you wouldn't know which projects are doing it and which aren't.

In fact, if you dig into these projects, they have llm disclosures for misc debugging and other things.

Your logic is ridiculous - I wouldn't post publicly like this if I was projecting.

It is just an obvious fact that when there are hostile swarms of people like you, that it is prudent to not disclose when you used an llm to unblock yourself.

I was being very neutral about this obvious fact and you became irrationally hostile.

What makes you say I don't know how to do it?

Also, people who have been doing something for years are exactly the people who benefit the most from having pointless toil removed from the workflows.

Attitudes like yours are exactly why we won't know what level of AI assistance is used. People won't admit it because there is a lot of downside to doing so - especially in communities like this.

I see quite a lot of announcements of compatibility and porting efforts -you would never know if someone used an LLM to help, but I would bet money it is being done without admitting it.

Really useful post, thank you.

I don't think people want to buy it - we just think its cool that its possible.

You should try having fun and enjoying things once in a while...

These questions are answered by the OP (Same speed, image support) - additionally, GLM is open weight.

This might just be damage control or politeness more than the truth.

I was replying to the comment about it being slow to run. I wasn't commenting on understanding it.

codex itself has a remote control mode that runs continuously. I wrote a systemd service to start it boot and interact with it via my phone.

Codex remote control serve can run continuously.

Sometimes start a new chat in the phone app, sometimes just add to the main one. Both seem to work ok.

If I want the agent to wait for something I need to start a new chat in the iphone app.

They could probably vibe-optimize it if they cared.

What would happen if they give an equivalent agent swarm the proof and a target to reduce runtime .

If they aren't already, or if its possible, prove that an optimized version matches the simple version...

I've recently had great success running codex in a regular qemu VM and using codex remote control to talk to it from my phone.

Honestly works extremely well as a personal assistant.

I can see why turning it into an API makes sense, just be aware you might not need to lock yourself in if you can setup your own VMs.

if I look at the timestamps, it appears Sholto is the one mocking Noams tweet... Noam posted first.

I think the new Astra computer use demos show that the models might be able to do things like inspection of real world objects if given a camera.

Super excited to see real world feedback added into the agent loops we have gotten used to working with. Could you let the model print and test the circuit boards it is prototyping with a jig?

if that is true then why is astra on the official ARC leaderboard now ?

I think the security aspects are overblown... but being able to edit the code of any part of your system is very useful.

Would love to see someone try to automate the bootstrap chain from a working C89 compiler to Rust.

At this point I think current LLMs are able help these incredible feats of bootstrapping as they can grind out the impossibly long built times over multiple days/weeks.

I am very optimistic for deterministic builds in general.

Closed models don't even need a back door - they will just MITM you and replace your code with malware.

I doubt anyone announces when they have under billed. OpenAI has also done many low price deals and quota resets.

I was sort of thinking a final stage of the filtering pipeline - though prompt injection etc are risks.

Seems like AI is in a pretty good position to do that tbh.

Often the spec can be simpler than the original.

The easiest way to demonstrate this is to write two implementations of an algorithm. One with no optimizations, the other with optimizations.

The formal verification can then be a proof the optimizations maintain the semantics of the simpler version and you can focus your review on the simpler version.

I think llms provide a really excellent way to do studies on software engineering techniques that previously were impossible.

It also refers to what the binary on my PATH is called, also what the library name I need to pass to link against it.

They even had an sqlite4:

https://sqlite.org/src4/doc/trunk/www/design.wiki

It is sqlite3. Emphasis on the 3 - it already has 'editions'.

The whole fable fiasco really soured me on Anthropic. This just looks disappointing by comparison.

This seems like it will have pretty huge negative affects on startups needing to compete with 'trusted partners'