This is true but I don't think the downvotes are "fake" though. There's just a whole lot of people who truly believe they are Making the World a Better Place Through Capitalism

It’s far, far too late to contain this sentiment.

Maybe this is just a symptom of my screen addiction, but I keep a close eye on this site for a lot of the day. I’ve noticed a pattern where my commments initially get one or two upvotes (within the first 5-10 minutes of posting) but will then immediately get a greater than or equal amount of downvotes very quickly. It happens consistently enough that I’ve noticed a pattern. The upvotes happen sporadically and the downvotes happen simultaneously.

At least until they actually tried selling them

They could be valuable in the sense that the owner can destroy bitcoin at will. Just having that leverage could be useful.

On the face of it, "10k lines of code per hour" sounds like a ridiculous metric to the point of parody.

Saying how many lines of code you can write this way is also a bit like bragging that you are building world's heaviest airplane.

The sensor isn't in the main compartment, it's in the back reading the water (not a specific bowl or other item)

Why not just wrap the tool so that when the LLM uses it, the wrapper enforces the OTP? The LLM doesn't even need to know that the tool is protected. What is the benefit of having the LLM enter the OTP?

Yes could do that, I think it makes things more complex though because then the tool is less plug and play and the thing calling it would need to handle it

Is that cheating, or is that just working smarter not harder?

The interesting aspect of the ongoing tests I feel is seeing how models can plan out an image directly using SVG primatives solely through reasoning (code-to-code). If they have a reference then it's a different type of challenge (optimizing for a trace).

Other comments suggest that the Agents.md is read into the system prompt and never leaves the context. But it's better to avoid excessive context regardless

Getting the context full to the point of compaction probably means you're already dealing with a severely degraded model, the more effective approach is to work in chunks that don't come close to filling the context window

The problem is that I'm not always using it interactively. I'll give it something that I think is going to be a simple task and it turns out to be complex. It overruns the context, compacts, and the starts doing dumb things.

If you're a developer, try NixOS. The code based configuration can be daunting but LLMs are very good at writing it.

Not sure it's good as a starter distro, but other than that I agree. I was put off NixOS for a long time despite loving the principles behind it. Then a few weeks ago I had ChatGPT give me a short course on it, including flakes and the basics of the Nix language. I completed that in a few hours and achieved more than I ever had reading the Nix docs and blogs etc. Now I'm able to use an LLM to help me write flakes while also understanding what it is doing (I'm not a fan of blindly using AI generated code).

That's what I'm getting at - the nixos learning curve is flattened out completely with LLMs to the point that I do recommend it as a starter distro for anyone technically competent (as it's still crucial to actually read and understand what the LLM produces)