adastra22

4 ms·

1.
by adastra22 6d ago
Cursor and aider were definitely compelling two years ago. Nothing compared to now, of course, but still better than raw dogging it.
2.
by adastra22 6d ago
Have another model (or even another instance of the same model) review the output of the first. Models will hallucinate. They are also quite good at spotting hallucinations in other models' output (with some more hallucinations thrown
3.
by adastra22 6d ago
A few things, but generally more chain of thought before generating a response. So the model is tuned to think more. Given that what it outputs for this task is a summary of its thinking, tuning it to think more will just make a more verbos
4.
by adastra22 6d ago
I don't see how that would improve the situation? If I do it with the same agent/same context, it tends to just report what it did. Which is the diff, repeated in prose.
5.
by adastra22 6d ago
The marginal electricity cost is zero.
6.
by adastra22 6d ago
That’s a reasonable position to hold. The way it was stated earlier would have included the content of the paper (the research) not just the words in the paper.
7.
by adastra22 6d ago
That is what the AVP is. They’re pretty open about this at launch and since.
8.
by adastra22 6d ago
Yeah the Vision Pro is actually really nice, if expensive.
9.
by adastra22 6d ago
So you can’t interact with an LLM, or use any product that has LLM features when performing the research?
10.
by adastra22 6d ago
Nobody outside TypeSafe even knows what RLCD is, so that’s impossible to answer.
11.
by adastra22 6d ago
Not quite. You use either “a” or “the” when you are specifying a class of things. “Our town has a castle” could be any castle. “Our town has the castle” would be a bit more rare and sounds weird as a general statement, but is valid in certa
12.
by adastra22 7d ago
Radicule does something similar. Also git-bug.
13.
by adastra22 7d ago
Or just stand on the other side of the wall?
14.
by adastra22 7d ago
The idea the proofs are the heart and soul of mathematics is an unfortunate unforced error, and will lead to the death of the professions now that machines are better at making proofs.
15.
by adastra22 7d ago
Not the person you’re replying to, but I think that was a bad example. Because an LLM‘s output is iterative, the output ends up being influenced by various attractors. That doesn’t happen when you one-shot a single prediction (or multiple p
16.
by adastra22 7d ago
BERT run on what? You would need training data, no? The things would use Jev for have no training data. Not that kind of problem.
17.
by adastra22 7d ago
That has been done, like back in the 1960’s.
18.
by adastra22 7d ago
If what you teach is proofs, then wheat you will filter for are students who live proofs.
19.
by adastra22 7d ago
It may have come from an old textbook, but its origin as a meme is a 2009 joke blog post that gets its humor from making fun of how obtuse and pedagogically inept that explanation is. When people reference it today, this is what they are re
20.
by adastra22 7d ago
What I meant is that the LLM isn’t the one that loads the agent file. It’s not trained to look for CLAUDE.md. It’s only training is to edit that file, and the harness does the magic of injecting it. So I would predict that with those instru
21.
by adastra22 7d ago
CLAUDE.md is handled by the harness, not the agent.
22.
by adastra22 8d ago
The point is that the terminology surrounding this is impenetrable and non intuitive. Rather than acknowledge that, we get a lesson on category theory, which is missing the point.
23.
by adastra22 9d ago
Iterative aren’t a thing in C, where that code smell notion comes from.
24.
by adastra22 9d ago
You have become the meme.
25.
by adastra22 10d ago
No, I don't believe so. Hallucinations are not "high probability" in a real sense. They are an artifact of the random walk the inference algorithm takes, which causes it to latch on to and chase attractors in the noise. This
26.
by adastra22 10d ago
AFAICT it is the same interface as you describe, but the underlying inference algorithm is fundamentally different, hence the speed gains. There is an application I am currently working on right now where this typed output predictor is the
27.
by adastra22 10d ago
Forgive my ignorance. Tabular foundation model?
28.
by adastra22 11d ago
Hacking other companies is illegal as far as I understand.
29.
by adastra22 12d ago
OpenClaw? Ok literally there's an initial prompt during bootstrapping, but the agent rewrites its starting prompts, in a way that is nondeterministic even with same inputs. It doesn't take very long for the random walk to have it
30.
by adastra22 12d ago
This is simply not true. People have been running autonomous systems for at least a year now.

More ›