JustFinishedBSG

Well if you own the harness ( and you should ) you see everything you want.

But I’ve seen approaches using eBPF to collect telemetry to OTEL

Only state I keep is filesystem and last message (but even that is persisted in the filesystem). Each agent gets its own btrfs volume, when it’s done the “next” agent get the previous agent work mounted in its own filesystem ( and told about it ). Agent is also able to “promote” files if it wants and they are mounted in a more prominent place.

I don’t know yet if it’s a good solution. But only thinking in terms of files / filesystem sure make things easier.

Also makes branching “easier”: no handling of merging or conflicts, the receiving agent just gets N file systems and decides how to handle things.

I'll study it as I am toying with "what should a workflow definition language look like".

My current vision, and prototype, is that it should be as close as possible to a "real" language as possible so that both the user and the agent know immediately how to use it and how it functions.

So for `pi` it means using typescript.

Then the UI is derived from the AST / code as much as possible and for things that aren't neatly possible like that I eventually add small semantic helpers that define the UI.

For example "plan -> execute" is:

      await flow.unroll(
      remaining.map(point => ({
          key: point.id,
          label: point.objective,
      })),
      async () => {
          for (const point of remaining) {
          await flow.item(point.id, async () =>
              await flow.agent(executePoint, {
              title: `Point ${point.id}`,
              prompt: point.objective,
              }));
          }
      },
      { title: `Plan r${planRevision}` },
      );

( simplified code ) in my implementation and `unroll` is only there to have a nice

      ● Plan r1 · 1/3 · active
       1. Inspect parser behavior
      ● 2. Add empty-input coverage
      ○ 3. Run focused checks

UI instead of a plain "Plan · 1/3" UI with no detail (which would happen if I just used a for loop, yes it works)

They still run the same thirdparty firmware. Seeed Studio has the reTerminal Sticky too.

M5 device are cool but they do have that, as ugly as it is useful, functional diagram printed on the back...

That's a whole lot of text to say "it's markdown".

> almost as if it was trained to spit out as complex of sentences as it can.

It probably was, at least inadvertently. Very easy to imagine that one of the post-training step (RLHF, DPO etc) reinforced the "sounds clever" behaviour.

This doesn't (and cannot) solve what the post is talking about

At a very high level mathematics is basically 100% text/symbolic rewriting. You start from some set of postulate assumed true and you do your thing to get a new different set of equivalent assertions in a form that is more useful.

I don’t know if LLMs will kill the working-mathematicians but at least seem like that it doesn’t seem absurd to imagine LLMs will be good at math…

My experience may not be entirely representative because to be entirely honest I’m not exactly a great researcher and there are brilliant PhD students. That said it indeed was my experience that in the pre-PhD / early PhD period ( or even longer … ) your advisor proposes (gives) you pretty low hanging stuff that he mostly already knows how to solve, at least at a high level, with the expectation that it will teach you to use the mathematical tools you need.

It's very hard those days to think of companies/people more arrogant than Anthropic/Dario, which is quite the achievement as the bar is very high.

If that arrogance was well placed at least you could somewhat excuse it, but the fact that it is so overtly hypocritical and based on false premises just makes it so much worse.

Interesting, it doesn't seem intuitive at all to me.

My (wrong?) understanding was that there was a positive correlation between how "good" a tokenizer is in terms of compression and the downstream model performance. Guess not.

Most of the cost (to the government) for Windows is "support" (in a very general sense) and that cost isn't disappearing with Linux.

Especially since it is easier to find badly underpaid (and not particularly competent) Windows sysadmins than it is to find badly underpaid Linux admins.

> for the same reason math problems require one to show their working.

We don't put our transitional proofs in papers, only the final best one we have. So that analogy doesn't work.

For every proof in a paper there is probably 100 non-working / ugly sketches or just snippets of proofs that exist somewhere in a notebook or erased on a blackboard.

> While it's noisy and complicated for humans to read through, this session info is primarily for future AI to read and use as additional input for their tasks.

Context rot is very much a thing. May still be for future agents. Dumping tens/hundreds of thousand of trash tokens into context very much worsen the performance of the agent

I understand the idea but the way I work, a commit isn't "a" session, it's potentially tens of sessions with branching in each session.

I honestly don't know if I'm doing something very wrong or if I have a very different working style than many people, but for me "just give the prompt/session" isn't a possibility because there isn't one.

I'm probably incredibly inefficient, because even when I don't use AI it is the same, a single commit is usually many different working states / ideas / branches of things I tried and explored that have been amended / squashed.