helsinki

I've been building github.com/fugue-labs/gollem for five months.

Some of us pay by the token.

The best agent framework: https://github.com/fugue-labs/gollem

That's why I added an invariant tool to my Go agent framework, fugue-labs/gollem:

https://github.com/fugue-labs/gollem/blob/main/ext/codetool/...

I've been working on gollem — it's a Go agent framework with type-safe agents, structured output, multi-provider support (Anthropic, OpenAI, Gemini, Vertex AI), MCP integration, and multi-agent team swarms. The big idea is compile-time guarantees instead of runtime validation, zero core dependencies, and single-binary deploys. Think of it as what you'd want if you were building production agent systems in Go instead of Python.

To stress-test it I pointed it at a dumb task: generate an entire novel. It wrote ~40K words across 11 chapters in about 10 minutes. The "author-bot" uses gollem's agent primitives — chained agents for chapter planning, continuity checking, iterative drafting. Basically an orchestration.ChainRun pipeline where each stage feeds into the next.

The novel ("The Year Unedited" — dystopian memory-editing premise) came out surprisingly coherent. Characters kept distinct voices across chapters, the plot threads mostly held together, and it didn't just wrap everything up neatly at the end. Not saying it's good literature, but it's a decent stress test for long-running agent orchestration.

I threw it on Amazon mostly as an experiment: https://a.co/d/037EOH88

The framework itself is the actual point of this post though. It does cost tracking, guardrails, middleware chains, streaming with Go 1.23+ iterators, background process management, graph workflows, eval framework, the works. Happy to dig into any of the architecture details.

Jump host with restricted commands / access. Agents SSH into a jump host and execute what they are allowed to execute.

It has been in the source code for like two months. I've been using it for a while now.

Can you invite me? In the off chance, my email is in my profile, but reversed.

Yeah, makes perfect sense, but you really lose a lot.

No, I don’t. I built it inside of my employer’s walls (hedge fund), so I’d have to rebuild it from scratch in order to acquire paying customers (not a bad idea).

Social currency, I guess. I work on the AI team, so it’s just part of the job.

I built a GitHub-native coding / review agent that outperforms CodeRabbit (by a lot), and it only took me like two hundred hours. There are a few of us using it at work, meanwhile CodeRabbit is valued at $550M.

Would anyone at Google be willing to tell me how many people are working on this project? I’ve been building something functionally similar for my employer, but it’s a nights and weekends project with only one contributor (me).

A cheese knife.

How do you use two of these massive monitors? They are stacked vertically?

I wonder how it’s different from using playwright MCP? The prompts and screenshots, I guess?

“will transfirm you from being a consumer to a producer”

You mean, “will teach you to make REST calls instead of letting a script do it for you”.

We are all consumers. Unless, of course, you work at one of three companies.

I think that’s a decent approach, but doesn’t the performance of a Neovim terminal bother you? It simply does not feel as good as a native terminal pane. It’s not as bad as VSCode’s terminal pane, but it still leaves something to be desired.

It’s more common that they lead technological advancements in IDEs, not follow. Neovim in particular.

This is not a shared sentiment across the buy side. I’m guessing you work at a bank?

Trash

This works in theory and somewhat in practice but it is not as clean as people make it seem, as someone who has spent tens of thousands on Opus tokens and worktrees - it’s just not that great. It works, but it’s just, ugh, boring, super tedious, etc. at the end of it all, you’re still sitting around waiting for Claude to merge conflicts.

So far, it’s relatively bug-free well-written code that I’ve forked to work behind the walls of a hedge fund, and it works, but the reality is that it doesn’t provide anything that some terminal windows and git worktrees can’t offer. Am I missing something?

You really need to add more features, because I struggle to find a compelling reason for advanced users to use it.

I use Vertex and work at a hedge fund. I just spam Claude Code Opus all day long. There’s not much to it, other than I sit at a chair for 12-16 hours and spam poor (actually, rich) Claude. I don’t use the cc usage too - I just look at my GCP bill :(

My curmudgeonly genius Q/Kdb+ programmer of a co-worker, whom claims to be immune to the impact of LLMs, is going to be fucking pissed when he hears about Qython.

Interesting. I use Opus exclusively (like $1000/day in tokens) via Claude Code. Do you really think Sonnet is better for programming? I’m not sure I agree, though I’d love to save $900/day by taking you up on it.

Doesn’t Apple’s new model do the same thing? https://huggingface.co/apple/DiffuCoder-7B-cpGRPO

I’ve been writing an equivalent of uv for the R language at work, and it’s quite daunting / unwieldy. I was feeling bad about it, but then I reminded myself that uv has hundreds of contributors and my project only has one.

If you’re making real money trading, you’re not telling people about it.

Thanks, GPT 4.1. It told me the same thing twelve hours ago when I asked it what was at the top. “what’s north of the North Pole”?