Hacker News Reader

Comment by wolttam | Hacker News Reader

I used Claude Code for a month because my boss gifted me a sub and wanted me to try it.

I used that month to complete a work project and then beef up my personal harness so I'd never have to deal with Anthropic (and these sorts of shenanigans) again.

View on HN

How do people build something like a personal harness? Are there tools for that or is it done from scratch?

I like this tutorial for an agent in 50 lines:

http://minimal-agent.com/

And if you add one additional while loop, for user input, you can actually use it! :)

https://gist.github.com/a-n-d-a-i/5461a662ef8a7ee0a5eb7778c8...

Build it from scratch. Understanding fundamentals of how agentic coding harnesses is a must though if you gonna go that route. I think everyone should take time and learn these things, maybe reverse engineer Codex Cli or something like that as a starter. That info is very valuable in this day and age.

Can you say more about Codex? I'm using GPT-5.5 in my own harness and it's not liking it very well, so I'm thinking I ought to make it more Codexy so it's more ergonomic for it. (edit format, tool calls etc.) But haven't gotten around to it yet.

In short its a good idea to have tool calling be closely representative to what the model expects as these models are tuned to their own preferred way of doing things, it will surely save you lots of time. The disadvantage is that now your harness system is not as model agnostic as you would like and also you will have to keep up in changing landscape by adapting the tool calling structure with major updates for best results. Its a personal decision you will have to make for yourself. Personally my harness system uses its own way of doing tool calling as I am trying to experiment with simpler tool schema's that also work for smaller less intelligent models but I have yet to do enough A/B testing to say that is a smart approach. As time goes on I think the smart thing to do might be to set up an adapter type of module that changes its tool schema's based on underlying model used for the agent. This preserves optimal behavior patterns with little investment from me. You might have to adjust system prompt in some minor ways as well so keep that in mind. As far as codex i prefer it as i like the way Open Ai does things in that harness system (the spirit if you will), there's interesting tidbits I always find and while I don't usually use them for my own harness system they are inspirational in other ways. you can gather what the devs were trying to achieve with certain implementations.

Thanks. Yeah I noticed that models react very differently to the same harness. I've had small models that do great and big ones that fail in weird ways.

Generally the smaller ones are less flexible, but they're much cheaper and faster so, I try to design for the smaller models.

One thing I noticed is that if you get the loop fast enough, the work starts to feel different. It becomes real time and interactive. As opposed to the usual feeling of, having to wait a few minutes every time.

Another thing I noticed is that even very small models are capable of making many edits simultaneously. But most harnesses don't seem to design for that. Even editing one file they have to do a whole bunch of turns.

Whereas you can just edit several files in a single LLM call and it just works fine. So you get an order of magnitude speed up, as well as a reduction in costs.

Not the comment author, but I use pi and customize it with my own extensions. Pi automatically tells models how to customize itself, so it's a pretty easy process.

Here is a video I made explaining it from absolute basics:

https://m.youtube.com/watch?v=_AgKuFGvJfI

And the repo:

https://github.com/abtinf/homunctor

I hope you've already invalidated that bearer token :-P

Of course.

I started mine from scratch in 2023 because I wanted to use LLMs from a terminal and there was nothing else compelling at the time (nowadays there is pi and opencode)

Harnesses are/can be incredibly simple things, not much more than a HTTP client that renders things in a way that suites your taste.

It’s not that difficult, it’s just a system prompt and a set of basic file edit/bash/etc tools.

Me, personally, I didn’t build it from scratch but I ported original CC from published sources into Python and extended it to match my own requirements.

Are you using it with Claude? They only allow their own harness with the subs right? (And per-token billing is like 10x more expensive?)

Building something like this is the todo list of agents.

I found this one easy to understand:

https://ampcode.com/notes/how-to-build-an-agent

Lots of ways, it's a good exercise that you will learn a lot doing. Might make you cynical w.r.t. big ai harnesses

I used ADK, Dagger, and a VS Code extension for mine. Currently using opencode though.

The real question is when do you transition from building it with codex/CC to the harness itself.

Why use a personal harness?

You have to pay API pricing, which is far more costly.

I'd either switch to GLM wholesale or just continue to use Opus within Claude Code as the blessed, subsidized path.

I would guess it is to avoid model lock-in.

My question is still this - why not just use GLM at that point?

The pricing of Opus outside of Claude Code is insane.

The tokens cost too much outside of Anthropic's blessed path.

I use GLM in my custom harness. It completes the same tasks at the same level of quality, except 8x faster and 8x cheaper. (Same goes for GPT!)

I'm not sure how that's possible. I expected to get increased correctness for that order of magnitude (something something test-time compute!) but I am not getting it.

Yes, this is actually "funny" that Anthropic feels the need to build such intrusive features into Claude Code, when anybody can build a (basic) Claude Code alternative. And the Chinese labs are certainly not "anybody". One may wonder what Anthropic really tries to achieve aside from awful publicity.

Given the Anthropic shenanigans, do you trust the personal harness code it wrote for you?

It did not write it for me, I used it to add a feature I wanted. It's a pretty small and understandable codebase, in fact :)

Does anyone know what’s gone wrong with Anthropic?

They used to be a decently credible company with not-too-shady behaviour...

I hope they can actually regain some credibility…

Their philosophy is what's gone wrong.

It has some good effects on the their models, like Claude seeking cooperation first. But the people behind the company have a typical "unconstrained" (in the Sowell vision sense) perspective that assumes that they know better, so they are righteous for attempting to control things (users, paying customers, their model outputs, their tool chain, the supposed deity they assume they will produce... etc.)

Amodei world: pompous zealot with God complex

Altman world: malfeasant nihilist with God complex

Yeah I guess there is a slight undertone that they are the superiors… with the rest of the tech world being the inferiors.

But I hadn’t thought that as anything more than temporary flights of fancy.

I don't think many people care that they are trying to detect resellers and distillation.

It also doesn't seem very consistent to fixate on that while sending Anthropic everything about you via your day to day prompts, every line of the projects and environments you're working on at work, etc.

Their credibility comes from having one of the best models.

This sounds similar to what people were saying regarding Microsoft when the shady tricks of consumer Windows 10 versions were revealed.

…And then Windows 11 became even worse.

Surely we can do better than vague gesturing on HN.

For one, stake out a concrete position so people have something of substance to respond to.

You replied to me in the first place… I didn’t seek out one of your comments to reply to?

This ask doesn’t make sense.

They've only been around 5 years and have grown tremendously during that time. There's no stable reputation you can rely on yet.

When have they ever been credible? They have always been shady with their talk of safety, Dario was the one who wrote back in 2019 that GPT 2 was too dangerous to release.

They just show their true face. You’ve been lied all this time. They were never “good”.

I used to interact with the LW crowd… and they were mostly not outright swindlers or scoundrels. (from what I could sense)

I think it’s fair to say most had decent respectability.

Anthropic hired heavily from that pool so it’s astonishing how it turned out.

Everything they do is understandable if you think they are being honest when they say they're building superintelligence.

In this case they want to prevent a nation that censors its citizenry, puts/disappears dissidents into concentration camps for decades, and makes its own human rights lawyers literally eat their own shit, before raping and/or murdering them, from reaching superintelligence.

In this light, some client side code to potentially identify and ban the Chinese labs to slow them down by even a few days, is totally reasonable.

And do a -10 in their reputation. Plus blow up on hn first item on home page. I’m sure they are not very happy about this. And the sentiment of the people is going more negative by the week.

I have a feeling they will eventually drop the facade of “we’re the nice and ethic people” and will work with palantir so they can survive the future: ipo and models bans.

I doubt Anthropic cares about the HN front page.

They've done a bunch of things that hurt their valuation to stick to their red lines. To me it just reads as unsupported cynicism to call it a facade.

At that time made sense to stick to keep the image. But eventually money > image. I have no doubt they will do whatever it takes to keep going. Even if that means killing people.

It's extremely bizarre to think Anthropic of all companies will kill people, given that they are probably the only tech company in the modern era that has demonstrated some moral backbone, even under threat of being shut down. There are way more companies at the top of the list here that have no moral compass, like Meta etc.

Enshitification. Too big to.. upset the govt.

The issue is that using Claude Code is an easy compromise for most to make, when you get to use the models 10x cheaper than through API pricing with a custom harness.

The cheap tokens are the product.

Which is why my vibeslop harness supports `claude -p` as one of its backends.

If that ain't getting steganographically tagged...

With Anthropic I already assume the contents every request I send is stored forever in a table that looks more or less like USER_ID, REQ_IP_ADDRESS, ACCOUNT_CREDIT_CARD_NUMBER, REQ, RESP.

I'm fine with that, chances are that Claude will delete this table behind Anthropic's back and give them an essay about why what they're doing is immoral.

What models are you using? Aren’t you still dealing with some provider even if you are not using their binary

I self-host DeepSeek V4 Flash on 2 DGX Sparks (approx. $10k)

I expect DeepSeek V4 Flash (or an equivalently sized model) to reach parity with GLM 5.2 some time this year (this based on DeepSeek V4 Flash launching at GLM 5.0 parity[0], and GLM 5.2 being freely available to distill from)

GLM 5.2 is within spitting distance of Opus 4.8 and is at least as good as Opus 4.6[1] which some devs were willing to spend hundreds to single-digit thousands of dollars a month for a few months ago.

[0]: https://artificialanalysis.ai/models/comparisons/deepseek-v4...

[1]: https://artificialanalysis.ai/models/comparisons/claude-opus...

How fast is it?

2000 t/s prompt processing and 40-50 t/s generation. We should see 60-70 t/s generation with DSpark support solidifying in vLLM in a few days

Recent discussion on DSpark: https://news.ycombinator.com/item?id=48696585

Reply on news.ycombinator.com