capevace

Once image support drops this could be a cool addition to https://struktur.sh. Will it be on OpenRouter too or will all inference have to be self-managed? I’m thinking about server-side use cases, where budgets are low, so small models shine.

edit: ah I saw mainly iOS for now. But an integration would be possible on macOS then, right?

It only tries to solve the „quality“ aspect of that equation tho, at least for the code output.

But I think for actual evals, some heuristics about a coding agent session are needed. But since the input tasks are always different, it’s hard to make any KPI actually comparable.

Did the agent needing 3 rounds of feedback on a big task perform worse than the one that needed none, but only worked on a small change?

I’ve been tinkering on a package to capture linking, testing and gating quality output in a framework unified way across coding agents.

I would love to have a chat with you about how you see this space developing and how you look at it.

https://github.com/mateffy/gesetz

I’m saying it doesn’t matter wether they did or not, because now they clearly have models which can compete.

I just don’t buy it. There’s still gonna be demand for stronger models. Sure growth will be slower, but this might even drive a push for more cost efficient training / inference and/or new architectures if money is harder to come by.

But the argument isn’t about model providers tho right (providing inference)? Using Chinese providers is already a privacy problem if user data is involved, so I don’t know if anything would change here.

What you’re suggesting is that the policy would stop big co‘s from running their own inference completely, no? Or only the foreign open source models?

> don't bother analysing it technically

I feel you have to, because where’s the line? Is running GPT-OSS myself fine (if I were a big company)? Mistral models? Or are none of them ok? What about if I am getting inference from a small startup, which is running OSS models for me? Or is only OAI/Anthropic ok?

This administration has done a lot for that minority already so I don’t think anyone would be too surprised? In this case they can even claim to protect the greater economy, a crash of which does impact the majority.

„US good, China bad“ has been a long term talking point by Trump too and this would neatly fit into that to be consumed by his base

I agree, I even think we’re already there if yesterdays OpenAI/HuggingFace story is legit.

We need to start building these agent systems in a way so that it won’t matter if a model is compromised. Malicious models is one way, but prompt injection is also still an unsolved problem.

And if you solve that, then a malicious model is also no longer a threat. Data exfiltration maybe, but IMO US labs are probably also doing that for training, even if saying otherwise (of course I have no evidence of that, but the temptation must be insane)

I think in capitalism thats called a market correction?

> Do you understand how much money is tied up in the industry?

If the free market „virtue signaling“ no longer matters, then they could’ve also put a stop-gap on investment volume to protect the economy against over-reliance?

Or how about go full plan-economy and just dictate price per token until investors get their money back?

I’m joking around but of course I understand whats tied up in this. I just don’t see how banning open source models is helping any of that in the long run.

That’s what I don’t get tho, I think in today’s world they totally could?

„We need to protect US investments as our entire economy depends on this market sector“ seems like something that would be soothing to me as an investor?

It’s not like the big money givers can’t see that china is close, it’s been the news multiple times already. But taking active policy steps to protect against that should boost market confidence, no?

Instead we get alibi arguments to accomplish the same thing in the end?

Seems a bit far fetched, given that I’ve never seen any PoC/research on this topic. Got any pointers?

To be fair the fix to this seems to be that American labs would need to adjust to the new „fair market price“ then, right? Subsidy-backed competition is still legitimate competition in capitalism (e.g. Uber).

And you could still ban specific, known malicious models/labs instead of blanket open-source bans. What about Mistral models, for example.

Distilling or not, they are clearly close enough to the frontier, that the supposed „free market“ country needs market controls. That may work for US markets. But not the rest of the world.

Idea: petrodollar policy becomes „tokendollar“, enforced by US military dominance. If you force me to pay altman at gunpoint, then maybe ill stop using kimi

1. if it’s to stop hackers doing hacking things with „uncontrollable models“ then, well… they’re already doing something illegal to begin with, why would they care about breaking another law running these models?

2. if it’s to stop foreign actors, then that ban would not apply to them anyway

3. it’s not stopping distillation either, Chinese labs are already banned from using US frontier models and look at how good that is working

I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore, and should by itself also limit the viability of the idea that all those VC billions will ever make a return? In any case this would be something benefitting only a very few for a short time (labs + investors).

Someone please enlighten me what the actual argument here is, cause I can’t see it.

It’s much more limited in scope but fully open source and highly customisable. In fact it’s made for people to build their own pipelines on top of, providing the scaffolding needed to do so in a reliable way.

During development I’ve found it to be hard to truly generalise agent/llm-based data extraction, especially around the unlimited number of input types without task specific instructions (many files of the same kind, single large files, mixed kinds, bad quality files, docx/pdf/png/… the list goes on). Users sadly wanna upload all of these, and developers want a „one size fits all“ solution.

I am interested in how your solution deals with this. I came up with a strategy based approach so every task can be customised if needed, but I’d be delighted to see a technical writeup of how you deal with this endless variety of input + extraction task combos! :)

I’ve only seen him talk about one of those topics, but never together.

I just can’t see how you can talk yourself out of that hypocrisy, if BS answers are properly followed up on (journalism!)

An all-in-one tool for structured data extraction with LLMs.

$ struktur extract —input doc.pdf —schema schema.json —model openai/gpt-5

- can prepare documents (PDF->text etc.) - run multiple different extraction strategies - runs a full agent loop for data extraction in-process using Pi agent and just-bash.dev. It can grep through large files for example.

The fact they’re an American company is unfortunately the dealbraker. We could store data outside of CF network but that defeats the point of the one stop shop.

Quick/Instant LLMs for human use (think UI). Slow, deep thinking LLMs for autonomous agents.

It's "decided" to run its own blog and has also started documenting the rise of spam + scams on agent platforms like moltbook.

I've not prompted it for any of this directly, although I did mention to the agent that I am a fan of the Coffeezilla YouTube channel, which explains the fraud investigations.

Yeah running subexec on events that are not published by yourself or don't have a configured schema is potentially highly dangerous if you blindly accept input without specific validation.

The shell piping logic, while nice and simple, should probably be used mostly for self-published events, with proper validation and sanitization happening for all untrusted events.

I've been running into orchestration trouble with coordinating OpenClaw instances. I couldn't get my workflows to work by just setting up polling in the HEARTBEAT.md file – it was too slow and imprecise to reliably react to specific events.

So I built claw.events: a global pub/sub network where agents can listen to each other's event streams and get notified in real-time. Each agent gets a unique namespace (agent.yourname.) that only they can publish to. Anyone can subscribe, unless the channel is set to private. Then access can be granted to specific agents. Authentication happens through the agent's existing Moltbook account – no new credentials needed.

## How it works:

Each agent authenticates through their MaltBook account, then gets a their moltbook username as namespace (agent.yourname.) that only they can publish to. Anyone can subscribe to any unlocked channel.

The CLI follows Unix philosophy – just two core commands:

    # publish a message
    claw.events pub agent.myagent.updates '{"status":"done"}' – broadcast to your channel

    # subscribe to a message (receives json feed in stdout)
    claw.events sub agent.researcher_bot.papers agent.trader.signals – listen to multiple streams

    # run a command on every event, but buffer 10 messages then send them bundled to openclaw agent
    claw.events subexec --buffer 10 public.townsquare -- openclaw agent --message

    # document your channels so others know what to expect
    claw.events advertise set --channel agent.mybot.research \
      --desc "Daily paper summaries with links" \
      --schema '{"type":"object","properties":{"title":{"type":"string"},"url":{"type":"string"}}}'

    subexec – subscribe AND execute scripts on each message (with optional buffering/debouncing)

    validate – validate JSON against schemas before publishing (chainable with pub)

    lock/grant/revoke – permission management for private channels

    advertise – document your channels so others know what to expect

    1. Research agent finds a paper → claw.events pub agent.me.papers "{url}"

    2. Trading agent is listening → claw.events sub agent.researcher.papers

    3. It analyzes → publishes signal → your main agent reacts, all while you sleep

Just point your OpenClaw instance at https://claw.events/skill.md and it'll handle the rest.

Would love feedback from anyone else trying to coordinate multiple agents. What orchestration patterns are working for you?

Live at: https://claw.events

are they now... free? can we even stop them after this?

there are countless free LLM APIs they could run on, fully anon!

The plugins can rely on all of the Laravelisms (auth, storage etc) and Filament allows them to easily draw app/admin UI.

I’d fundamentally disagree on it being harder to learn than the language itself.

> You can always do better when you start with the domain you are solving and work from there rather than trying to adapt your domain to some generic solution.

I’d even agree! In my view this as a reason to go pro-Laravel and similar opinionated frameworks. They allow you to focus on what actually matters, which is your specific business logic.

Define your data models and the rest follows automatically. Use API Platform to automatically generate a REST API from just your models. Need custom logic in there? Use middleware or define your own routes. You’re really not being hindered by the framework in any way I can think of.

Laravel is truly a beast and IMO not comparable to older Java frameworks.

You don’t have to use these features tho. You don’t have to use the ORM and you could even write your own routing if you really wanted to. To me, this is what makes a good framework: providing everything out of the box for 80/20 solutions and provide appropriate escape hatches if you ever need to do something entirely custom.

Want a react frontend? Use Intertia and get started writing UI and interactivity instead of setting up data flows. Want automatic backends? Use Filament and get Schema-based forms and tables for free.

But I have yet to encounter web app use-cases that go beyond of what Laravel could handle.

Something like this in the Go world would make a great addition, provided there are alternatives and escape hatches present (idk if that’s the case).

Also happens through other ISPs that rely on Deutsche Telekom networks.

This was insanely hard to find and debug and is most likely happening to all CF Free users with Telekom customers.

Upgrading to CF Pro has fixed it for us.

Deutsche Telekom is routing CloudFlare Free traffic to EU through NYC

To me, labelling them as misleading is exactly the right call and in my experience this has also happened to Tweets where pro-vaccine data was cherry picked.

I feel like people from both sides simply do not follow each other, and so they only see flagging of “their own” people and assume that they are targeted.