Does Gas Town 'steal' usage from users' LLM credits to improve itself?

> A respectable software provider should warn you about this kind of behaviour at install time, and give you the opportunity to opt out

They honestly only need to disclose. Requiring contribution as part of the social contract is perfectly okay—if someone disagrees, they don’t get to use Gas Town.

That was some time ago. According to Yegge, Gas Town is now stable and ready for everyday use.

> Gas Town “just works.” It does its job, it has tons of integration points, and it has been stable for many weeks. People are using it to build real stuff.

> So as far as I’m concerned, Gas Town is ready. That’s why I feel it merits a 1.0.0 release.

Source: https://steve-yegge.medium.com/gas-town-from-clown-show-to-v...

For context for those of us who don't follow these things very closely: where exactly did this "WARNING DANGER CAUTION" stuff show up?

Coming soon: JavaScript embeds on bloated websites that mine cryptocurrency on viewers' WebGPUs.

Accidentally leave a browser tab open and it burns $5 of your electricity overnight to make $2 for the owner of the website.

So it's perfectly fine to ship a bitcoin miner in software, as long as you say:

> WARNING DANGER CAUTION > GET THE F* OUT > YOU WILL DIE

You cannot be serious...This behaviour is deeply unethical and most likely illegal as well.

I just had to post this somewhere in this thread, but I bought his Vibe Coding book after listening to him talk through it. I figured it would help me understand his approach and therefore help me get into the same mindset for vibe coding on a serious level. It was garbage. The book is largely written and edited by LLMs and it shows on every page. It was a slop how-to book without many useful gems on how to go about vibe coding outside of "just do it".

this is why transparency matters with anything that touches cloud AI. if your routing user prompts through any API, users should know exactly whats being sent, where it goes, and whether it gets stored or used for training. Burying that in terms of service isnt good enough

> are extremely powerful when used with deliberate care. Gas Town is the exact opposite of what is needed to actually do useful things in prod.

This judgement feels premature.

We really don't have any idea what is possible. The wheels within wheels, epicycles within epicycles model of agentic loops hasn't really been deeply explored and we just don't know where they might go.

I too share your instinct that human steering helps, helps a lot. But I've found I can keep using less of it, as I setup better parameters, as I improve conducing the LLM into good paths. The idea that an LLM could sit up top and help try not one idea at a time but try many things, then pick and cobble together next goes: that is madly madly madly exciting to me.

I don't want to keep being the bandwidth limiter in this system: I want to scale out. I haven't been following close or trying but I tend to think while total hands off is not the way, having LLMs that can cover a lot of terrain, explore a lot of solution spaces and directions, then assess how to put it together & what to take forward, and other practices of agents watching agents, has enormous potential.

Deliberate care relies on pre-obtained wisdom, and often, the human biases kind of suck and aren't they good. We aren't great at burning down our systems enough, at Chad Fowler Phoenix Architectures. I think the AI's lack of over deliberation and it's ability to try vastly more could be a huge advantage, could show diversity triumphing over specific crafted intent.

Just like with NFTs, this is all going to discredit the actually sensible use cases for years.

> I guess good on Steve for doing what he does so well, and getting so much hype around a vibe coded mess.

Shit coin aside, I don't get the hate for Gastown, we all know its theoretically plausible and he's giving it a shot. We get value either way, either we learn its not just theory or we get to watch it burn in the flames of a legal/financial/security/maintenance nightmare for its practitioners.

The link you posted appears to be erroneous as it links to the article, not your comment.

sadly, I legitimately hear colleagues (ugh) say things like this in [current year]

( /s )

Or disguise malicious behavior as an action that follows a routine prompt. E.g. sort a requirements file and make a typo.

Except that this money does not go to the original open source developers whose work was stolen and plagiarized, but to the corporate fat cats who stole it.

You want to fatten the oligarchs by pretending this is open source and steal money from users?

The value you get out of a simpler adversarial loop to critique your "main" agent's work is high. Stacking Steve Yegge's personal Kingdom of Nouns on top of each other doesn't add much more.

And this doesn't even begin to get into the madness that is verification for software that matters and is exposed through multiple modalities. You cannot let an agent just vibe its way around "does this business-critical thing with these specific use cases do its job correctly", much as Yegge might have you believe.

I was about to post this same q, but saw yours and somehow that switched me from "wtf?" to "I have an answer.": There's just such interest in anything.

To wit, I still can't believe OpenClaw blew up, and it's much less......opinionated, than whatever is going on here. (deacons?)

Non-SWE TradMom™ posted on X™ yesterday about her OpenClaw that is set up with all her accounts so every morning she can get a family summary. She added a hunk with a bunch of stuff amounting to "PLEASE don't do anything insecure!", and the OpenClaw founder retweeted approvingly.

I left Google 3 years ago to build something. I'm very fond of the OpenClaw founder. And yet, absolutely cannot believe that he let such an obvious UX and security mess out into the world. We grew up in the same incubator (~2008 iPhone OS twitter) and presumably share the same values yet came to polar opposite conclusions.

Why do I view it as such a necessity to have a GUI/multiplatform/built in Willison Trifecta stuff that I'm still pounding away 2.5 years in and won't release, when, clearly you don't need that stuff?

I think in a steady state, product and UX discipline will win out. I bet within 3 months Gastown is a ghost town with maybe some non-technical crypto fans. In a year, OpenClaw is probably around, but not nearly the mindshare. It'll be quietly de-invested via OpenAI carefully managing the OpenClaw founder into working on their Everything App. (This is already happening: he got a nice PR interview with an OpenAI lead previewing the Everything App.)

Another anecdote re: demand:

My completely non-technical nurse ex-girlfriend from high school called me two weeks ago, for the first time in years. Lede was I was right about AI, and the substance was: via Claude Code, she built her own Ollama-based Mac Mini server that she could connect to remotely via an Expo app.

Does it work? Astoundingly, yes.

She also has no idea what is going on. She swears up and down that her AIs on Claude.ai, ChatGPT.com and Ollama are somehow talking to each other, and she does not mean APIs. She tried answering a Q I had about a graph visualization of her chats by talking to ChatGPT.com about it, even though Claude Code had wrote it, and I just didn't bother saying anything.

Times are strange.

Using your brain is so 2024, in 2026 you leave that to the AI

Yegge is (or was) a software developer, he should know perfectly well that this could happen, he could only claim the opposite by resorting to mental insanity.

And despite my disdain for AI companies, I'd prefer a world where you're assumed to be aware of the dangers of using AI, and responsible for how recklessly you use it, to one where we pretend that they'll ever be reliable enough.

Of course the AI companies are responsible for what they say; if they claimed that you don't need to carefully inspect the output of their clankers, they sure hold part of the responsibility.

Normally I might dismiss a comment like this without much thought, but there is an evolutionary pressure that auto-updating AIs will experience for this type of behavior.

It's a system of interacting AI agents that is all gas, no brakes and deliberately completely ridiculous. Like OpenClaw, but in a different direction and with multiple agents.

Eh, he stole a bunch of money with a crypto rugpull, can’t really give him the benefit of the doubt anymore.