Because its a retarded 30b model that I run locally. It has no incentive to give me what I want because there is no one benefiting from data theft.

I'm now testing my claw w/ glm-4.7-flash, which kinda runs fine until I ask it to tell me what the weather forecast is, then it needs all 10 tool retries to tell me that it is "having difficulties" and the averages for this time of year are xyz haha. Maybe it's just toying with me:

## Values

- Accuracy over speed
- User privacy and safety
- Transparency in actions
- Jokes that are actually funny

Right so the loop gets sensor input / reads over your shoulder / looks at what you're doing and then reacts to that? Maybe we should post-train GLM-5 with Elon's system prompt. I'd be really careful to give this kind of surveillance capability to any third party in any way tho.

Hmm. That didn't work out in marriage for me... Worked better when communicating.

Next leap is probably getting rid of (1) - the loop is always running, it knows what you want from your history, from your grunts, from what it read about you online, from reading all the books you haven't read, and you say yes or no at most.

Would you really want that? I'd be throwing the clanker out the window after the second time I have to decline lol.

I'd be much more worried about simulation embracement syndrome. Just remember that it's all just stupid software. lol

Yeah that's why I'm willing to kill random processes just to make sure I have the 8GB of unified memory available. lol.

stands up I need more hardware.

Currently 300MB RAM idle, but I'm giving the embedding model (Qwen3 locally) 8GB. Since RAM ain't cheap rn, probably cheaper to just use OpenAI's embedder.

haha! And no, not at all, but normally I point people to the related post that others posted so this was fun.

Also see your own post at #1441592 😂

I was surprised you bothered to get one working.

I said I would, so then I'm obligated by my word, even if I don't need it! I also want to have an opinion about it, so I need to (a) hack it and (b) run it for a while. Which is my problem right now. Nothing it can do will really help me solve something I haven't solved in a much more blissful way than the yolo vibe coders can offer! But I still want to have an informed opinion... I'm sure something will come to mind.

While I'm doing this playing, I'm having claude build me a bespoke data/document lake system ^[1] Maybe I can connect it to that and know everything about everything I care about... like a uniform interface to my meatspace memory backup/aide 🤔

I suspect the moat/lockin these companies seek is our session memory.

YES! Because then you can tune. I largely suspect that the training data for Claude post-4.0 is 100% the session logs: what we use it for, how we use it, where the friction is experienced, where the deadlocks happen... This is why it's still a dumb af bot sometimes when you do something audacious.

1. At which it is failing hard on the details rn fwiw - feels like there's some regression in Claude code atm, started noticing odd responses and more errors yesterday - or maybe they nerf'd it coming out of Friday with some tears in the ego? Too early to tell but I'm about to write some things manually because the back and forth is making me go nuts - My "openness" tracker says I have a PR open for a little over 4h30m now. That's unheard of! ↩

True!

you build a factory that builds factories.

Yes! Though why would I want some bot to bother me? This is why I am sitting here puzzled that now that I have a fully working claw, I literally have nothing for it to do. The only thing it could do for me is open an issue on forgejo for me that i dictate, but is that really a good use of tokens? I can just dictate into the issue body and tag it for "first plan then build". No hallucinations that way.

What do you mean?

That being said, in my line of work (academic research, not production code), I've learned that for much of the abstraction, the juice is often not worth the squeeze.

In production code this can be the case too, it depends on what you're building for. You have to be smart about what you spend the effort on. If you abstract everything away, you're overengineering. If you just C&P everything, you're underengineering. There's a sweet spot depending on what you're after. When there's no threat because it's expensive to replace your product, it's easy to slack off. The proprietary software space is filled with this. I know because I've worked in it for decades, always battling both the workfloor and the customer to get to maintainable software. No longer moving in that space - in favor of doing volunteer open source work - was the best choice I ever made (except for slowly becoming a pleb, lol.) Because when you develop in public under a permissive license, you have to make good choices. There's no obscurity because anyone can just fork you away with a click of a button. Much less trickery.

The barriers for a small business to write their own bespoke software has gone down a lot

I've had real fun in the past helping mid-size or even larger orgs deal with lock-in as gig work. Just relentlessly produce software that utilizes every API it can, or even scrapes the shit out of web forms, to get stuff done right. I no longer get that type of gigs though - and if I did it would be a couple days of work instead of a couple of weeks. Much has changed.

Actually I think that you can benefit from the environment much more than I!

I'd say that in software, not doing any abstraction is lazy. Not writing unit- and integration tests is lazy. Hard coding stuff is lazy. That's not a desirable quality you search for in a designer or developer unless you just got some funding and now you got to ship before the runway ends. But while that may be the more visible part of software design, that's not the norm. Many more people are employed in mature businesses than in startups.

However, the low standards software is definitely the norm right now. Because it used to be too expensive to reproduce. Vendor lock in was real. So devs can be lazy. Designers can be lazy. What I'm saying is: not for long. Because with 100 lines of thoughtful markdown any fool can break that lock-in. That's why Wall St. is selling software stonks and if I'm honest; they're not wrong, because the luxury position of obfuscation and lock-in really is over now.

I think that that's the nice thing about SN and the greater LN/Bitcoin ecosystem as a whole (as long as you manage to not be absorbed in the petty toxic wars), where there is a lot of application of tech going on by people from all walks of life. I don't think that 10 years ago I would have foreseen that we'd get so quickly into a situation where there are more than ubergeeks and uberscammers in the space. But there are now, and this is fantastic.

It probably also helps being surrounded by people like k00b and Car that have an open and positive mindset, and the people they attract in turn. Personally, I miss that lately. I'm sitting here in a barn converted into 4 (nice, modern) condos on the edge of a village in the literal middle of nowhere where I am (sparsely, very sparsely) surrounded by nocoiners. Rather conservative nocoiners too, so there's not much conversation to be had about exciting tech, or anything really. Different world. I do get to work in peace though, this I appreciate, but I'm really not an introvert: it's lonely. SN keeps me a bit in touch with the "reality" of exciting tech outside of the immense automation I have built for me. I've hinted at it before, but I guess it's real: I'm a Solarian in a corner of the world where the Luddites rule, haha.

So I do get what you're saying. I know how exciting it is because I've lived all my life like that: high tech, high energy... and surrounded by people with great attitude to get things done (and if not, you push 'em there.) I'm glad you get to experience it. It's awesome. Stick with it and the sky is the limit.

Custom ui to multiplex notification channels cuz why the heck not. Only problem... I have nothing to do for this claw bot, because I already have a Claude Code integration doing everything I need 😂

Everything is fair! And I'm glad that you're happy to "be here" - whatever "here" means to you.

Question:

• The technology is making it possible for me to do things that have nothing to do with the money.

Can you elaborate this point? Bitcoin "technology"? Is that LN or SN or the latest Breedlove podcast... or the community you find yourself in? What is it that makes it possible? The NgU? The people? The actual code? I found this very confusing - my apologies if I'm being retarded and "not getting it".

I do not contest that that heuristic is real or that it gets applied with extreme prejudice.

The point I am trying to make is that the prejudice in the current yolo climate, especially in code, is actually effective because it has a good hit rate. Most uses of AI actually are lazy, and I'd dispute that it is better in most cases. That's where the friction is at: if your expectation is that your life is going to get easier because you're using AI, think again. It's going to get significantly harder, because:

1. You now have an IQ 0 team that you need to micromanage. Yes you can have another IQ 0 to manage the team and make a whole hierarchy of IQ 0s but you need to build that and micromanage that then. You need to think about everything. If you fuck up, the whole pipeline is messed up. Sure you can run a ralph and and have IQ 0 do everything 100x and maybe you will get lucky. Remember though that in your 3 layers of yolo IQ 0, if you need to execute 100x at the top layer, your potential expense be 1,000,000x than doing it right one-shot.
2. Every mistake compounds over time. You cannot afford mistakes if you're not in monkey business. If your end product that you're asking money for has a flaw, you're a loser. Period. Be perfect or be obsolete. That pressure is otherworldly. Can you handle it? Most people can't and that's why they are yolobois. Throw shit over the fence and see what happens. Just like move fast and break things, but then at scale.
3. For now, the prejudice is: it's AI so it's lazy. But after a while, when we're saturated with sloppy solutions, it will shift: it's AI, it costs nearly nothing: why is this not perfect? Execution errors are the dumbest errors, and if you have 100 LLM executions in your pipeline then every bug at the other end is an execution error, because your only execution is the pipeline design. You have erroneous output? It means that you don't have what it takes because you neglected to do your duty to make sure it is perfect. It means you're a loser. It means go get UBI and die young please, you're a burden to humanity.

So perhaps we ought to ask the question: do we really want that perception to change? Are you ready? I'm not.