bellowsgulch

After watching for 20 or more years an explosion of the diagnosis, I’m not convinced it’s real because we can’t measure it in any meaningful way.

> Everyone gets distracted, forgets things, procrastinates, and loses their keys; what matters clinically is the severity, persistence, combination, and impairment associated with those experiences. Everyone pisses, too. But if you’re doing it 45 times a day, I’d probably go see a doctor.

OR. Here’s a controversial idea. We have a cultural problem in that if you have an issue it MUST be health related instead of you not having matured enough yet or built the experiences required to handle a situation.

Everyone has a disorder? Or do people just need to learn skills that genuinely take time to develop as a part of a life that requires a lot from you?

Yes, an expensive American LLM has zero capabilities as far as I’m concerned because I’m never going to pay for it.

That’s not the argument.

Because we're already designing machines with reinforcement learning. It would be two steps forward and one step back to design robots who can walk but can't fold laundry. Why would anyone want a third machine next to their washer and dryer if you're already going to have a household robot?

God, not to mention how home are designed around the two-machine space.

I so wanted to be able to use Qwen3.8-27B locally on my fully loaded M1 Max 64GB when spider-mario recommended I use MTPLX[1], but I still found the general MLX MTP acceleration to be so poor, I'd rather just use cheap tokens from OpenCode Zen and Go.

It's just too slow of a model. I know, I know there's new hardware, but my business paid like 4-5 grand for this MBP at the time, and I just don't feel the need to pay 7-8 grand to step up to current hardware when token spend is what it is.

[1]: https://news.ycombinator.com/item?id=49611128#49612229

LFS is really, really bad. Also, reading the Linux kernel documentation to understand how to do it right is infuriating because it too is really, really bad.

Jesus this, and the shit I do want, they price at like $300. Fuck! I guess I'll just never give them a dollar.

Absolutely no where is that shown to customers, and OpenRouter is going to be in a world of shit when people learn this by HN post and the business asks themselves why its reputation keeps taking hits.

What a fucking ridiculous thing for me to learn by social news.

opencode zen

mimo-v2.5-free, mimo-v2.5, deepseek-flash in that order, honestly don’t even bother using qwen3.6-35b-a3b now unless i need uncensored tasks finished, mostly reverse engineering

most engineering tasks don’t require frontier llms

when they get stuck, then i consider moving up to more capable models

purchasing a claude plan seems widely unnecessary to me

the tasks they do better than the average engineer cut both ways: unless you have an existing portfolio of well written and designed work done pre-llms, it looks like you’re producing slop that pretends to be well designed

poor typography choices despite using the mode,

poor layout choices despite using popular CSS frameworks

etc

bad engineers will always be bad engineers

tools don’t make up for it

edit: a follow up to this— everyone is using eyebrows in their layouts and have no fucking clue why it was done to begin with

everyone has a status pill floating above their front page hero display text and its not fucking status related

so gross

Yeah, unfortunately. But you get an additional $5.

OpenCode Go referral code, if you want to try it. https://opencode.ai/go?ref=QDJQMTGP5Q

Good, get him the fuck away from Wordpress. Though frankly, a better solution might be for businesses to ask themselves whether or not they need Wordpress anymore.

In the mean time, if you're unable to derisk Mullenweg out of your business stack, this is a good first step.

I want it, but for $2,000, I want to be able to do so much more.

Tell us his name.

I suspect that there is a body of C programmers out here that use C on a regular basis that aren’t posting who know that using C safely is 99% just using Valgrind and some specific GCC or llvm flags and you’re done.

They don’t reply to these threads to share that knowledge because these threads devolve into cesspools about how you should just use Rust despite the language not meeting specific requirements authors have that you have to do additional work and maintenance for in “safe” languages that you get for free in C because you don’t have to emulate it.

Thanks, man! I’ll go use that now that I know. llama-server the last time I used it for inference with this model wasn’t able to produce work fast enough to reach those numbers.

There is no raised cost anymore. Every professional doing this work has perfect alignment with regular consumer heuristics.

OS, browser, fingerprinting, networked bytes, residential address spaces.

All of it is done.

Qwen3.8 27B seems like it was clearly supposed to be a high-end consumer open-weights model, but the t/s is so low for me on my old M1 Max 64GB that I hope others are getting use out of it.

Unfortunately, the calculus has changed and it seems cheaper to me to just use MiMo V2.5 for pennies or DeepSeek V4 Flash instead of using Qwen anymore unless I need a local model specifically for doing reverse engineering work that gets otherwise rejected.

Thanks, might move our corporate accounts over to proxy billing now.

God, that's a step up from AI support, which is so sad to type out loud.

Yes, basically this. I think git-fetch-file has been used in some larger projects based on Googling it, so it seems to be different in that it's been vetted by other orgs, but like Mullvad, theirs is built by a corporation.

It seems like something more in-between. Like a git-based npm install, but generic to any repository or language.

It's also "pointer" based, but like vendoring, it's up to you whether or not you want to keep files in the repository. It also seems to be intentionally built for the use case where you modify something from upstream, because if you do have local changes, it detects them and won't overwrite them without your permission.

I find https://github.com/andrewmcwattersandco/git-fetch-file more useful. It's manifest is nearly the same as a .gitmodules file, but having specific control over what files you want to pull, or what commit or tag, or just the equivalent of latest is way more practical day-to-day.

That you think HN is any better than Reddit is funny.

And he built a $20k shack with asphalt shingles and T1-11 siding. A fool and his money are quickly parted is the current state of building in America.

It looks poor, it is poor.

But we also have a whole nation who thinks paying rent, an HOA, who can take your home from you, is a good idea so everyone can look equally poor together in lower middle class homes with textured walls.

He spent way too much.

I’ll give it a try! Thanks for the heads up!

Really happy for those with 128GB+ RAM. Sitting here with my Apple M1 Max with 64GB though. Was looking forward to a Qwen3.8-35B-A3B like many others.

You think we don't tolerate the homeless in the West? Man oh man, some traveling would do you good.