Silagi

Realistically, it's never going to be Pi, for the same reason that video game consoles are always more performant than PCs for a given hardware spec.

There's no canonical Pi harness, it's too variable. There might be a world where something Pi based is the target, like OMP, but then you have to hesitate when you start extending the harness because you don't know what effect any given extension will have on model performance.

The irony is, this would be a pretty trivial one shot project for a modern LLM to create that interfaces with the source control repository of your choice. It's basically an advanced note-taking app, which LLMs have been trained to death on. I bet you could get it to write it as a VS Code extension as well (or emacs macro, or... etc.)

I've been running 2 threads of Opus 5.5 for ~16 hours on a server C++ to Rust translation/optimization project and used 8% of the 20x sub. Going to test out gpt6 sol over the weekend, but it really seems like we're back in the realm of having to try to burn a 20x sub.

I think this is where Deepseek has nailed the mark; DS4.1 Flash is really, really fast, and really, really cheap. If you give it small, structured goals, it completes them crazy quick, at negligible cost. There's different vectors to differentiate along to stay in the conversation.

I've taken to using them as micro-review subagents at development milestones, where a "frontier - 1" model like Opus or Sol launches 10-15 of them on small review tasks that each run for ~10 minutes. Costs about $1 per cycle, and they usually catch something Astra or Fable didn't. Then the orchestrator validates each claim before passing it back to the planning session so we can fold the findings in.

Adjacent, I work in a lab so we have dozens of ovens and freezers running constantly, so I'm intimately familiar with the constant hum, even though my office is a couple of doors away from the oven room.

We had a full site shutdown last week, which involved all power being shut off, which meant no running ovens.

The lack of sound felt so strange, like I was somewhere I shouldn't be, or in some sort of liminal space. It was unsettling in the silliest way.

This is really the only reason I have a MacOS system around. The M4 Air feels like what I imagined a laptop could be 15 years ago. No fans to block, can grab it and carry it around almost like a tablet, works great as a thin client into a more powerful system over Moonlight, it's light enough that I don't even think about carrying it around in my backpack even when I'm also carrying my work laptop, and it was cheap enough that if anything happens to it I won't be too concerned.

Compared to the last XPS I had, it's substantially better for my use case, since I'll generally be remoting into systems anyway.

You could realistically be burning through ~5k of tokens without hitting the quota. The subscriptions are crazy subsidized relative to the API rates.

That model simplifies thought into a painfully linear process, and overestimates the creativity people put into the "small decisions" that push a project forward. Most decisions are arbitrary and need to be reconsidered later anyway. And any real creative work has a "push it far enough to find the edge cases, then go back to the initial design and iterate" loop cycle anyway.

AI significantly speeds up the "Find where this spec breaks down, then lets go back to the design stage" in a way that should enable any creative person to create more interesting and useful work. If the output is slop, that reflects on the operator, not the tool.

I'm convinced this "signal" has already been hijacked. Maybe a Baader-Meinhof phenomenon, but I've noticed more and more egregious spelling errors that make little sense from a human perspective. Hop into whatever chatbot you'd like and ask it to "write a paragraph with subtle misspellings on long but common words", and you'll notice misspellings that just feel wrong, because they don't map to a clear misunderstanding that a person could have.

Or maybe I'm losing it after reading too much slop. Also distinctly possible.

Did you consider the R9700 or B70 when you went for the MI100? If so, what made you choose the MI100?

I've been playing with picking up a card in this class but haven't been able to justify it when running the Qwen3.6 MOE model on a 6800xt is tolerable for the type of projects I've been willing to point local AI at.

Are you increasing your water intake when you do? That sounds like dehydration. Creatine takes a large amount of water to appropriately process, and during the loading phase, your body is pushing substantially more water into your muscles. Anecdotally, if I don't drink something like an extra half gallon of water a day while loading at 15g/day, I show symptoms of dehydration. And I'm already drinking somewhere between a half gallon to a gallon a day.

I think we're roughly on the same side here? I'm explicitly against moralizing about AI by people who benefit from other forms of tech exploitation, e.g. using Amazon for hosting; I'm not against the idea of AI as a whole; I think it's an inevitability.

I just believe all of us should be willing to accept our own responsibility for the state of it. I also believe it could eventually work out to be a net benefit to the species; but in the short term, it's going to hurt us badly, as most technological revolutions have. I'm saying everyone involved, directly or indirectly, should be willing to accept their fraction of responsibility for the people who are suffering in the interim, and that moralizing about it is disingenuous.

I'm pretty sure we're not disagreeing at all. Over time, most technological revolutions have benefited society. That doesn't mean they didn't exploit people, though. Even in the current world, there's more slaves than there have ever been. People are still being exploited by technology. It's just more diffuse, and (I'm hoping) the average suffering in the world has gone down.

Sure, but those people who "considered the downsides" shouldn't then moralize about the exploitation of workers; they're contributing to it. It's explicitly hypocritical. They're explicitly deciding the exploitation is worth the upsides for their or their company's benefit.

I'm not excluding myself from this. Just chafing against this grandstanding by people actively contributing to the same problems, and especially annoyed by the people saying we should pause development because it's going to affect people's jobs. This isn't new.

Edit: I'd also like to point out that "Public cloud computing" doesn't really capture what I said; the OP of the article is specifically building on Amazon, which has a well documented history of worker exploitation. Even building on Azure or Google cloud would be more defensible in the context of the article they wrote.

Most do, in my experience? I worked at Argonne for a while, and they absolutely treated their profession with deadly seriousness. I didn't meet anyone who took the stance of "Move fast and break things". Most spent too many late nights rechecking their work to make sure it was correct. Even when being wrong would be entirely inconsequential in the grand scheme of things.

I get what you're saying, that not everyone who is in tech contributed, which is fair to an extent, see my other comments.

More specifically, someone actively building on top of Amazon decrying the exploitation of workers and the environment is demonstrably hypocritical.

And "engineers building things to spec" are not the same as those giving the orders, but they should take a measure of responsibility for the things they build. I think most people generally agree about the culpability of those following orders when people are harmed. I'm not even saying they should necessarily be held accountable; just that moralizing about it is hypocritical.

I vehemently disagree that meta or airbnb have done more to benefit society than not, but I'll take the overall argument; that technology, on the whole, benefits society overall.

Which is true, on the long term. But we have no reason to believe that AI will be different in that case. In the short term, technologies have absolutely been used to exploit the average person; the industrial revolution benefited us all over time, but tell that to the kids killed in early industrial manufacturing centers.

Look at how the transition to globalization went in the 80's-10's; entire sections of the US were essentially shut down because of the improvements in communication technology, and unless you're in support of the current state of the US, you'd agree we're still dealing with the consequences of that.

Even in your own case, there's an argument to be made that CFCs meant the overall damage to humanity was greater than the ice trade, just spread out over more people. The exploitation was similar, but it was less visible. Even if we've eventually reached a point where people were better off, you can't argue that the health of the average person was never exploited for the benefit of the few.

To be clear, I'm not separating myself from this; I'm fully aware that work I've done has displaced people. I'm just chafing against the moralizing around it. It feels like the people making these arguments are trying to remove themselves from responsibility while continuing to build on top of companies like Amazon, that are built on top of exploiting people that absolutely cannot advocate for themselves.

I'm specifically saying "I'm entitled to displace people with automation built on previous work, but automation that affects me shouldn't be allowed" is a particularly hypocritical take.

The implications of AI aren't as novel as tech circles would like to believe. The same trends in employment and automation have been happening across industries for decades in slightly different forms. This is just the first time it might actually affect the people doing the work, instead of being conveniently separated from their inner circle.

And you're cutting an awfully wide swath in the opposite direction; most tech gains value by exploiting or displacing people. Economies of scale don't just exist at the absolute top of the economy. The computer cut out entire classes of people from jobs they had specialized in by decreasing the education or effort required to successfully complete tasks, at the cost of massively increased infrastructure costs.

I'm all for pushing back against what AI might do, but doing it in this massively dishonest way just opens the door to obvious counters.

Controversial take: It's weird to see people in tech taking this stance. They've been riding the same wave of exploiting the average person through economies of scale for the last 20+ years, but now that it affects them, it's suddenly catastrophic.

You dont get to benefit from the expansion of companies like Uber, airbnb or meta, then pretend like you were always focused on the success of the average person. You didn't care when you could get ahead, don't pretend like you care now. It's childishly performative. This is an evolution of the same automation and communication tech that has been growing for as long as most people have been alive. Just now it might actually affect the technologically literate class. You did this. Own it.

I've always thought about doing something like this in the Midwest US, but was always a bit nervous about condensation damaging the components over time; did you run with that sort of setup consistently, or only when pushing high scores? Ever run into issues with components failing?