Hacker Newsnew | past | comments | ask | show | jobs | submit | woadwarrior01's commentslogin

> Goal: 200mb RAM usage.

We've come such a long way from: "Eighty Megs and Constantly Swapping" :D


I was expecting something closer to Van Emde Boas trees. :D

I'd called out fraud (blatant lying in investor updates) at a VC backed startup where I was a technical co-founder, once. I emailed all the investors and presented all the evidence to them. They decided to not rock the boat and keep my charlatan co-founder. So, I left. Now, the company is slowly bleeding to death.

> Now, the company is slowly bleeding to death.

There are thousands of companies where the shady practices are rewarded, the companies thrive and make money for the investors. So the investors are incentivized to reward this behavior just on the chance that they are rewarded back.

Whistleblowing sinks those chances and the investors and VCs know it. It doesn' just take away the money, it even takes away the plausible deniability. They put a lot of effort to absolutely punish any whistleblower to discourage the rest. Anything for a dollar. and this is probably all you'll ever need to know about almost every VC out there. Beyond the witty "I'm rich so I'm smart" blog posts and tweets, they're very much just the "anything for a dollar" type of people.


if they touch the federal government, feel free to ping me. I can walk you through how to report to people who will actually do something about it and do so anonymously

To be fair, I’m not sure blatant lying in investor updates alone constitutes fraud. There needs to be harm (or the intent thereof) AFAIK. The other party needs to be using that information to make a decision. If you give me a dollar and then later I tell you I’m actually Beyonce, is that fraud? Or am I just a lying sonofabitch?

If I give you a dollar and you say it’s being spent wisely, Beyonce loves the product, you’re about to land Taylor Swift as pro bono public ambassador… yeah that’s fraud.

It's encouraging future investment on a false pretext. I'd say that's fraud.

Lying in investor update was merely the tip of the iceberg. There was lots more, fabricating customer traction pre-investment, paying oneself back-pay for months spent twiddling thumbs pre-investment (before I was involved), etc.

My lesson from the whole kerfuffle was that investors (at least the ones I’d dealt with) prefer hustle over integrity and execution abilities.


This makes sense because investors in startups just care that they aren't left holding the bag. As long as they aren't the final fool in the buy in chain, they don't care.

Also, Qualcomm and ARM aren't quite in good terms.

https://www.qualcomm.com/news/releases/2025/09/qualcomm-achi...


1-bit g128 with a shared 16-bit scale for every group. So, effectively 1.125 bit.

Yeah, this reminiscent of Bandura's moral disengagement work. The differentiating factor isn't "do you follow authority?" but "do you maintain attentiveness to the consequences of your own actions."

The refusers likely weren't necessarily more heroic or principled in some deep character sense. They were just still paying attention.

https://albertbandura.com/albert-bandura-moral-disengagement...


> Sorry to shatter your bubble, but this is patently false, LLMs are far more efficient on hardware that simultaneously serves many requests at once.

You might want to read this: https://arxiv.org/abs/2502.05317v2


There's a very similar afm CLI that can be installed via Homebrew.

https://github.com/scouzi1966/maclocal-api


done

  brew tap Arthur-Ficial/tap
  brew install Arthur-Ficial/tap/apfel

No need for the extra tap step, this works fine alone:

    brew install Arthur-Ficial/tap/apfel

It's been around for a long time. It's from well before vibe-coding was a thing. I first saw it ~3 years ago.

oh okay. I had no idea and was lazy to check the commit logs.

Yeah, even if one efficiency trick lands, people will end up spending the saved budget right back on bigger models, and/or more "thinking" tokens.

Not if the bigger models have diminishing returns. Lets say you figure out a way to reduce RAM requirements 100X, but 2x increasing RAM usage by 2x only gets you a 1% increase in effectiveness and 3x does not get you any noticeable increase over 2x at all. Sure you can reduce the price per token, but you might have already saturated the market. Even if you haven't saturated the market, your hardware based moat just got smaller and this is going to reduce your margins even more.

Just noticed that pydry made a similar point: https://news.ycombinator.com/item?id=47574216


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: