logoalt Hacker News

dgreensp • today at 2:17 AM • 2 replies • view on HN

A lot of this sounded (from a quick read) interesting and insightful.

I wasn't sure about these two paragraphs:

> If AI had started a new wave, there would have been an extended period of uncertainty and experimentation. There would have been a population of early adopters experimenting with their own models. When thousands or millions of tinkerers use the tech to solve problems in entirely new ways, its uses proliferate. But because they are using models owned by the big AI companies, their ability to fully experiment is limited to what’s allowed by the incumbents, who have no desire to permit an extended challenge to the status quo.

> This doesn’t mean AI can’t start the next technological revolution. It might, if experimentation becomes cheap, distributed and permissionless—like Wozniak cobbling together computers in his garage, Ford building his first internal combustion engine in his kitchen, or Trevithick building his high-pressure steam engine as soon as James Watt’s patents expired. When any would-be innovator can build and train an LLM on their laptop and put it to use in any way their imagination dictates, it might be the seed of the next big set of changes—something revolutionary rather than evolutionary. But until and unless that happens, there can be no irruption.

> ... In the 1970s, it took time for people to wrap their heads around the desirability of powerful and ubiquitous computing. But in 2025, machines that think better than previous machines are easy for people to understand.

Even if tinkerers aren't currently training frontier LLMs from scratch, there's still a lot of experimentation happening, and I don't think everyone even realizes yet how transformative this technology will be. On the contrary, there are a lot of skeptics.

That said, it's a good question, who captures the value.


Replies

strken • today at 2:27 AM

On the question of tinkerers, I'm reminded of the man who just copy-pasted the middle layers of an LLM on top of each other: https://dnhkng.github.io/posts/rys/.

Turns out you can just do that!

johnnyanmac • today at 2:21 AM

>Even if tinkerers aren't currently training frontier LLMs from scratch, there's still a lot of experimentation happening, and I don't think everyone even realizes yet how transformative this technology will be.

The big issue of vendor lock-in remains for the frontier models. We saw a small piece of this in the games industry when people started reverse compiling games at unprecedented speeds, and companies' answer was to push Anthropic to refusing to process such prompts. A very easy workaround for now, but they played their hand and revealed their allegiance.

Any innovation that threatens to disrupt with current structures will be stifled by the old guard.