[edoceo]

I'm also feeling like that, not a quick burst like dot-com but a deflating bubble over next 24 months. During which AI costs rise and IPO valuation comes down.

Maybe my memory of dot-com is fuzzy (likely) and perhaps I didn't see early warning (or even didn't know how to see it). I feel like it all transpired over 3 months. And for those of us in Seattle the "death-knell" was the Nisqually quake that made much of Pioneer Square unsafe to occupy for a while.

[dofm]

One of the things that really strikes me is that there are so many people talking about how great the frontier models are but there's nowhere near as many people talking about open weights and local models.

There is this huge knowledge gap here about what the local models could do in terms of consumer queries with a tiny bit of agentic support. If more people could tinker with local models and see what they can do, I think there would be far less belief that only the big two/three hold the keys to the kingdom and far more that the future is a bit more distributed.

On that basis I am always on the lookout for coverage that gets into this — Cal Newport touches on it a bit — because I think it is one of the deflationary factors, as is the quality of open weights models, which in turn is why Anthropic is now getting credulous journalists to write about "distillation attacks".

Re: the dot-com crash, I was in the UK working for an internet startup at the time though I was on a sort of health break/sabbatical at the time of the pets.com crash, so I was reading and researching, and while I do remember the coverage of their valuation being insane, and agreeing with that, I also think it did not have such a long run-up. We did know in 1998 and 1999 that Broadvision's valuation seemed absolutely insane, for example, and that some e-commerce developer and design startups were radically overvalued, but what is so different about this time is how much more broadly things were distributed then. So many players. Even the major duality (which was Microsoft vs ABM — a loose coalition of "Anyone But Microsoft") had many more players.

Whereas now… it feels like a prequel to Rollerball.

[JumpCrisscross]

> If more people could tinker with local models and see what they can do, I think there would be far less belief that only the big two/three hold the keys to the kingdom and far more that the future is a bit more distributed

Eh, I've played with them. They take way more babysitting to do the work reliably, and I have to much-more closely monitor the outputs for hallucinations.

They'll get there. But it might take a decade or so for (a) consumer hardware and (b) SOTA distilled open-source models to converge with the closed-source stuff, and that's enough time for both a slow deflation and plenty of profits to be made.

[dofm]

I was specifically referring to consumer AI, not agentic coding, and I really think Apple have a good chance of demonstrating smaller on-device LLMs plus traditional symbolic AI work (even just rules systems like evolved forms of Automator and Shortcuts in harnesses) can do much of what consumers want. Agentic coding is a giant red herring, in a way.

I don’t think these vast LLMs will be the dominant approach in a decade, frankly, because the performance gap between a 35B MoE and these absolutely gigantic models just is not proportional to the scale difference, but at this point we are just pitching belief against belief.