> any factor of 10 being a new science / new product category,

I often remind people two orders of quantitative change is a qualitative change.

> The thing that I’m really very skeptical of is the 2 month turnaround. To get leading edge geometry turned around on arbitrary 2 month schedules is .. ambitious. Hopeful. We could use other words as well.

The real product they have is automation. They figured out a way to compile a large model into a circuit. That's, in itself, pretty impressive. If they can do this, they can also compile models to an HDL and deploy them to large FPGA simulators for quick validation. If we see models maturing at a "good enough" state, even a longer turnaround between model release and silicon makes sense.

While I also see lots of these systems running standalone, I think they'll really shine combined with more flexible inference engines, running the unchanging parts of the model while the coupled inference engine deals with whatever is too new to have been baked into silicon.

I'm concerned with the environmental impact. Chip manufacture is not very clean and these chips will need to be swapped out and replaced at a cadence higher than we currently do with GPUs.

I think the next major innovation is going to be intelligent model routing. I've been exploring OpenClaw and OpenRouter, and there is a real lack of options to select the best model for the job and execute. The providers are trying to do that with their own models, but none of them offer everything to everyone at all times. I see a future with increasingly niche models being offered for all kinds of novel use cases. We need a way to fluidly apply the right model for the job.

> speculative decoding for bread and butter frontier models. The thing that I’m really very skeptical of is the 2 month turnaround. To get leading edge geometry turned around on arbitrary 2 month schedules is .. ambitious

Can we use older (previous generation, smaller) models as a speculative decoder for the current model? I don't know whether the randomness in training (weight init, data ordering, etc) will affect this kind of use. To the extent that these models are learning the "true underlying token distribution" this should be possible, in principle. If that's the case, speculative decoding is an elegant vector to introduce this kind of tech, and the turnaround time is even less of a problem.

For speculative decoding, wouldn’t this be of limited use for frontier models that don’t have the same tokenizer as Llama 3.1? Or would it be so good that retokenization/bridging would be worth it?

> The thing that I’m really very skeptical of is the 2 month turnaround. To get leading edge geometry turned around on arbitrary 2 month schedules is .. ambitious. Hopeful. We could use other words as well.

They may be using Rapidus, which is a Japanese government backed foundry built around all single wafer processing vs traditional batching. They advertise ~2 month turnaround time as standard, and as short as 2 weeks for priority.

Think about this for solving questions in math where you need to explore a search space. You can run 100 of these for the same cost and time of doing one api call to open ai.

The guts of a LLM isn't something I'm well versed in, but

> to get the first N tokens sorted, only when the big model and small model diverge do you infer on the big model

suggests there is something I'm unaware of. If you compare the small and big model, don't you have to wait for the big model anyway and then what's the point? I assume I'm missing some detail here, but what?