[trollbridge]

DeepSeep V4 Pro is 99% cheaper than similarly performing models were 2 years ago (if such a model even existed).

Computing has always been about how to wring out more efficiency. The ENIAC was 150,000 watts, with 3 phase 240 volt power, and cost about $500,000.

My day to day laptop (a year old) is 35 watts, with 1 phase 20 volt power, and cost $1,000, so that's 99.98% less power consumption, 99.8% cheaper, and it has about 10 orders of magnitude more computing power, all on a time span of 80 years.

[cratermoon]

Moore’s law is dead.

[HappMacDonald]

It died before AI came around and today's coding agents are somewhere upwards of twice as competent as whatever the state of the art of automatic coding was in 2020. 8I

[mrandish]

A good chunk of that was one-time gains from shifting GPU and memory architectures to better match what LLMs need at scale as well as some algorithmic improvements. Most of the low-hanging architecture optimization has already been harvested. We'll certainly have more algorithmic gains but the consensus is they'll generally be smaller and less frequent.

There's always a chance we'll have some dramatic gains far larger than DeepSeek's optimizations a year ago, but it hasn't happened again yet at even that scale. It would be nice but I certainly wouldn't count on it.