Chinese IQuest-Coder Rivals Claude, AI Adoption Faster Than Internet, Compute Doubles Every 7 Months

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:00

   Good morning Legion. Happy Monday. The mailbag has officially closed, and you can expect responses to the mailbag sometime this week. We weren't expecting over 500 submitted questions, so this might take some time, but it's a good problem to have.

2. 0:13

   As a general reminder, we have our first Eco Collab Call of the year is on Wednesday, January 14th at 11 a.m. eastern Time. It will feature a look at the NEAR marketing team's plan for 2026, as well as a first look at the NEAR Merch Store, followed by an introduction to Hako Labs and their intent based trading product. We hope to see you there.

3. 0:33

   Three stories from the AI world today, and a lot more on the docket for this week. First up, a new Chinese coding model rivals even Claude Code. IQuest-coder from Quest Research, backed by China's quant hedge fund giant UBIQUANT recently dropped with some phenomenal benchmarks and evals.

4. 0:50

   Importantly, all models dropped are natively supported, with up to 128,000 tokens without requiring additional scaling techniques. That's great for context. This comes at a time when a new DeepSeek coding model supposedly on the horizon as well. Is 2026 the year of Chinese open source models?

5. 1:07

   Time will tell, but surely don't sleep on IQuest coder from Quest Research. Our second story is a quick one. It's a simple graphic that shows the adoption curve of generative AI. Three years in, relative to things like the internet and the mobile phone.

6. 1:20

   What they should tell you is that AI adoption is vastly accelerating relative to the adoption of other technologies. If adoption continues on this pace for humanoid robots and more advanced forms of AI, then an AI dominated future is absolutely coming sooner rather than later.

7. 1:37

   Last but not least, epic AI finally gives us an estimate of the doubling time for global compute. Total compute is doubling every seven months, and 60% of that compute is comprised of Nvidia chips. Since 2022, total compute has grown roughly 3.3 times per year, and it shows no signs of slowing down in near future data centers.

8. 1:58

   Water, energy, and materials for data centers are expected to be in ample demand with this AI future ahead of us. That's all for today, folks. We’ll do this all again real soon.