Research
Research
Long-form notes on building agents, training models, hardware, and the AI industry.
Agents
How to design, build and evaluate agents: runtimes, knowledge, memory, evaluation and product routes.
- Agent Engineering Atlas
Agent architecture, enterprise knowledge, memory, evaluation and product technical routes across 200+ connected pages.
In Chinese
Related writing
Models & training
How models are trained: data, objectives, world models, robot learning and control.
- Robot models, data & control
World models, VLA, robot training data, real-time control, and Stanford/MIT learning resources.
In Chinese
Related writing
Hardware & infrastructure
Hardware design and blueprints: materials, fabs, chips, memory, optical interconnect and data centers.
- AI Infrastructure Atlas
The physical AI stack, from materials and chips to optical interconnects and data centers, with traceable sources.
In Chinese - Wafers & fab manufacturing
Blank wafers, foundries, logic, memory and photonics manufacturing.
In Chinese - Optical interconnect
From lasers and silicon-photonic modulation to receivers.
In Chinese
Related writing
Industry & future
The AI industry as a whole: companies, capital, analysis, and where AI is heading.
- AI World
An evidence-led atlas of AI companies, funding and the ideas shaping the industry.
In English