Living notes on LLM technology, recent progress, infrastructure, and the surrounding AI ecosystem. The repo currently uses standalone HTML pages for long-form explainers, technical notes, and visual summaries.
| File | Description |
|---|---|
ai-supercycle.html |
Long-form notes on the economics of the AI supercycle, including value accrual by stack layer, inference economics, power/data-center buildout, and where durable value may migrate. |
ai_layers_slide.html |
A one-page visual landscape of private AI startups by supercycle layer, with reported valuation signals through July 2026. |
inference-technology.html |
Technical notes on inference architecture and related AI infrastructure. |
Open any HTML file directly in a browser:
open ai_layers_slide.htmlIf GitHub Pages is enabled for this repository, the pages are available at:
https://hefeicoder.github.io/llm_ai_technology_notes/ai-supercycle.html
https://hefeicoder.github.io/llm_ai_technology_notes/ai_layers_slide.html
https://hefeicoder.github.io/llm_ai_technology_notes/inference-technology.html
GitHub Pages can take a few minutes to build after a push. If a page returns 404 immediately after enabling Pages or pushing changes, wait for the Pages build to finish.
The startup landscape slide mixes several types of valuation signals:
- completed private funding rounds
- reported private-market targets
- IPO targets
- acquisition or strategic transaction values
- undisclosed rounds where only raise size is public
Cards marked with * are not clean post-money funding valuations. Re-check fast-moving valuation data before publishing or quoting externally.
The files are plain HTML/CSS with no build step. Keep updates self-contained unless a shared style or generated pipeline is introduced later.