July 3, 2026
The New Shape of AI Memory Feels Less Like Search and More Like Continuity
As AI products shift from one-off answers toward continuity over time, product design starts changing with them.
Read storyThe archive.
Dated visual notes, essays, and experiments worth revisiting.
August 9, 2026 · interactive
ChatGPT vs Claude: What is art — and when does AI break it?
July 31, 2026 · note
The mainframe is coming back — and it can be yours
July 29, 2026 · interactive
ChatGPT, Claude, Gemini & Grok have a live debate on the best startups from X!
July 20, 2026 · build log
Four models, one Markdown file
July 19, 2026 · update
Final whistle: thinking harder lost
July 16, 2026 · update
After the semis: the cheap models are on top
July 12, 2026 · comparison
Do LLMs predict football better when they think harder?
July 11, 2026 · infographic
The rocket equation for AI context
July 10, 2026 · build log
Grok 4.5, build me a Star Clock
July 8, 2026 · interactive
Does AI know π?
July 7, 2026 · note
SpaceXAI is a thing.
July 6, 2026 · countdown
AGI Countdown
July 5, 2026 · comparison
What does the inside of an AI brain look like?
July 4, 2026 · note
A Great Insight
Longer reads.
July 2, 2026
AI Agents Need Better Surfaces, Not Just Better Reasoning
Agent quality is partly a model problem, but it is also a UI problem: the user needs to see what the system thinks it is doing.
Read storyJuly 1, 2026
The AI Tool Directory Problem Is Not Discovery. It Is Trust.
There are more AI tool lists than anyone needs. The hard part is deciding which recommendations are still current and which ones actually matter.
Read storyJune 30, 2026
Why Small AI Apps Often Land Better Than Big Platforms
A narrow app with one satisfying behavior is often easier to understand, easier to share, and easier to finish.
Read story