The day the market priced in efficiency
Nvidia lost roughly $600 billion of market value in a session. The trigger was a paper about training costs.
page 12
older dispatches from the README archive.
Stargate is a $500 billion bet on a bottleneck
OpenAI, Oracle, SoftBank and MGX announce an infrastructure vehicle. The real constraint isn't chips — it's power and steel.
DeepSeek R1 puts a reasoning model under an MIT license
Open weights, a published training recipe, and API pricing that reads like a typo. The reasoning-model moat just got a lot shallower.
Bun 1.2 stops being a runtime and starts being a platform
Native S3 and Postgres clients, a real Node compatibility push, and a bundled package manager that keeps getting faster.
CES 2025 was a GPU keynote in a robot costume
RTX 50-series, a desktop supercomputer named DIGITS, and an entire show floor that decided AI is a hardware category now.
Rust 1.84 and the quiet rewrite underneath
The next-generation trait solver goes live for coherence checking. Most people will never notice, which is the point.
The year the model comes to the laptop
2025 opens with open weights good enough to matter and consumer hardware finally sized for them.