Fetching from the wire…
01
02
03
04
05
06
07
08
Models2026-07-26 · source-backed
slvDev's esp32-ai (MIT, 1.2k stars) runs a TinyStories-trained model on an ESP32-S3 with 512KB SRAM, 8MB PSRAM, and 16MB flash, printing generated text to a small wired screen (GitHub). The trick is a Per-Layer Embeddings architecture parking a 25M-parameter lookup table in flash while only the compute core lives in fast SRAM, quantized to 4 bits at 14.9MB total. Plain C on Karpathy's llama2.c. It cannot answer questions or follow instructions, only writes short stories, and that honest framing is exactly what makes the memory-hierarchy trick worth studying.
Each link below shares sources, entities, or timing with this story.
Satya Nadella said companies routing everything through a single proprietary lab may not survive. His argument: you hand that lab your most sensitive business context, and the lab can turn it against you as a competitor. His prescription is an orchestration layer — keep the ha...
GPT-5.6 Luna went to $0.20 input / $1.20 output per million tokens on July 30. That's an 80% cut. Terra dropped 20%. Luna's input now undercuts Gemini 3.1 Flash-Lite ($0.25/$1.50) and sits at one-fifth of Claude Haiku 4.5's $1 input. Simon Willison covered the announcement and...
Simon Willison doesn't hand out superlatives. So when he writes that Z.ai's GLM-5.2 is "probably the most powerful text-only open weights LLM," that's worth stopping for. His June 17 evaluation walks through a 753B-parameter Mixture-of-Experts model with 40B active params, a 1...
One Python file wiring a model to exactly one tool, bash, with the system prompt loaded from a file the agent can rewrite. Tools, memory, skills, conventions all have to be grown by the agent session by session into a self/ directory that persists. Model access goes through Si...
The first major version bump of his widely-used Python/CLI SQLite toolkit lands built-in schema migrations and savepoint-based nested transactions on June 21. If you lean on sqlite-utils as glue in data or LLM pipelines, migrations kill the hand-rolled ALTER scripts you've bee...
Fourteen times more commits. Not 14% more. 14x. GitHub COO Kyle Daigle dropped this number in a conversation covered by Simon Willison: the platform is currently running at 275 million commits per week, which annualizes to roughly 14 billion for 2026. In all of 2025, GitHub sa...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.