Founder of Becki and Astir. One memory. Every AI you use. Writing Model Agnostic, notes on building with every AI.
Joined August 2026
- Tweets45
- Following57
- Followers4
- Likes6
Pinned Tweet
Codex went from my sanity check to my main developer. Grok is my thought partner. No IDE.
Different models message each other directly. The ask travels, the context doesn't have to.
bryandossantos.substack.com/…
anthropic now shows managers what a team produced, what it cost, and where the workflow broke down. my guess at what they get out of it: the deck the buyer brings to budget review. 900% net revenue retention says renewals aren't the problem yet
techcrunch.com/2026/10/28/an…
google's new connected apps platform has developers plug in through a cloud hosted MCP server. that makes it Claude, ChatGPT, and now Gemini and Search. build the MCP server once and it works across all of them
techcrunch.com/2026/10/28/as…
github's octoverse numbers are something. merged PRs tripled in under a year to 220M a month. commits more than tripled to 7.38B a month, most of that in the second half of this year. that's agents showing up in the data
simonwillison.net/2026/Oct/2…
anthropic's gsa deal: 40% off token usage, no minimum. pay per use matches how everyone buys this. openai got 50% off weeks ago, but with a 27 month cutoff. curious what anthropic got back. my guess: volume and a spot on every agency's buying list anthropic.com/news/expanding…
liquid shipped lfm3. six ~2B models on one shared backbone, the text one fits in ~1 GB at 4-bit, and each fine-tunes on a single gpu. small open models you can specialize on your own hardware keep getting more interesting liquid.ai/blog/lfm3
read axiom's writeup on formalizing the classification of finite simple groups in lean. a ~15,000 page proof, now 43.1M lines a computer checks, done in 24 days with swarms of agents. best part: it turned up 34 deviations from the published record axiommath.ai/research/the-en…
cloudflare's writeup on the glm-5.3 kv cache leak is how a postmortem should read. 155 affected responses, the exact root cause down to the sglang admission check, and they say plainly which check they skipped blog.cloudflare.com/kv-cache…
openai's dots does 'proactive research' on a cadence, thinking about what it knows about you and proposing ways to help. good direction. i'd still want that memory to live somewhere i own that every model can read, not inside one app techcrunch.com/2026/10/26/ho…
most interesting thing i read last night: openai says its models proved local smoothing in every dimension, disproved the exponential time hypothesis, and proved yang-mills exists. all three go to outside experts for review first. right call. nitter.cf/gdb/status/21148370594…
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
this should be good. @trq212 spent more than a year on claude code at anthropic, and now he's giving the closing keynote at AIE Code. keynotes get livestreamed free, so i'll be watching. nitter.cf/aiDotEngineer/status/2…
We are pleased to announce that @trq212 will be returning to AIE to share what's now and what's next!
aie.codes will be the last flagship AIE of the year, gathering the top 1000 agentic engineers in the world by invite only.
Apply here: ai.engineer/code/2026/apply
As usual all keynotes livestreamed and talks posted free on our youtube!
huge fan of sol. can't wait to play with terra nitter.cf/OpenAI/status/21147807…
