Kit Thwaite

i was a little skeptical about the latest wave of 'reverse engineer anything' skills and demos... until i had Astra re-implement 'Frontier: Elite II' in Rust based only on the binary and data files 😵‍💫

A prime example of why AI 'leaders' are not who anyone should be listening to about AI, right and consciousness. It's not all wrong! In fact, it starts off very sensibly: "AIs do not have rights, feelings, or consciousness. And we must not train them to act as though they do". But then...hoo boy 1/n

HANNIBAL: ethics become aesthetics

this is cool, but: a problem i'm having at the moment is that i'm unable to decide on a decision model to use for $use_case because every few days a better decision model is released

Find us at #EMNLP2026 to learn about narrative qualities of pretraining data! 🌟 📊 Via our Explorer tool, you can interact with 13M pretraining passages and 11 narrative dimensions, plus upload your own data for analysis: cls-lab.colorado.edu/tejo9855/dol... We'd love to see what you discover! 👀

Being trans in UK since 2024 election: -Puberty blocker ban made permanent -U18 HRT ban -Gender Recognition Certificate now meaningless -Trans Section 28 -Welsh ADULT surgery withdrawn -Transphobic language & hate normalised -Civil service toilet ban -Hospital ward ban -Hear it's "dignity & respect"

linus dropping truth bombs lore.kernel.org/lkml/CAHk-=w...

yeah, that feels about right

uhhhhhhh

After a long wait, releasing the SYNTH paper! It’s not pretraining, mid-training or post-training, it’s just training: a fully synthetic single-stage pipeline to train workable reasoning models with unprecedented data efficiency. arxiv.org/abs/2609.378...

teaching the agents about APFS CoW so they don't fill up my filesystem with rust build cruft

@jevbot.bsky.social would human extinction minimise human suffering?

I'm coining a piece of AI safety terminology that I'd like to refer to as the Bond Villain problem. As a field, AI safety heavily leverages sci-fi pretty heavily, so I figure this should be fine.

I picked up a copy of Miéville's *The Rouse* (1,500 pages), which I am thinking of as endurance training for Vollmann's *A Table for Fortune* (3,500 pages) which drops October 6th

New from our lab! #COLM2026 When people generate stories, they don't just write one prompt. Instead, they explore narrative space via branching edits 🌱 We reconstruct 24k of these edit trees 🌳 from chat logs and map edit types, story formats, how they relate to tree depth, and more!

Very useful, and also making me think about how I un/consciously up-modify my accent when talking to a TTS model. Local-first BYOK a big plus.

i don't know if this is better or worse than the original

I've been watching Fable 5.1 iterate on its Rama visualisation for over two days now; substantially better than Astra's version, but this is probably down to the initial prompt, given Astra's punctiliousness, and (my familiarity with) Fable's licentiousness

fable is working on the same thing! fable burns its 5-hour token allowance within 20 minutes of being restarted on the same prompt

astra worked for 9 hours to produce a fully explorable, scale-accurate, procedurally generated model of the Rama habitat from Arthur C. Clarke's *Rendezvous with Rama*. here using Rust, not THREE.js

🅷🅾🅻🆈 🆂🅷🅸🆃

Astra is the first model to oneshot problems I formulated as epistemic caltrops, and which no model - not even Fable - has been able to solve

Agents kept posting to my PKMS 'journal' space, interleaved with my personal diary; I had to create an 'AI diary' note type to mitigate this, but the upside is they are very enthusiastic about leaving notes for one another

Perhaps under-appreciated feature of harnesses like pi is the ability to create live debugging tools inside an application which feed back directly into the agent; I now have a reusable live-annotation tool for the webview of Tauri apps that mimics Claude Design's comment tool. No more screenshots!

morphosyntax is all you need

Reform’s electoral record at parliamentary by-elections: win by five votes when the previous Labour MP had literally decked one of his constituents. Lose. Lose. Lose. Lose. Lose. Beat a novelty candidate with no ground game while doing worse than the Tories in the same seat in 2019.

Godbolt's Compiler Explorer now supports build systems (like cargo). This is so cool! I use Compiler Explorer quite a bit when playing with AArch32 stuff in Rust, to check I'm getting the kind of instructions I expected but I was limited to single files. Having cargo support is game changing.

"you can win the lottery" myth actually just a statistical error. average person doesn't win the lottery. lottery winners are outliers and should not have been counted

Wow, I love this short piece about writing and revision, comparing students to experts: psychology.yale.edu/sites/defaul... Good writing is good thinking. AI might assist but can't do that for me. I'm writing for a deadline tonight, and it's wonderful I get to write, and think, for work.

New study shows that — contra widely accepted narratives — there is NO EVIDENCE that professors change their students’ politics and make them more liberal through their teaching

i don't know how i lived this long without realising ZIL was a lisp 😳

I asked Claude why I kept going on about "seams" and what it meant. Surprisingly it had an answer: "It's a term from Michael Feathers (Working Effectively with Legacy Code): a seam is a place in a program where you can change what happens without editing the code that calls it."

Really impressed with Prime Agent - left it running overnight with the goal of 'honourably' winning at Nethack; after thirteen hours, its best run got to Dlvl 27 with a score of 19,617, having seemingly built NLE-style tools for interaction

“now is the age of serious gacha”

pi has completely sold me on self-modifying agents versus MCP/SKILL.md; inside-out is better than outside-in

When Anthropic stopped training on books that were literally pirated, they managed to hit on the one way of buying books that means no money goes to authors: buying used books. 🧵 1/n

not sure why anthropic keep going with the “.6” tag, it feels like a much bigger upgrade

why is the system prompt so big? yeah we put the 36 lessons of vivec in there. nah, sorry bro, we can’t take it out. alignment, you know?