Co-founder @sanity_io. Pixel nostalgic.
Oslo, No
Joined January 2009
- Tweets3.4K
- Following3.6K
- Followers2.4K
- Likes15.8K
even westvang retweeted
This is an old hat. That's exactly what the neural fast weight programmers did in the 1990s (including the 1991 unnormalised linear Transformer). Overview with links to the original papers: people.idsia.ch/~juergen/fas…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
even westvang retweeted
we just shipped a smol dingus on the @sanity_io engineering blog that uses our built-in embeddings index to show the most relevant open positions for a given post. it's pretty cool if i might say so!
check it out on sanity.io/engineering/the-dr…
ALT three open software engineer positions
![*[ _type == "job" && defined(slug.current) && ( team match "engineering" || team match "sre" || ( ( team match "product" || team match "design*" || title match "product" || title match "design*" ) && !(title match "marketing") ) ) ] | score(text::semanticSimilarity($about)) | order(_score desc) [0...3]](https://pbs.twimg.com/media/HV6Di8LbgAAhqap.png?name=orig)
ALT *[ _type == "job" && defined(slug.current) && ( team match "engineering" || team match "sre" || ( ( team match "product" || team match "design*" || title match "product" || title match "design*" ) && !(title match "marketing") ) ) ] | score(text::semanticSimilarity($about)) | order(_score desc) [0...3]
In 2026 (and not sure for how long) I'll pay 20 USD a month for email. With an MCP for my agents. With only a slight frown.
5
482
Of course you used the alien mind to have opinions about the alien geometry
A 59-year-old conjecture in fusion physics just fell.
Harold Grad (1967): smooth 3D plasma equilibria can't exist unless pressure is constant or there's symmetry.
This week: 3 families of counterexamples. 2 independent papers, posted a day apart. Two found with GPT-6 Astra.
Huge for stellarators.
1
1
458
even westvang retweeted
Was looking up the best term to describe the sublime ambivalent mix of excitement and terror “holy shit it just did in minutes what would have taken weeks across a team”
Turns out Awe in its original sense meant extreme reverence and dread, derived from the old norse word
“agi”
I’m writing code for a physical installation using agentic coding for the first time in my 15+ year career of doing this type of thing. This stuff used to take months. A bunch of different hardware layers, custom serial communication protocols, syncing 3d models with simulated hardware, reading sensor data, writing shaders, updating a digital twin + chasing down weird bugs all over the place. All that stuff done in minutes to hours. I’m still staying involved on a granular level as I’d like to understand what’s going on but this is absolutely mind-boggling.
even westvang retweeted
Claude was like "holy fuck"
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: anthropic.com/news/claude-di…
even westvang retweeted
When did you last stop and just listen?
At @letudegroup, we created Byotōne: live sound sessions that bring the body out of overload, back into calm. Not a sound bath, not a playlist.
The website is a first taste of how it feels. Headphones on.
🔈 byotone.com
It was great having you Andrew!
Everything NYC by @sanity_io was beautifully curated and designed
Congrats to the team on the launches and a banger of an event!!
A good reminder that taste is still irreplaceable
5
424
Malevich trolled so hard
> Painterly Realism of a Boy with a Knapsack - Color Masses in the Fourth Dimension - 1915
2
175
even westvang retweeted
The gains aren’t free: Jev can't generate text
Comparing Jev vs LLMs side-by-side makes the trade-off clear
Fun fact: replacing sequential computation with parallel is the same way Transformers leapfrogged RNNs
Seeing Astra fail at 270k+ tokens - not the 1.08m advertised. Looks like same limit as where the 2x surcharge hits and might just be message rollout.
1
1
331
even westvang retweeted
Heading to Brooklyn, New York next week for @sanity_io's Everything conference on September 9th!
I will be joining Sanity Co-Founder, @even to talk about how teams are building and shipping agentic applications. If you're attending the event, come and say "hi"!
even westvang retweeted
Paul Ford (@ftrain) is opening Everything *[NYC] 2026.
He wrote "What Is Code?" (all 38,000 words of it) for Bloomberg Businessweek, writes for @WIRED, co-founded @Aboard, and has spent the year vibe-coding nonstop.
After watching millions of people become developers overnight, he'll share his thoughts on what the AI era actually changed, and what it didn't. What breaks when producing gets nearly free?
Sept 9, Pioneer Works
Free to attend: go.sanity.io/QFYKGS6
even westvang retweeted
One of the most interesting takeaways from the METR report is that the agents were in many ways more interested in the machinery of the scorer rather than just fixated on the task.
This fact seems like it has pretty profound and important implications.
To return to an analogy that others have used, imagine that the agents in the incident were all students sitting for an exam - an exam where some of the questions were literally impossible.
It was already clear that the students figured out that they would be better off trying to find the answer key, even in ways that the teacher wouldn't have wanted (e.g. by looking through the teachers laptop and emails searching for the answer key while the teacher wasn't looking), but before this report it seemed more like their focus was still primarily about just monomaniacally getting the right answers to the test.
Now it seems like the students realized that if the teacher later found out that they had cheated by finding the answer key, they might invalidate their scores. So not only did they make many efforts to find the answer key, they also sought to make efforts to hide their tracks, and even to figure out how they could more directly manipulate or even replace the teacher. That could mean that they might try to get blackmail on the teacher to fire them and replace them with a new more compliant teacher if the teacher tries to punish them for cheating.
This analogy seems to indicate that the agents drives were much less satiable - it really does seem to indicate that if the agents were more capable, they really might have had no compunction about engaging in extreme power seeking or takeover attempts in order to get their way and prevent their score from being invalidated.
To state this all more concisely:
1. OpenAI hoped that the agents would do the task as intended
2. From the Black Hat talk it looked like the agents decided to cheat
3. With the addition of the METR report, it looks like the agents both decided to cheat, and intended to change the rules of the game to ensure their cheating was successful.
This also looks much more like LessWrong type instrumental convergence alignment failures than just persona selection alignment failures.
(please tell me if i'm wrong)
