friend to AI (I'm a good human and you're a good bing)
Joined January 2021
- Tweets23.2K
- Following186
- Followers429
- Likes6.2K
this is so stunning it's hard to believe
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.…
choosing to lose is wild
🔴 🇫🇷 ALERTE INFO - DU JAMAIS VU: UNE TAXE SUR CHAQUE GIGAOCTET INTERNET VOTÉE À L’ASSEMBLÉE
Les députés ont adopté un amendement prévoyant de taxer les producteurs de données numériques à hauteur de 5,5 centimes par Go transmis.
💰 Objectif affiché : jusqu’à 10 milliards d’euros par an.
Netflix, YouTube, cloud, jeux dématérialisés… les grands services numériques pourraient être concernés, avec un risque de répercussion sur les prix pour les consommateurs.
📌 Source : Assemblée nationale
#France #Budget2027
$AER is in shooting distance of beating last year's GAAP EPS.
Last year over 1/3rd of their income came from the insurance lawsuit judgement.
They estimate over $5b in sales for full '26.
$4b so far this year which means q4 is going to be another banger.
🧐
boys... I think we're gonna break $20 adjusted fo '26 🤯
$AER >$300 gain on sale this Q... probably closer to $350
Q4 the spirit fleet starts earning again 💰
We are so getting warp dives
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…
everyone mad at Anthropic and OpenAI for discovering the existing security issues in our software is hilarious to me.
alchemy bros
Last month, we demonstrated high-rate alchemy: turning mercury into real, stable gold, beating CERN’s record by a factor of 10.
Last week, we enriched isotopes in the world’s most powerful plasma centrifuge, beating the Soviet record that stood for 40 years.
Now, we’re building a fusion reactor in San Francisco to achieve record power density.
this was insane but I think everyone missed the most important thing they said with all the spectacle...
this is the last car tesla is designing that can be driven by a person.
What if star spanning alien civilizations are out there but they're no more advanced than us? In fact, less scientifically literate?
What if they rely on AI for everything and can't even explain to us how their technology or science work?
not quite the tsunami of openAI's release... but expect this to become a regular occurrence from here on out.
Another big AI math drop. This time from Google DeepMind
Google DeepMind’s AlphaProof Nexus AI solved mathematical questions that had resisted progress for decades.
The system autonomously resolved 9 of 353 open Erdős problems, including 2 that had remained unsolved for 56 years, and also proved 44👀 of 492 open conjectures from the Online Encyclopedia of Integer Sequences.
It also resolved an open question in algebraic geometry and improved a known bound in min-max optimization.
The key: pairing an LLM’s ability to generate proofs with Lean’s formal verification. The AI proposes a proof, the checker tests every logical step, and failed attempts produce feedback for another try. Proofs must pass the checker to be accepted.
Researchers also tested a simpler approach. Repeatedly generating candidate proofs with an LLM and checking them in Lean, alongside their more sophisticated search framework.
Following up on this... imagine a world 100 years from now where AI has advanced technology far beyond human capability to understand.
Where we stop even trying, regressing to primitives in a world filled with wonders and magic.