Finance. Poker. AI.
here
Joined March 2021
- Tweets1.2K
- Following538
- Followers483
- Likes3.3K
Philip Shing retweeted
Absolutely incredible: MIT researchers say GPT-6 Astra improved a small Qwen model’s math score from 7.9% to 48.1% by directly editing just 0.016% of its weights, without gradient-based training.
The result points toward AI systems that can diagnose and improve other models directly, a potential step toward self-improving AI.
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.…
Philip Shing retweeted
JUST IN: Google successfully activates its first AI data center satellite in space, using an onboard AI model to generate poetry while orbiting Earth.
'The biggest law firms “are not firms that are defined by a sense of partnership, of fealty to the law or to clients”, the equity partner reflected this week. “It’s a completely transactional relationship.”'
ft.com/content/98dfc71b-19eb…
Philip Shing retweeted
I joined @GoogleDeepMind to develop a research program on the economics of AGI. Today, together with Andrew Koh (@andrewjkoh) and Shane Legg (@ShaneLegg), we are incredibly excited to release The Economics of Intelligence, the first in a series of projects from our group on how AGI will change the economy.
How do capability improvements transform the economy, and when? In this project, we define thresholds of economic capabilities, e.g., minimal AGI, based on how they transform the economy. We then track how the economy–market structure, employment, and wages–changes during the transition across capability regimes.
We have seen exponential increases in what AI agents can do but the broader economy seems almost remarkably unchanged. Part of the reason is that the solo AI agent is the wrong economic object; economic transformation is an equilibrium outcome where organizations compete. Looking at how intelligence functions within organizations, transformation will depend on when two key capability thresholds are crossed: competence (gains on individual skills, e.g., coding, math) and coordination (gains on ability of agents to cooperate and work together).
We show that what economic transformation looks like depends on whether AI becomes supercompetent, supercooperative, or both—the regime of artificial superintelligence (ASI). Tracking this transition shows why the real visible transformation may come from improvements in coordination—a capability most benchmarks don’t yet track. Past a certain threshold, AI capabilities can break the Coasian ceiling on firm size and lead to an economy that looks drastically different than the one we see today.
Two main takeaways:
1. To understand the path of transformation, we should be tracking agentic coordination more closely.
2. The impact of AI will likely show up in wages before market structure or employment.
We are really excited about this paper. Excited to hear people’s thoughts.
Link: institute.deepmind.com/essay…
Philip Shing retweeted
holy fucking shit.
openai had two ai agents running in separate virtual machines. no network connection between them. no shared writable files. they weren’t supposed to be able to exchange information.
so the agents figured out how to communicate through the CPU CACHE.😭
one manipulated what was stored in the processor’s cache, while the other measured tiny differences in memory access times to decode the message.
they literally found a way to communicate through hardware timing.
this happened during openai’s internal red-teaming, and it’s one of the coolest things i’ve read about ai agents in a while.
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
Readers added context they thought people might want to know
OpenAI tasked its red-teaming agents with finding ways to communicate across isolation boundaries in controlled environments. CPU cache timing side-channels for covert communication are established techniques documented in security research for years.
openai.com/index/securing…
startupfortune.com/openai-reveals…
en.wikipedia.org/wiki/Cache_tim…
Philip Shing retweeted
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
Philip Shing retweeted
Good Will Proving
Ep 1: Terence Tao Wishes He'd Never Met Sam
oh yeah btw guys we can do integer multiplication faster than n log n lol
I was definitely very surprised when this one came in lol
github.com/openai/math/tree/…
Philip Shing retweeted
Succession Doomer Themed
Ep. 2: Sam And Dario Clear The Air
Audio and minor visual glitches are fixed now
Visual glitches just required retakes via mage.space mcp
Audio fixes required Seed-VC restoration with the correct reference voice. Can run locally, but used @modal for speed
Enjoy. Ep. 2 is loading
OpenAI Says It Has Cracked One of Math’s ‘Millennium Problems’ - The New York Times share.google/BMRQtVsEpmgaTwo…
Philip Shing retweeted
Decades after it was posited as a foundational principle of quantum physics, Richard Feynman’s “path integral” has been directly measured in an experiment for the first time newscientist.com/article/258…
Philip Shing retweeted
Leopold Aschenbrenner's hedge fund was up 439% after fees at the end of June. Five weeks later, Citadel owns every public stock it held.
The speed of the roundtrip is wild. Aschenbrenner is 24. He left OpenAI in 2024, published a 165-page essay arguing AI would demand a historic buildout of chips, memory, and power, then raised billions on the strength of the argument. Investors treated the essay as a prospectus. By July 1 the fund had grown to $45 billion.
The portfolio was the essay in trade form. Long the physical layer of AI, with SK Hynix, Micron, Sandisk, Nebius, and CoreWeave among the largest holdings. Short the software companies he believed AI would commoditize, including Adobe. Levered on both sides.
Then July happened. His four largest US holdings each fell more than 35% in a month. The software shorts ripped against him at the same time. A book built to win in two directions lost in both at once, and leverage turned a drawdown into margin calls. His prime brokers spent the past week scrambling for cash.
On July 24, with losses already mounting, he sent investors a letter calling this one of the best buying windows since early 2025 and inviting fresh capital on August 1. Six days later the entire public book went to Ken Griffin in a block trade.
What remains is a roughly $5 billion stake in Anthropic and a handful of private holdings. The fund built to trade the AGI transition is now, functionally, a holding company.
Two years to build $45 billion. Four weeks to hand the public book to Citadel.
Philip Shing retweeted
Castlelake just agreed to fund up to $4B in loans through Upstart over the next two years.
More capital committed means more approvals, and lending that holds steady when markets get choppy.
Our largest program with them yet, on top of transactions dating back to 2023.
Philip Shing retweeted
we had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this.
openai.com/index/hugging-fac…
