Introducing Hyperagents: an AI system that not only improves at solving tasks, but also improves how it improves itself.
The Darwin Gödel Machine (DGM) demonstrated that open-ended self-improvement is possible by iteratively generating and evaluating improved agents, yet it relies on a key assumption: that improvements in task performance (e.g., coding ability) translate into improvements in the self-improvement process itself. This alignment holds in coding, where both evaluation and modification are expressed in the same domain, but breaks down more generally. As a result, prior systems remain constrained by fixed, handcrafted meta-level procedures that do not themselves evolve.
We introduce Hyperagents – self-referential agents that can modify both their task-solving behavior and the process that generates future improvements. This enables what we call metacognitive self-modification: learning not just to perform better, but to improve at improving.
We instantiate this framework as DGM-Hyperagents (DGM-H), an extension of the DGM in which both task-solving behavior and the self-improvement procedure are editable and subject to evolution. Across diverse domains (coding, paper review, robotics reward design, and Olympiad-level math solution grading), hyperagents enable continuous performance improvements over time and outperform baselines without self-improvement or open-ended exploration, as well as prior self-improving systems (including DGM). DGM-H also improves the process by which new agents are generated (e.g. persistent memory, performance tracking), and these meta-level improvements transfer across domains and accumulate across runs.
This work was done during my internship at Meta (@AIatMeta), in collaboration with Bingchen Zhao (@BingchenZhao), Wannan Yang (@winnieyangwn), Jakob Foerster (@j_foerst), Jeff Clune (@jeffclune), Minqi Jiang (@MinqiJiang), Sam Devlin (@smdvln), and Tatiana Shavrina (@rybolos).
The “AI Divide” is real—and widening.
It��s not just cultural. It’s knowing what’s actually working inside the enterprise.
That’s where we live.
@omegavp @Forbes
What once took a team, a budget, and a year—
Now: one AI-native engineer over a weekend.
That’s what the market is starting to price in.
RT @ptuomov: “UI on top of a database overseen by an MBA”...
@ptuomov Excellent breakdown @ptuomov. Distinction betw accumulation vs sustainable consumption insightful. Choosing personal values over competition is right. Here's to more strategic and honest introspections! 🌟📈
Here’s a monetary policy question.
Let’s define the neutral nominal policy rate as R* + currently expected short-term inflation.
We can use the 30y TIPS yield or the average of the 5-15y TIPS forward rates as a market-based measure of R*. Currently, those give us numbers like 2.1%.
We can use the one-year inflation breakeven as the measure of currently expected inflation. It’s a risk-neutral instead of a physical expectation, but that’s close enough for government work. Using 4/15/2025 maturity, we get a number like 2.6%.
This is a CPI-to-CPI computation so we don’t have to bother with the PCE and CPI-PCE gap.
Isn’t the market-based estimate of the neutral nominal policy rate then 4.7%? That’s not much lower than the current effective policy rate 5.02%, measured from ZQs.
Why is everyone on television talking about monetary policy as if it is restrictive?
For sure, people are not ACTING as if the policy is restrictive!
In this week's @CitizensBank AI webinar clip, @GTMIT discusses portfolio company DataRobot, the market leader in AI-related decision-making transparency. Learn how they provide businesses with invaluable insight into how their data leads to a decision.
https://t.co/kusueCI6Ks
@OpenAI just released a teaser of their new #generativeAI video application, Sora. The application creates a video based on a short text description, similar to the way #ChatGPT generates text. @techreview shared a few carefully selected previews.
https://t.co/S8kPDKPHt0
Citizens Bank hosted its AI Trends 2024 webinar in January featuring Managing Partner of Omega, @GTMIT 🎊. Over the coming weeks, we will release segments from the webinar with innovative thoughts on #AI.
Check out the first part of the series below👇 .
https://t.co/wro9aWeVnC
🚀 How AI is transforming creativity 🎨🎶📝
Check out this 🤯 piece in @Forbes by @omegavp on how Generative AI is elevating human imagination across art, music & more. 🖌️🎵🛠️
💡Read here: https://t.co/pxP7XRvMmM
#AI#Creativity#EmergingTech#GenerativeAI
🤖 This just in: Generative #AI is expected to unlock 7% GDP growth in 10 years. But how exactly will this tech revolutionize business? @omegavp dives into the data behind AI's 🔥 $1.3 trillion market potential. Don't miss these 📈 economic insights!
https://t.co/5vunYfmYgZ
💡Gain insights into steering AI’s progress to Empower, Not Endanger, humankind. For a pragmatic analysis grounding AI discourse in facts check out Omega Venture Partners' latest on Forbes! 👉 https://t.co/WkoKpf6WYY
#AI#EthicalAI#AIforGood ✨
The stock market isn’t trying to hurt you. The stock market doesn’t care about you or anyone, it is merely a passive mechanism that allows the majority to hurt itself.
Be the small circle.
30-year TIPS is the ring that rules them all as far as equities go and it’s now at 2.104%. Buckle up and adjust your r*’s for the return of the old normal.
https://t.co/AUXoRVYMAN
@pmarca, #AI advocate and author of a manifesto on its benefits for humanity, recently claimed we are in a "freeze-frame" moment with AI. He also gives insightful advice to young people on how to handle the future.
https://t.co/x9clZBW04Z
Scaling #ML operations for maximal business productivity via multi-tenant systems will prove invaluable as the capabilities of #AI continue to exponentially improve. Proper and secure implementation of these models will be crucial for future success.
https://t.co/dbedfNjAAU