Nenad Tomasev

Take a sneak peak at The Paradox of the Organism at Google books books.google.com/books?hl=sv&...

Don’t forget to pre-order!

None of this is guaranteed to work well, or work at all, of course. It is up to us to decide if such AI agent markets ought to be created, how they should be safeguarded and regulated, and how to ensure that everyone can benefit from the economic value created by future advanced AI agents.

While it is tempting to think about major societal challenges like sustainability, climate change, or human well-being more broadly, it may well be more feasible to realize these concepts at the community scale.

If we intentionally design these virtual markets to support persistent agent identities, verifiable transactions and action outcomes, reputation, smart contracts between agents, and currencies and mechanisms that can capture positive social outcomes, this may open new possibilities.

Counter-intuitively, it may well be easier to coordinate AI agents at scale, than it is to coordinate people.

This is by no means a new idea - similar proposals have been directed at the human economy in the past. Yet, it is interesting to reassess its utility in the context of advanced AI agents in particular.

Another concept that we explore is that of Mission Economies realized through steerable AI agent markets.

This is contingent on the underlying infrastructure being in place - fairly distributed bidding budgets, safety and governance mechanisms, verifiable and auditable transactions, oversight, and others.

Markets offer unique opportunities for steering these interactions towards favourable outcomes - and hopefully future research can make some of these conceptual ideas more concrete.

The reason why this may be of interest - is that we need to move beyond aligning single agents to instructions in isolation, and recognize the complex arising web of interactions in multi-agent systems at scale.

AI agents could thereby engage in auctions towards shared but otherwise limited resources, and the market mechanisms could help establish a fair allocation of goods.

This would be one of the emerging possibilities within Virtual Agent Economies imagined as steerable agent markets, somewhat analogous to High-Frequency Trading in the existing human economy.

One of the concepts that we discuss is that of a hypothetical High-Frequency Negotiation (HFN) framework between Personal AI Assistants, on behalf of their users, and in line with individual preferences.

Finally, we outline the technical and governance infrastructure—such as verifiable credentials for establishing trust—required to safely and robustly scale agentic AI deployments. These are necessary to address systemic market risks, and prevent exacerbating inequalities.

We can therefore think about orchestrating AI agent interactions towards achieving major societal or community goals, or better aligning with user preferences. Market-based mechanisms like auctions may also be employed for fair resource allocation when accessing shared goods.

The arising virtual (sandbox) AI agent economy may offer us new opportunities for insulation and safeguarding, as well as establishing potentially unprecedented coordination between AI agents.

The rapid adoption of AI agents points to a future where AI agents may be able to produce economic value independently of human labor. Coupled with the development of new interoperability standards like the A2A and MCP, this signals the inevitable emergence of a new economic layer.

Virtual Agent Economies The rapid adoption of autonomous AI agents is giving rise to a new economic layer where agents transact and coordinate at scales and speeds beyond direct human oversight. We propose the "sandbox econo...

Happy to share a new preprint:

Virtual Agent Economies
arxiv.org/abs/2509.10147

where we discuss a number of possible frameworks for establishing steerable agent markets.

For those with interest in mental health and AI, and in particular on how potentially sensitive data gets collected and used there - there is now a Delphi survey that you can fill out to inform how this gets done, as a part of PARQAIR-MH project (www.parqair.org/home), via:
redcap.link/2l91l043

Our work on concept discovery towards bridging the human-AI knowledge gap in AlphaZero has now been published in PNAS.

As future AI systems become even more capable, we should be thinking of ways of utilizing them not only to perform tasks, but also to further our own knowledge and understanding.

This was cool: "Gemini 2.5, create a sim that is cross between Jujujajaki networks and cellular automata"

Gemini: "What's a Jujujajaki network?" I paste in a paper.

Gemini: "Got it, a dynamic network with local search & exploration." Worked in one shot.

Me: "Make it nicer"

Some cleverness here

Self-organizing systems: what, how, and why? - npj Complexity npj Complexity - Self-organizing systems: what, how, and why?

An interesting read on self-organizing systems, and how order and disorder may manifest differently at different scales: www.nature.com/articles/s44...

Dataset Distillation (2018/2020)

They show that it is possible to compress 60,000 MNIST training images into just 10 synthetic distilled images (one per class) and achieve close to original performance with only a few gradient descent steps, given a fixed network initialization.

Please feel free to reach out to me in case you're interested, and I can forward the applications along to our recruiting team for further evaluation.

Ideally, the candidates would have already applied through the SR posting - but if you are not already in the pipeline, this can be accommodated.

I can say from my own experience that it is a highly motivating and inclusive environment, and would therefore recommend it as an excellent opportunity for pursuing ambitious open-ended research on agentic technology with foundation models.

The Agent Frontiers team leans into a diverse set of research backgrounds, and brings them together towards the common goal.

This is an on-site position at our London office, for a projected 24-week duration.

We are looking for strong candidates currently pursuing their PhD, to work on an exciting project looking into adaptive planning and reasoning under inference-time constraints - to help us develop agents (and evaluations) with strong reasoning capabilities.

I'm happy to advertise an upcoming Student Researcher position on my Agent Frontiers team here at the Google DeepMind Foundational Research Unit, aimed for a start date early in the summer (currently listed as late June, but obviously somewhat flexible).

Constrained roads to complex brains Neural development and brain circuit evolution converged in birds and mammals

Constrained roads to complex brains -
Neural development and brain circuit evolution converged in birds and mammals www.science.org/doi/10.1126/...