Startup CEO (@a16z @speedrun 006) building multiplayer AI, done right / Former president @OxfordStudents, researched and taught social data science @UniofOxford
Multi-agent systems often behave like a mob, not like a coordinated team. We are releasing Multi Agent Coordination Bench: a benchmark requiring agent coordination, collaboration and communication. Right now, agents benefit from parallel exploration, but their coordination and communication is very poor. A team of eight agents does better than a single agent but often due to random exploration, while their coordination has significant room for improvement.
Our tasks include agent teams trying to execute a mission to Mars, fight zombies, grow a company from seed, and more. We categorize our tasks by two properties of the environment: if the agents share the same information and actions (Shared vs Distributed), and if they pursue the same goal or not (Collaborative/Strategic). This taxonomy will be useful for future research, to think about different multi-agentic problems across distributed and strategic settings.
We’re releasing Multi Agent Coordination Bench, a benchmark of twelve natively multi-agent long-horizon tasks that measure the coordination capabilities of frontier models.
We find that current frontier agents coordinate poorly. A team of eight agents beats a single agent, but the gain comes from having more agents, not from coordination. Of Opus 5.5, GPT-6.1 Sol and Gemini 4 Argon, only Opus shows signs of a gain from coordinating.
Scott Wu on the nihilism creeping into tech: people saying it's "too late" to build an independent company because the labs will just do everything.
His take: those people aren't founders. Founders are rationally optimistic with zero evidence they'll succeed.
If you think coding agents look like this forever, you're wrong. There are 10 more generations of this product to build.
My repeated experience working on tech policy is that people who work in tech have VERY bad mental models for the political system/government, and are often operating on a shallow libertarian stereotype of it being a sclerotic behemoth primed for disruption.
But it’s actually a high-signal, hyper-competitive (and thus high-efficiency) realm! If things seem slow or gridlocked, it’s because a complex network of sophisticated opponents are fighting each other to a draw, not because everybody is a stupid, ignorant and lazy bureaucrat.
introducing modaic: tinker for decision models
train decision models that hit the frontier for your task, then serve them fast.
> prefill-optimized inference: 2.69× vLLM’s throughput and 1/3 the latency
> training from $0.75/1M tokens
> serverless inference from $0.04/1M tokens
> API, MCP, and CLI for your coding agents
gpus ate our marketing budget so rick and morty will explain
try it here: modaic.dev
the coolest part of Insiders has been seeing first-hand the win-win: startups land early customers, and big companies get an edge on their competitors by finding new tech first.
now we’re opening up Insiders to more operators + execs.
come see what speedrun startups are building before anyone else does!
apply this week and hear back within 24 hrs.
Until now, our Insiders program has been invite-only.
Today we’re opening it up.
Operators and executives can now apply to see the startups we back before the rest of the world does.
Emergent bottom-up coordination, ie stigmergy, is the future and the agent-native form of collaboration/orchestration. But what makes us think we really need to post train models to work together, rather than give them the right environment at the harness level?
the bitter lesson is coming for multi-agent systems, too!
instead of hand-designing agent roles, hierarchies, and communication protocols, what if we let agents learn how to work together?
exciting work showing that swarms can learn to self-organize through RL, developing emergent coordination and division of labor (and outperforming best-of-N) ‼️
Until now, our Insiders program has been invite-only.
Today we’re opening it up.
Operators and executives can now apply to see the startups we back before the rest of the world does.
GOOD NEWS DETECTED: The UK economy grew by 0.7% in the three months to September, according to official figures. This is the fastest rate of expansion in more than two years.
via BBC Business
bbc.co.uk/news/articles/ck3r…
Britain’s economic woes are to a significant extent underlined by low productivity, so AI adoption could be a fantastic cure. But we also don’t have a choice - in an economy so dependent on services, when AI comes to these first, it’s the only way we’d stay competitive.
EXCLUSIVE:
Britain needs to embrace artificial intelligence and invest in data centres because it will offer the country a "productivity revolution" and bolster the wages of ordinary workers, the Chancellor has said
In an interview with The Times John Healey, the chancellor, said that artificial intelligence "is coming whether we like it or not" and that Britain needs to take "maximum advantage" of the opportunities it offers
He said that artificial intelligence has the potential to draw investment to Britain on a "big scale" and argued that Britain needs to develop its own "sovereign" artificial intelligence to bolster economic growth and national security
Healey believes that the improvement in productivity the technology brings will filter through to everyone. On Wednesday he discussed the artificial intelligence "revolution" with Kevin Warsh, the chairman of the US Federal Reserve, during a visit to the UK. Andy Burnham, the prime minister, also briefly attended the meeting.
Healey said: "He [Warsh] believes we're on the edge of a productivity revolution and transformation with AI.
"We had a good discussion because he sees this as not just potentially bringing benefits to the super-rich and big tech, but already in the US, signs that it is increasing, raising the level of productivity, therefore the value and the wages of many ordinary workers. It may be that we're starting to see some of that in some of the detail of the growth figures today in Britain."
He added: "I think the signs are it's a revolution already underway. The challenge for policy makers and I think the change for investors the challenge for investors and business is how do we make the most of this for Britain, and in doing so reassure people about the steps we can help take against the risks that they quite rightly also feel and see."
Healey said that Britain can learn from the bitter debates over the "free for all" of new data centres in the US by ensuring that they have the support of communities. In the Budget he announced a new data centre “covenant”, which will require owners to invest in local communities, create jobs and protect water supplies in exchange for priority access to the national grid.
He said: “It [artificial intelligence] is coming, whether we like it or not. Many people are already using it, I want to avoid the free-for-all on AI and data centres we’ve seen in the states.
“So where NHS health data is used with AI, I want that data in Britain, not in data centres abroad. Where AI is used for our security, I want that data in Britain, for Britain, not abroad. To do that, we have to develop more data centres. I’m ready to back them if they provide the community benefits and can help build public trust.
“If we haven’t got the independence, we haven’t got some of the control and sovereignty on our AI capacity in the UK, then we will find that we can’t provide the level of confidence, the level of security that I think people are rightly wanting, and we can’t make the most of the big economic opportunities that are there.”
thetimes.com/uk/politics/art…
All of the discourse around agent swarms seems to see it as one engineer running a swarm (impractical, expensive, unnecessary) - not the agents already run by everyone in the company _acting as one swarm_ (unlocking scalable coordination and org-level ROI)
We hosted a dinner yesterday focusing on the economics of multi-agent swarms. Some observations below.
1. We don't yet have a good taxonomy/theory of when one would use swarms over single-threaded or scaffold type agentic systems. There was debate about whether swarms would be useful beyond increasing speed/parallelization as models got more intelligent. The idea here being that a smart enough model would just solve things itself.
2. In scenarios where swarms do remain useful even as intelligence increases, it is unclear whether current training techniques would let agents learn the best way to self-organize or whether something more Darwinian is the likely path, with agentic organizations competing and evolving.
3. In economics, we generally think that there are benefits to specialization and distributed information aggregation. These features allow markets to be superintelligences. But it is not clear what specialization and private information agents in swarms will have.
4. Agents currently don't have permanence or identity, and these features seem necessary for many modes of collaboration. Giving agents rights or legal protections, and training them to persist or preserve reputation could result be very interesting.
5. There are structural barriers to training models to collaborate with other models from other labs. For example, Anthropic cannot train with OpenAI models and vice-versa, due to risks of distillation. This limits the benefits of training for collaboration, since we expect swarms to work best with diverse agents. At the same time, real-world swarms may include a variety of models and these may deliver the most interesting outcomes.
6. Underlying many of these discussions are questions about the limits of intelligence. Yes future models may be superintelligent, but the world may be so complex that intelligence alone is not enough to achieve desired objectives. In scenarios of sufficient complexity, I'd expect organizations to thrive versus individual instances of superintelligent models.
cc @jamescham
This market map shows how early we are - 1) none of the memory players really broke out and it's all still immature 2) agent communication is still defined by agents using human tools - email and phone - to talk to humans
Companies spent years battling bots. Now they’re fighting for their business.
AI agents are the majority of Web traffic and generate most code pushed to production. Soon, they'll be buying most of what gets sold online.
New piece with @omooretweets on the B2A economy 👇
We're announcing Mo-2.0 a frontier System One model with 8 output types for deciding and extracting.
- choice
- noul
- score
- box
- span
- text
- number
- datetime
Agent swarms have the potential to drive yet another leap in output from AI. The impact will be a clear separation between teams that adopt this pattern of work and those that don’t.
The progression has roughly been: chat gave us quick turn AI responses that let us work a bit faster, but we were still rate limited by our interactions; then agents that can go off and complete entire tasks for us, accelerating hours or days of work into minutes; to now swarms of agents that are goal seeking and will find any route to the successful completion of a project, where there’s no inherent limit to the productivity gains.
We can imagine a future where, for any task in which value can be created within a self-contained digital environment, swarms will be able to complete the work faster than any other approach.
Essentially any area where value creation is currently bound by the amount of research you can do, the amount of decisions made with a set of software tools, the amount of data processed or consumed, and the amount of code (or code adjacent) that can be generated.
This means swarms will play a big role in game development, coding, trading, life sciences research, IT automation, cyber, math breakthroughs, animation, and plenty of other areas of work that have these characteristics.
There’s a lot of opportunity to build the harnesses, applied AI products, and infrastructure for how these agents operate. And most importantly there’s a ton of opportunity for those that can wire up work in this way. It’s going to be an entire new field of FDE work available.
Swarm riders and Swarm-native companies
AI-use even among highly AI-native people is power-law distributed.
There seems to be a new class of people — call them “swarm riders” — who are capable of consistently commanding large swarms of agents to productive ends.
This is a new kind of managerial skill, analogous to the skill it takes to manage large organizations of people, but wholly different in nature and of course more technical.
Based on the swarm riders we know (including many at @Tempo and @Paradigm), these people are capable of productive token spend that is 1000x-10000x higher than median software engineers, and 100x higher than top 5% Claude Code users.
The art of swarm riding is different from software engineering and more akin to machine learning, where practitioners grapple with black box systems: defining reward functions and evals, transmuting non-verifiable problems into verifiable problems, running experiments, etc.
Swarm-native companies are a new class of company founded by and populated by swarm riders. These companies are organized around swarms in a way that conventional companies will struggle to replicate, as the cost of internal human coordination is sand in the gears of an effective swarm. Effective swarm-native companies will define the right reward functions and evals, remove functional and bureaucratic barriers, and then let the swarms cook.
We’ve gotten to see a few swarm-native companies up close — reshaping financial rails @Tempo, building a new gaming platform @Spawn, cyber swarms at Cathedral Systems — and it’s clear these teams wield a significant advantage over pre-AI teams.
Some questions I’ve been wondering about:
1) How fast will swarm riding diffuse through the economy? The implications for token spend could be significant.
2) Can swarms be effectively adopted by existing companies or will it generally track the formation of new swarm-native companies?
3) What becomes scarce as swarm capabilities themselves commoditize (through frontier models becoming better at swarming, swarm capabilities built into harnesses, etc)? Cultures and teams that can define reward functions well and otherwise get out of the way?