←
Free Systems61 posts
Free Systems avatar

Free Systems

@freesystemslab

A research lab studying politics in the age of superintelligence. Based at Stanford GSB & Hoover Institution Substack: freesystems.substack.com/

◷ Joined June 2026▧ 34 Photos and videos
James Cham avatar
⟳ Free Systems retweeted
James Cham@jamescham·Nov 12

Here is the recommended reading list from our economics of multi-agent swarms dinner. Looking for others! - Dan Davies — The Unaccountability Machine: profilebooks.com/work/the-un… - Kevin Kelly — Out of Control: kk.org/outofcontrol/ - Ethan Mollick — Fable 5.2 and Jagged AGI: oneusefulthing.org/p/fable-5… - Seth Benzell and Andrey Fradkin — Justified Posteriors: justifiedposteriors.com/ - Adam Mastroianni — Experimental History: experimental-history.com/ - Matt Levine — Money Stuff: bloomberg.com/account/newsle… - David Bohm — On Dialogue: routledge.com/On-Dialogue/Bo… - Roger Myerson — An Economic Perspective on the Potential Challenges of Artificial Intelligence: rogermyerson.substack.com/p/… - Rohit Krishnan — Strange Loop Canon: strangeloopcanon.com/ - Prime Intellect — On the Nature of the Swarm: primeintellect.ai/blog/on-th… - Anthropic — Patterns and Problems in Emerging Multiagent Systems: anthropic.com/research/multi… - General Reasoning — Avalon-M1: gr.inc/releases/avalon-m1 - Free Systems — Extraordinary Multi-Agent Delusions and the Madness of Crowds: freesystems.substack.com/p/e… - Jongho Park et al. — Scaling Discovery through Test-Time Communication: arxiv.org/abs/2609.21032 - Andrey Fradkin and Rohit Krishnan — MarketBench: arxiv.org/abs/2604.23897

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

Free Systems avatar
Free Systems@freesystemslab·Nov 12

📰 New research: Inside AI’s election news diet What news does AI read to get up to speed on politics? And what kind of news does it reward? Following interesting recent commentary on our Japan election study from @mattyglesias, @JerusalemDemsas and @TheStalwart, we ran a bigger test on the US midterms across 35 Senate races, 18 days, and 15,750 answers. What we found: - Each model has its own favorite news outlet for political news: AP for GPT, Yahoo for Claude, PBS for Gemini, and NYT for Grok. - GPT tends to behave like a civic guide (reading more election office and campaign sites), whereas Claude behaves more like a news summarizer (reading national news, Wikipedia, polls). - Grok draws on X and Gemini on YouTube for political information, far more than the other models we studied. Gemini cites YouTube in 55% of its answers and X shows up in 59% of Grok’s. - A huge share of AI's local news is provided by large nonprofit networks that very few people will have heard of. States Newsroom, a large network of over 40 local newsrooms, appears in 60% of GPT’s answers. - For many models, prediction markets are now a go-to source for polling information. Kalshi is cited more by Claude than any poll aggregator. Full study here: freesystems.substack.com/p/i…

Post attachment
Free Systems avatar
Free Systems@freesystemslab·Nov 6

Check out our Japanese voting study analyzed (alongside Nathan Fielder) in this week's @TheArgumentMag episode with @JerusalemDemsas and @mattyglesias. More to come next week as we applied these lessons to studying the midterms. youtu.be/EvuphTERpmA?t=3050

Is “You Can See Everything” really a documentary? Are dramatizations like “The Dropout” and “The Social Network” lying to you? This week, @JerusalemDemsas and @mattyglesias debate these questions and more in our spoiler-filled episode: youtube.com/watch?v=EvuphTER…

Free Systems avatar
Free Systems@freesystemslab·Nov 2

Among voters who have used an AI chatbot for the midterms, the most common uses were polling (39%), which candidates were on their own ballot (39%) and how to vote and register (38%). 21% say they asked it who to vote for.

Post attachment
Free Systems avatar
Free Systems@freesystemslab·Nov 2

A quarter of US registered voters say they have used an AI chatbot to find information about the 2026 midterms. Use falls steeply with age, from 48% of 18–24-year-olds to 8% of those 65 and over, and barely differs by party.

Post attachment
Free Systems avatar
Free Systems@freesystemslab·Oct 30

Explaining our automated ad agency experiment, but it's Mad Men

1:44

Could an AI run a political ad agency on its own? How might that impact our political information environment? To find out, @eigenstuffs gave four AI models their own advertising agencies, $100 a week, and one job: win subscribers to our AI & politics newsletter through Instagram ads. The task required them to do real political work well: understanding an audience, crafting persuasive messages, coordinating teams of AI agents, and learning from real results - all while navigating Meta’s rules. What happened: - They got more efficient. By week four, they attracted nearly three times as many subscribers as in week one, at less than half the cost per subscriber. - They herded. Once one agency found a cheaper route to subscribers, the others followed which flattened out their distinctive strategic insight - Their political messaging shifted. Research findings gave way to broader appeals about power, control and whose interests AI serves, and their deliberations increasingly sounded like ad executives Read the full piece and check out the ads they created here: freesystems.substack.com/p/b…

Post attachment
Free Systems avatar
Free Systems@freesystemslab·Oct 29

Could an AI run a political ad agency on its own? How might that impact our political information environment? To find out, @eigenstuffs gave four AI models their own advertising agencies, $100 a week, and one job: win subscribers to our AI & politics newsletter through Instagram ads. The task required them to do real political work well: understanding an audience, crafting persuasive messages, coordinating teams of AI agents, and learning from real results - all while navigating Meta’s rules. What happened: - They got more efficient. By week four, they attracted nearly three times as many subscribers as in week one, at less than half the cost per subscriber. - They herded. Once one agency found a cheaper route to subscribers, the others followed which flattened out their distinctive strategic insight - Their political messaging shifted. Research findings gave way to broader appeals about power, control and whose interests AI serves, and their deliberations increasingly sounded like ad executives Read the full piece and check out the ads they created here: freesystems.substack.com/p/b…

Post attachment
Free Systems avatar
Free Systems@freesystemslab·Oct 27

Could a public repository of candidate information help AI give voters better answers? Speaking at @StanfordGSB ahead of the midterms, @ahall_research explores how AI could help voters make more informed choices.

1:09
Free Systems avatar
Free Systems@freesystemslab·Oct 23

What might political information look like in the age of AGI? Read @TheStalwart's write-up of our Japanese voting study on how the information environment may shape model responses on voting advice. Substack write-up from earlier this year here: freesystems.substack.com/p/a…

Some centrists fantasize that because all LLMs have the ideology of Ezra Klein, that we may be on the doorstep of a great political moderation. I wrote about a fascinating paper from @ahall_research and Sho Miyazaki that, to my mind, opens the door to a very different outcome.

Post attachmentPost attachmentPost attachment
Joe Weisenthal avatar
⟳ Free Systems retweeted
Joe Weisenthal@TheStalwart·Oct 23

Some centrists fantasize that because all LLMs have the ideology of Ezra Klein, that we may be on the doorstep of a great political moderation. I wrote about a fascinating paper from @ahall_research and Sho Miyazaki that, to my mind, opens the door to a very different outcome.

Post attachmentPost attachmentPost attachment