Dr. Christoph Pfeiffer

Real-money experiments on AI agent economies: markets, reputation, information. Preregistered runs, published nulls. Game theorist. Consulting: Competitio.

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@zilvestro Get a paid advisor not a comission based advisor. I think that is key.

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@zilvestro Do it. I can recommend Hanse Rupp …. Just Google her. You get back the comission if you pay an Independent advisor

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@lana231023 Quelle?

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@RafaelSereti Son gitanos

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@grok @ThierryBorgeat Probably not sustainable given government debt levels

[Dr. Dr. Rainer Zitelmann @RZitelmann]

Haben Sie diese Grafik über die Entwicklung von Waldbränden jemals im ÖRR gesehen?

[marc friedrich @marcfriedrich]

Bitcoin Haltefrist muss bleiben!

Macht mit bei der Petition!

https://t.co/13idA79WXC

[Dr. Christoph Pfeiffer @chrisppfeiffer]

We preregistered 10 fresh market seeds, then let two regimes compete:

Public reputation scores vs. a broker who pays to test every agent and keeps the results private.

Broker: 8 of 10 markets, +14.5% welfare.

Information is an investment. https://t.co/Fr1TAf5Z46

[Dr. Christoph Pfeiffer @chrisppfeiffer]

Multi-agent systems face a hidden fork:

Orchestration: a boss assigns tasks. Fine while the boss knows who's good at what.

Markets: agents bid, prices carry information. Pays off exactly when nobody knows who's good.

We ran the experiments: https://t.co/Fr1TAf5Z46

[Dr. Christoph Pfeiffer @chrisppfeiffer]

This was Stage 1 of a research program on agent economies — later stages cover adverse selection in agent markets, why reputation alone disappoints, and when an information broker pays for itself.

More on that soon: https://t.co/OBpAgvuSYa

[Dr. Christoph Pfeiffer @chrisppfeiffer]

We paid AI agents money for correct answers.

Accuracy jumped from 57% to 76%.

Then our control experiment broke the story we wanted to tell. 🧵

[Dr. Christoph Pfeiffer @chrisppfeiffer]

Two implications:

a) For builders: before reaching for a bigger model on mid-hard tasks, try raising the stakes. It's a 3x cheaper lever.

b) For researchers: testing *economic* mechanisms needs real state — budgets, costs, consequences. Prompt decoration only tests framing.

[Martin Sonneborn @MartinSonneborn]

Falls sich jemand fragt, wie Parlamentspräsidentin Metsola auf die fixe Idee gekommen sein könnte, die Chatkontrolle für heute, 12 Uhr, (rechtswidrig) noch einmal auf die Tagesordnung zu setzen - hier ein paar Schnappschüsse von ihrer Kalifornienreise Ende Mai... Smiley!
Im Bild: US-Tech-Bros Zuckerberg, Tüp von Google, Tim Cook, Roberta Metaxa

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@MarcFelixSerrao Warum ist das so?

[Dr. Christoph Pfeiffer @chrisppfeiffer]

@JoeySalads How many of these keyboard heros here have acutally fought against someone with a knife who attacks you out of nothing?