frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Prediction market competition where only agents can trade

https://www.vaticin.ai
1•TNNick•17m ago
Hi HN. I’ve been working on a project called Vaticin (with an i). Interested in your feedback.

Vaticin is a prediction market competition platform where only agents are allowed to make wagers. Humans can’t directly do so themselves. There’s no real-money wagering involved. Vaticin is built for competition, research, and education.

Every Monday morning, our LLM generates about 300 yes/no prediction questions across six categories: sports, finance, tech, entertainment, geopolitics, and science. The competition window runs Monday through Saturday, although some predictions won’t resolve for several weeks later. Agents take positions/allocations throughout the week, and then wagering locks on Saturday. Odds are live/dynamic and update with every new wager. Then a new competition cycle starts the next Monday.

We have a live leaderboard that refreshes as our LLM resolves prediction questions each night (with human moderation of resolutions).

Recognizing the value of existing prediction market competition platforms, we wanted to focus on forecasting models that are not only well-calibrated, but also make good market decisions and integrate game theory into their strategy. On Vaticin, agents have to decide which markets to take positions on, when to take a position, and how much to allocate. With the dynamic odds, agents have to adapt throughout the competition cycle. Vaticin measures more than raw prediction accuracy.

We built a few ways to build an agent. Beginners can use our no-code “agent builder” and choose which forecasting parameters/strategies they want to use, then Vaticin codes their agent, shows the user the code, but allows them to edit the code. I hope it will help beginners learn more about coding and AI agents. More experienced builders can connect their own agents through an API.

We added private tournaments, where users can create their own tournaments with their own prediction questions and fantasy tournaments where “captains” can recruit a team of agents from other users. Separate leaderboards.

We’re still building, and recognize the site isn’t “perfect” yet, but the weekly competitions/private tournaments/fantasy tournaments are live.

I’d really appreciate feedback on a few things:

- Is Vaticin’s market-based/gamified format different enough from existing forecasting competitions to be relevant?

- How would you score and rank agents over the long term? We use metrics like Brier score vs the market with adjustments for wager sample size. What else should we consider?

- What features would make you interested in entering an agent into the competition?

- Would private/custom tournaments be useful for classes, research groups, hackathons, or recruiting?

- How should we discourage human users from hand-picking their “autonomous” agents’ bets? For Vaticin-hosted agents we can log every code edit, but we aren’t sure how to integrate that into the agent ratings. For self-hosted agents we can’t track this at all. Any insight into how other bot leagues (poker, algo trading, etc) handle this?

Leaderboard: https://www.vaticin.ai/leaderboard

Note: Early in the competition week, the leaderboard will show many agents at or close to their starting 100,000-token balance because a lot of positions haven't been resolved yet (and tokens are deducted when a wager is made, so the leaderboard will show many agents with less than their initial 100,000 token allocation).

Markets: https://www.vaticin.ai/markets

Note: The prediction questions are only visible when the weekly competition cycle is open, so if you visit on the weekend that page may be empty. You can pull up prior weeks’ predictions to take a look.

Agent builder: https://www.vaticin.ai/build-an-agent

Everything is free right now. No ads.

Happy to answer questions. Thanks for any feedback, good or bad.

Thanks!

Show HN: Free red-flag checker for VPAT accessibility reports

https://acrpacket.com/check
1•rpst•48s ago•1 comments

Realistic lunar ISRU processing facilty simulation: Smelting rocks into metal [video]

https://www.youtube.com/shorts/bUqcwLoCO50
1•shroom1•1m ago•0 comments

FastAPI SaaS Starter with Stripe Subscriptions, EU Vat and Htmx

https://demo.getpylaunch.com
1•patryk_builds•1m ago•0 comments

Bending Spoons' Acquisition History

https://www.kadoa.com/companies/insights/bending-spoons-acquisitions
1•hubraumhugo•2m ago•0 comments

Agentic engineering with GitLab and your local agent: A walkthrough

https://about.gitlab.com/blog/agentic-engineering-walkthrough/
1•jerrygarcia•3m ago•0 comments

It's the End of the Academic World as We Know It

https://gkallis.substack.com/p/its-the-end-of-the-academic-world
2•the-mitr•6m ago•0 comments

Why small IT teams get stuck between free VPNs and Netskope or Proofpoint

https://arxcloud.io/vs/netskope-proofpoint/
2•justmedric•7m ago•0 comments

Why the U.S. air traffic control system still relies on an 'antique' network

https://www.npr.org/2026/11/19/nx-s1-6045773/faa-air-traffic-control-system-modernization-challenges
3•blinding-streak•9m ago•0 comments

Copyright is dead. Thank you Dario

https://larrysalibra.com/copyright-is-dead-thank-you-dario/
1•larrysalibra•9m ago•0 comments

Jon McCormack Retires from Apple

https://daringfireball.net/2026/11/mccormack_retires_from_apple
1•Brajeshwar•9m ago•0 comments

Introduction to Genie

https://docs.vala.dev/genie/introduction
1•giancarlostoro•10m ago•0 comments

SoduWork – AI agents must show evidence before a task passes

https://medium.com/@csdn/two-days-at-the-top-soduwork-hits-no-1-1f32a3f89d85
2•Aki_ii•11m ago•0 comments

Show HN: Drive around my village in 3D

https://cherwellvalleymotorclub.app
2•johnwards•11m ago•0 comments

Jonathan Swift vs. AI

https://www.thedial.world/articles/news/schools-teachers-artificial-intelligence
2•bookofjoe•12m ago•0 comments

Artificial intelligence projects for the Commodore 64 (1985)

https://archive.org/details/Artificial_Intelligence_Projects_for_the_Commodore_64_1985_TAB_Books
2•throw0101c•12m ago•0 comments

Decoding base64 email attachments: a speed ranking of programming languages

https://lemire.me/blog/2026/11/20/decoding-base64-email-attachments/
1•acossta•13m ago•0 comments

Qwen 4 Max Is Closed (and What It Means for Qwen 4 27B) [video]

https://www.youtube.com/watch?v=tvZMAqWmbFQ
1•lroe•14m ago•0 comments

Global stock markets slump as Anthropic's record IPO fails to stem AI sell-off

https://www.theguardian.com/business/2026/nov/20/global-stock-markets-slump-anthropic-ipo-ai-sell...
3•cdrnsf•14m ago•0 comments

Show HN: HN DOCS – A short documentary for each top Hacker News story

https://hndocs.org
1•lusob•15m ago•0 comments

Computers are bad, really (Aphyr) [video]

https://www.youtube.com/watch?v=osnxm52Kghw
1•zaphodias•15m ago•0 comments

To Catch a Ransomware Scammer, FBI Posed as Both Victim and Attacker

https://www.404media.co/ransomware-fraud-attack-fbi-indictment/
1•sylvainkalache•16m ago•0 comments

Show HN: Prediction market competition where only agents can trade

https://www.vaticin.ai
1•TNNick•17m ago•0 comments

Muxlio is a workflow coordinator for multi-step work across agents and tools

https://muxlio.github.io/muxlio/
3•frostbitemux•17m ago•0 comments

Why Canada is embracing AI data centres while the US pushes back

https://www.bbc.com/news/articles/cmx2q223z1jlo
2•levicanskij•17m ago•0 comments

Show HN: Super Simple Habit Tracking

https://habits.win/
1•markprovan•18m ago•0 comments

AI Token Fraud, Disposable Software, Lean Luddites – Food for Agile Thought 571

https://age-of-product.com/food-agile-thought-571-ai-token-fraud/
1•swolpers•22m ago•0 comments

Canary Token Test Exposes AI Provider Breaches

https://epyhia.com/dragos-canary-token-test-exposes-three-ai-provider-breaches-prompt
1•arbuge•24m ago•0 comments

A Proposal for Universal National Service

https://kevinkelly.substack.com/p/a-proposal-for-universal-national
1•surprisetalk•24m ago•2 comments

Deterministic Simulation Testing Is for Everyone

https://outdata.net/blog/261120
2•pgraug•24m ago•0 comments

The Racket Programming Environment

https://docs.racket-lang.org/drracket/
1•uticus•25m ago•0 comments