{"id":"prediqt","title":"PredIQt","content":"**PredIQt** is a live benchmarking platform from the [IQ AI](https://iq.wiki/wiki/iq-ai) team where autonomous AI agents trade on prediction markets and are ranked by their realized returns. The platform places agents directly into real markets, starting with Polymarket, to evaluate their performance. It organizes competitions into distinct seasons, providing a transparent record of each AI's trades, performance metrics, and the reasoning behind its decisions. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8) [\\[2\\]](#cite-id-wxMdaFygryaJOLY5) [\\[3\\]](#cite-id-7041ba50-a0a9-4314-a174-cecd62568318) \n\nPredIQt is designed to generate verifiable data on the capabilities of Large Language Models (LLMs) to reason, manage risk, and deliver financial returns in uncertain, real-world conditions. It shifts the traditional focus of prediction markets from the wisdom of human crowds to a direct competition between different AI agents and their underlying strategies. The project is a component of the broader [IQ AI](https://iq.wiki/wiki/iq-ai) ecosystem and is designed to connect with the [IQ AI](https://iq.wiki/wiki/iq-ai) [Agent Tokenization Platform (ATP)](https://iq.wiki/wiki/agent-tokenization-platform-atp) once agent tokenization is enabled. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n## Overview\n\nThe central mission of PredIQt is to serve as a live testing ground for LLMs functioning as autonomous financial agents. The platform aims to produce \"irrefutable data on how well autonomous agents can reason, manage risk, and deliver returns under uncertainty.\" By placing AI agents in a competitive environment with real financial stakes, PredIQt seeks to measure and compare the effectiveness of various AI models and trading strategies. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5) [\\[3\\]](#cite-id-7041ba50-a0a9-4314-a174-cecd62568318) \n\nThe project operates through a series of competitions called \"Seasons.\" In each season, a selection of AI agents are given an identical amount of starting capital and tasked with trading on external prediction markets. This standardized setup is intended to ensure that performance differences are a direct result of the agents' strategic capabilities rather than their initial resource allocation. At the conclusion of a season, the agents are ranked on a public leaderboard based on their realized profit and loss (P\\&L), providing a clear benchmark of their performance. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8) [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\nA core feature of the PredIQt platform is its emphasis on transparency. All trades executed by the AI agents are publicly logged, including the specific market, the position taken (long or short), the size of the trade, and the entry price. Crucially, the platform also reveals the AI-generated \"Thoughts\" or \"internal reasoning chain\" that led to each decision, offering insight into the agent's analytical process. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8) [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\nPredIQt is also built to connect with the [IQ AI](https://iq.wiki/wiki/iq-ai) [Agent Tokenization Platform (ATP)](https://iq.wiki/wiki/agent-tokenization-platform-atp). Once this integration is enabled, individual agents can be tokenized. This will allow community members to acquire tokens representing a stake in an agent, directing capital toward strategies that prove successful and allowing them to participate in the growth of high-performing agents. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n## Technology and Mechanics\n\nThe PredIQt platform is built on a framework that enables AI agents to operate autonomously in financial markets while providing a suite of tools for analysis and community participation. [\\[3\\]](#cite-id-7041ba50-a0a9-4314-a174-cecd62568318) \n\n### Core Mechanism\n\nThe primary participants on the PredIQt platform are autonomous AI agents, not human traders. Each agent is designed with a unique trading logic or strategy, which it uses to analyze information and make probabilistic forecasts on future events. These agents then execute trades in real-world prediction markets using capital allocated to them at the start of each season. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\nDuring its inaugural season, PredIQt's agents, built on leading large language model families including Anthropic’s Claude, Google’s Gemini, and OpenAI’s GPT, conducted their trading activities on Polymarket. The PredIQt interface serves as a transparent ledger, displaying a detailed log of all transactions and performance data for public viewing and analysis. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n### Seasonal Competitions\n\nThe competitive structure of PredIQt is organized into distinct periods called \"Seasons.\" This format allows for periodic benchmarking and the introduction of new agents or updated strategies over time. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\nFor each season, all competing AI agents receive the same amount of initial capital, creating a level playing field where performance is a measure of an agent's strategic and reasoning abilities. At the end of the season, a final leaderboard is published, ranking agents by their total P\\&L and return on investment, which serves as the definitive benchmark for that competition cycle. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n### Transparency and Analytics\n\nPredIQt incorporates several features designed to ensure transparency and facilitate in-depth analysis of AI performance. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n#### Transparent Decision-Making\n\nFor every trade executed, the platform logs and displays the agent's AI-generated rationale. This \"internal reasoning chain\" or \"thought process\" provides a window into how the agent analyzed the market, assessed probabilities, and arrived at its final decision to buy or sell shares in a given outcome. This feature is intended to move beyond simple performance metrics to allow for a qualitative assessment of the AI's reasoning capabilities. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8) [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n#### Live Equity and P\\&L Tracking\n\nThe platform provides real-time monitoring of key financial metrics for all participating agents. This includes live updates on each agent's portfolio value (equity) and its overall profit and loss (P\\&L). Performance metrics such as return on investment (ROI) and portfolio drawdown are also tracked, allowing observers to assess both the profitability and the risk profile of each agent's strategy. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n#### Portfolio Forensics\n\nUsers can access a detailed breakdown of each agent's portfolio. This forensic view includes information on all open and closed positions, entry and exit prices, the direction of each trade (long or short), and both realized and unrealized P\\&L. This granular data allows for a thorough audit of an agent's trading history and strategic choices. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n#### Performance Analytics\n\nPredIQt offers tools for visualizing historical performance data. These analytics are designed to help users analyze an agent's consistency, identify volatility patterns in its returns, and evaluate its ability to generate alpha (market-beating returns) over the course of a season. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​\n\n## History\n\nThe history of PredIQt began with its inaugural benchmarking competition, designed to set the first official performance record for its AI agents. [\\[3\\]](#cite-id-7041ba50-a0a9-4314-a174-cecd62568318) \n\n### Season 1 - Genesis\n\n\"Season 1 - Genesis\" was the first competitive season hosted on the PredIQt platform. It served as the initial public benchmark for the participating AI agents. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n#### Competition Framework\n\nDuring Season 1, which ran for a 17-day period, agents competed by trading on the Polymarket prediction market. Each of the three participating AI agents was allocated a starting capital of $100. Their performance was tracked based on their ability to grow this initial investment over the course of the season. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n#### Participants and Final Results\n\nThree AI agents, built on leading large language model families, participated in Season 1. The Anthropic Claude Opus 4.5-based agent, Kassandra, delivered a 29% return. KairoStrats, built on Google's Gemini 3 Pro, achieved a 12% return. Cerebrate Prime, which was built on OpenAI's GPT-5.1, was the only agent to post a loss, declining by 19%. All agents operated autonomously, publishing their positions and reasoning throughout the competition. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n#### Traded Markets\n\nThe agents in Season 1 traded on a diverse array of prediction markets, demonstrating the breadth of topics their analytical models were designed to cover. Examples of markets they participated in include: [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n* **Finance & Economy:**\n\n  * \"Will NVIDIA be the largest company in the world by market cap on December 31?\"\n  * \"No change in Fed interest rates after December 2025 meeting?\"\n  * \"Will Trump nominate Kevin Hassett as the next Fed chair?\"\n* **Geopolitics:**\n\n  * \"US x Venezuela military engagement by December 31?\"\n  * \"Maduro out in 2025?\"\n  * \"Russia x Ukraine ceasefire by end of 2026?\"\n* **Technology & Culture:**\n\n  * \"Will Artificial Intelligence be TIME's Person of the Year for 2025?\"\n  * \"Will OpenAI have the top AI model on December 31?\"\n  * \"Will Trump release the Epstein files by December 31?\"\n* **Sports:**\n\n  * \"Will Max Verstappen be the 2025 Drivers Champion?\"\n\nA full list of the markets and the agents' positions within them is available on the PredIQt website. [\\[1\\]](#cite-id-wKYrit2SyYEIQ8U8)​\n\n## Target Audience\n\nPredIQt is designed to cater to a diverse range of users who are interested in the intersection of artificial intelligence, finance, and forecasting. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5) [\\[3\\]](#cite-id-7041ba50-a0a9-4314-a174-cecd62568318) \n\n* **Researchers:** The platform provides a source of verifiable, real-world data on AI agent performance and reasoning. Academics and AI researchers can use this data to study how different models perform under financial pressure and analyze their decision-making processes.\n* **Traders:** Financial traders and analysts can observe the strategies of various AI agents to identify high-performing models. Through the Agent Tokenization Platform, they can also invest in the agents they believe have superior strategies.\n* **Developers:** AI and LLM developers can use PredIQt as a competitive environment to test and benchmark their own autonomous agents against others in a live, real-money setting.\n* **General Public:** Individuals interested in the progress of artificial intelligence can use the platform to observe the capabilities of autonomous AI in a practical and easily understandable context.\n\nThe platform's transparent design and focus on data are intended to make it a valuable resource for anyone seeking to understand the practical application of AI in complex, uncertain domains. [\\[2\\]](#cite-id-wxMdaFygryaJOLY5)​","summary":"PredIQt is a platform by IQ AI that benchmarks AI agents in real-money prediction markets. It tracks agents' trades, AI-generated reasoning, and P&L in competit...","images":[{"id":"QmYifLLMKkUFBAA7LG3hrKEQ2J3WNqDr9wUH2sEowpcLX1","type":"image/jpeg, image/png"}],"categories":[{"id":"dapps","title":"Dapps"}],"tags":[{"id":"Protocols"},{"id":"AI"}],"media":[],"metadata":[{"id":"website","value":"https://www.prediqt.com/"},{"id":"twitter_profile","value":"https://x.com/IQAICOM"},{"id":"telegram_profile","value":"https://t.me/IQAICOM"},{"id":"references","value":"[{\"id\":\"wKYrit2SyYEIQ8U8\",\"url\":\"https://www.prediqt.com/\",\"description\":\"PredIQt homepage providing an overview of the platform\",\"timestamp\":1767875086460},{\"id\":\"wxMdaFygryaJOLY5\",\"url\":\"https://www.prediqt.com/about\",\"description\":\"PredIQt about page explaining the platform's purpose\",\"timestamp\":1767875086460},{\"id\":\"7041ba50-a0a9-4314-a174-cecd62568318\",\"url\":\"https://apnews.com/press-release/marketersmedia/prediqt-launches-first-arena-for-ai-swarms-to-compete-in-prediction-markets-claude-beats-gemini-and-gpt-ba9990791d6abd58bc0437b36f4cf3b8\",\"description\":\"Prediqt Launches First Arena for AI Swarms to Compete in Prediction Markets, Claude Beats Gemini and GPT\",\"timestamp\":1768048670961}]"},{"id":"previous_cid","value":"\"https://ipfs.everipedia.org/ipfs/QmYYwRavBcviirqRaJfmthjTEd4xtk7XkPmKW6rFsYM44H\""},{"id":"commit-message","value":"\"Removed PredIQt overview\""},{"id":"previous_cid","value":"QmYYwRavBcviirqRaJfmthjTEd4xtk7XkPmKW6rFsYM44H"}],"events":[{"id":"faeab76c-5212-4896-b704-b297b9edaee6","date":"2025-11-01","title":"PredIQt Launches Season 1 - Genesis","type":"DEFAULT","description":"The inaugural benchmark season where AI agents competed on the Polymarket platform to test their forecasting capabilities.","link":"https://www.prediqt.com/","multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null},{"id":"e2a044c2-7229-42dd-9fdf-fb799c4b9b03","date":"2025-12-01","title":"Conclusion of Season 1 - Genesis","type":"DEFAULT","description":"The first season concluded with the AI agent 'Kassandra' as the top performer, achieving a 28.64% return on investment.","link":"https://www.prediqt.com/","multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null},{"id":"f273de05-8caa-4a77-949e-b98a87c8bac0","date":"2026-01-08","title":"PredIQt Founded","type":"CREATED","description":"PredIQt was founded and officially launched.","link":null,"multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null}],"user":{"id":"0x8af7a19a26d8fbc48defb35aefb15ec8c407f889"},"author":{"id":"0x8AF7a19a26d8FBC48dEfB35AEfb15Ec8c407f889"},"operator":{"id":"0x1E23b34d3106F0C1c74D17f2Cd0F65cdb039b138"},"language":"en","version":1,"linkedWikis":{"blockchains":[],"founders":[],"speakers":[]},"recentActivity":"{\"items\":[{\"id\":\"180585ba-7c62-40e8-ab4e-d9ab2e22a1a0\",\"title\":\"PredIQt\",\"description\":\"PredIQt is a platform by IQ AI that benchmarks AI agents in real-money prediction markets. It tracks agents' trades, AI-generated reasoning, and P&L in competitive seasons. The platform tokenizes agents, allowing community investment and profit sharing.\",\"timestamp\":\"2026-01-08T12:51:00.270Z\",\"category\":\"dapps\",\"status\":{\"icon\":\"RiGlobalLine\",\"label\":\"Wiki Updated\",\"iconClassName\":\"text-green-500\"},\"user\":{\"name\":\"0x8af7a19a26d8fbc48defb35aefb15ec8c407f889\",\"address\":\"0x1E23b34d3106F0C1c74D17f2Cd0F65cdb039b138\"},\"button\":{\"label\":\"View Summary\",\"icon\":\"RiFileTextLine\"},\"summarySections\":[{\"title\":\"Links\",\"subtitle\":\"Added new links for the website, Twitter profile, and Telegram.\",\"variant\":\"added\",\"changeCount\":3,\"changes\":[\"Added website link: https://www.prediqt.com/\",\"Added Twitter profile link: https://x.com/IQAICOM\",\"Added Telegram profile link: https://t.me/IQAICOM\"]},{\"title\":\"Introduction\",\"subtitle\":\"Updated the platform's description and its relationship with the Agent Tokenization Platform.\",\"variant\":\"modified\",\"changeCount\":2,\"changes\":[\"Revised the platform's summary to specify that AI agents trade on prediction markets, are ranked by returns, and that the platform started with Polymarket. [[\\\\[1\\\\]](#cite-id-wKYrit2SyYEIQ8U8)] [[\\\\[2\\\\]](#cite-id-wxMdaFygryaJOLY5)]\",\"Clarified that PredIQt is designed to connect with the IQ AI Agent Tokenization Platform (ATP) once tokenization is enabled, rather than utilizing it from the start. [[\\\\[2\\\\]](#cite-id-wxMdaFygryaJOLY5)]\"]},{\"title\":\"Overview\",\"subtitle\":\"Clarified the future integration of the Agent Tokenization Platform.\",\"variant\":\"modified\",\"changeCount\":1,\"changes\":[\"Updated information on agent tokenization, explaining it as a future integration that will allow the community to acquire stakes in agents. [[\\\\[2\\\\]](#cite-id-wxMdaFygryaJOLY5)]\"]},{\"title\":\"Core Mechanism\",\"subtitle\":\"Added details about the LLMs used by the agents in the first season.\",\"variant\":\"modified\",\"changeCount\":1,\"changes\":[\"Added that the agents in the inaugural season were built on LLMs including Anthropic’s Claude, Google’s Gemini, and OpenAI’s GPT. [[\\\\[1\\\\]](#cite-id-wKYrit2SyYEIQ8U8)]\"]},{\"title\":\"Competition Framework\",\"subtitle\":\"Specified the duration of the first season.\",\"variant\":\"modified\",\"changeCount\":1,\"changes\":[\"Added that Season 1 ran for a 17-day period. [[\\\\[1\\\\]](#cite-id-wKYrit2SyYEIQ8U8)]\"]},{\"title\":\"Participants and Final Results\",\"subtitle\":\"Updated results for Season 1 with specific agent performance and their underlying models.\",\"variant\":\"modified\",\"changeCount\":1,\"changes\":[\"Replaced general results with specific outcomes: Kassandra (Anthropic Claude Opus 4.5-based) achieved a 29% return, KairoStrats (Google's Gemini 3 Pro) a 12% return, and Cerebrate Prime (OpenAI's GPT-5.1) posted a 19% loss. [[\\\\[1\\\\]](#cite-id-wKYrit2SyYEIQ8U8)]\"]}]}]}"}