Jason Zhao announces AI-driven simulation platform WhatIfVerse, aims to model living economies
On July 22 2026 the X founder revealed a sandbox where AI creates persistent societies, previewing a multi‑citizen test before a public release.
By Ryan Merket · Published
Why it matters
WhatIfVerse shows how generative AI can power persistent virtual economies, a step toward more realistic synthetic data for research, training, and new business models.
On July 22 2026, Jason Zhao used his personal X account to introduce a new project he calls WhatIfVerse. In a two‑part thread, Zhao framed the venture with a series of speculative questions – “What if AI took over the world? What if Elon Musk became the president of Mars? What if dinosaurs never went extinct?” – before stating that most existing AI tools can answer such hypotheticals, but none can create the world and let it persist.
https://x.com/jzqcjz/status/2079773856064299054
WhatIfVerse, according to Zhao’s brief post, is an AI‑generated simulation where virtual citizens work, earn wages, and interact within a continuously evolving environment. The platform is positioned as a sandbox for testing the dynamics of economies, societies, and policy outcomes, all driven by generative models rather than scripted rules. Zhao includes a short video clip from the X post that shows a rudimentary UI with agents moving on a grid, underscoring the prototype stage of the effort.
In a follow‑up reply to RuntimeWire, Zhao disclosed the current scale of the test. He said the team is working with around eight citizens and is testing the potential for 60‑70 citizens before a broader launch. “Will be released soon!” he added, indicating that the public version will arrive later in 2026. No timeline beyond “soon” was provided, nor were pricing or revenue models discussed.
The announcement arrives amid a wave of interest in AI‑powered virtual worlds. Since OpenAI’s 2025 release of its World Model framework, developers have been experimenting with agents that learn and adapt in simulated economies. Academic papers from MIT and Stanford have highlighted how synthetic societies can generate training data for reinforcement‑learning systems, while startups such as HyperCities and SyntheticSocieties have been courting venture capital to build commercial platforms for policy‑simulation and gaming. WhatIfVerse’s claim of a self‑sustaining citizen base that earns wages directly ties into this broader trend of using AI not just to answer questions, but to live inside them.
Zhao’s background is not detailed in the X thread, and no corporate filing or prior coverage could be located in public databases. His X handle, @jzqcjz, does not link to a corporate page beyond his personal timeline, suggesting that WhatIfVerse is being built as a founding project rather than a spin‑off from an existing venture. The lack of disclosed investors or a corporate entity points to a bootstrap approach, which is common for early‑stage AI research tools that rely on open‑source models and cloud credits.
The modest scale of the current simulation – eight active agents – reflects the computational cost of running generative‑model‑driven economies at any meaningful size. Zhao’s reference to a potential 60‑70 citizen test hints at the scaling challenges: each additional agent multiplies the interactions the model must compute, and maintaining a coherent wage‑earning system requires a reliable economic engine. If the team can expand beyond the current prototype while preserving stability, the platform could become a valuable testbed for economists, game designers, and policy analysts seeking rapid, low‑cost iterations of societal models.
For founders and operators watching the evolution of AI‑generated content, WhatIfVerse illustrates a shift from answer‑generation to world‑creation. The project’s public release, whenever it occurs, will likely attract early adopters who need synthetic data for training autonomous agents or for creating immersive narrative experiences. It also raises questions about governance: a world that “keeps living” will need mechanisms to prevent runaway economic loops or unintended bias in the AI’s decision‑making.
The X thread garnered modest engagement – 11 likes, 2 retweets, and over 11,800 video views – indicating a niche but curious audience. While the platform is still in alpha, the combination of an explicit economic layer and the promise of scalable citizen simulations positions WhatIfVerse as a noteworthy experiment at the intersection of generative AI and virtual society modeling.
Why it matters – The effort shows how generative AI is moving from single‑prompt answers to persistent, interactive environments. If successful, WhatIfVerse could provide a low‑cost “sandbox” for researchers, game studios, and policy labs to explore complex socioeconomic scenarios without building custom simulation engines from scratch.