Pax Historia
AI-powered historical grand-strategy game with 35K daily active users.
On the record
Checkpoint: Fine-tuned model performance data and mobile launch via new Unity hire expected
The call
Pax Historia has fast early traction and no revenue disclosed yet, and it has not shown its fine-tuned model or community lock-in can outlast bigger, better-funded rivals.
Pax Historia lets players build and run imaginary worlds using AI that responds to their choices. Right now it works because it uses existing AI tools (like ChatGPT) in a clever way, but anyone with programming skills can copy this exact formula, as people already have on the internet. The game also faces bigger rivals: Paradox Interactive could add AI storytelling to their famous games, OpenAI or Google could build the same thing into their own products, and a competitor called Myth-OS is already selling something similar on Steam. The one thing that could save Pax Historia is if they train their own custom AI model on thousands of hours of how players actually play the game, making their version noticeably smarter and more fun than the copies. They say they are working on this, but no one outside the company has proven it works yet, and they probably have less than a year and a half to prove it before bigger competitors catch up.
Why
01
AI pressure: OpenAI, Google, or Anthropic shipping native 'build-your-own-AI-world' features. Pax Historia runs on frontier models via OpenRouter, so if those labs ship similar world-building tools natively, the core loop becomes a free feature. Could AI do the core job? Yes, likely.
02
Bigger players: Paradox Interactive. Paradox owns the grand-strategy audience and could bolt AI narrative tools onto existing franchises like Crusader Kings or Stellaris. Could a giant copy it? Yes, easily.
03
What protects it. Early lead in a new category: 35K daily users, 20M rounds, and 4,000+ user scenarios, though none of this is locked in yet.
What would change this call
Pax Historia ships a proprietary fine-tuned model beating frontier models on gameplay benchmarks by Q1 2027. Would show the roadmap moat is real, not just a plan.
Myth-OS or a frontier lab ships a comparable build-your-own-AI-world feature before Pax Historia's moat lands. Would show the open window closed before Pax Historia could defend it.
Pax Historia's DAU or 20M+ rounds figure declines sharply in later reporting. Would echo AI Dungeon's collapse and signal retention is not sticky.
Myth-OS or a frontier lab ships a comparable build-your-own-AI-world feature before Pax Historia's moat lands. Would show the open window closed before Pax Historia could defend it.
Pax Historia's DAU or 20M+ rounds figure declines sharply in later reporting. Would echo AI Dungeon's collapse and signal retention is not sticky.
Key numbers
- 35,000 daily active users
- 20 million+ rounds played
- 4,000+ user-published scenarios
Named risks
- Core loop already cloned by open-source projects on GitHub as of mid-2026, no proven technical moat.
- Engagement could collapse like AI Dungeon did, from ~500K peak to ~26 concurrent players by April 2026.
- Heavy dependence on third-party frontier LLM inference cost and availability via OpenRouter and Infron AI.
- Very small team (3 to 6 people) supporting 35,000 DAU and millions of tokens per hour, operational strain.
- Unproven monetization transition from token pricing to subscription passes as of April 2026.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| AI Dungeon | gaining on | AI Dungeon's Steam concurrent players fell to ~26 by 2026-04-18 while Pax Historia grew to 35K DAU |
| Myth-OS | contesting | Steam-listed AI sandbox directly overlapping Pax Historia's grand-strategy pitch, active 2026 |
| Paradox Interactive | gaining on | Paradox (2.2B SEK 2024 revenue) has not shipped a comparable AI-narrative engine as of late 2025 |
| Open-source clones (GitHub) | losing to | At least two clones explicitly inspired by Pax Historia public by 2026-06-17, eroding technical moat |
Questions people ask
- Is Pax Historia profitable?
- No profitability data has been disclosed. Pax Historia monetizes via token purchases and a 'Pax Patron' subscription pass (transition announced April 2026), but no revenue or ARR figure exists in any public source as of April 2026, unlike YC W26 peer Pocket's disclosed $27M ARR.
- Could AI replace Pax Historia?
- Pax Historia's entire product IS AI, LLMs generate all game outcomes, so there is no legacy workflow to defend. Its core loop is already replicable: at least two open-source clones explicitly inspired by Pax Historia appeared on GitHub by June 2026, and it relies on 28+ commodity models via OpenRouter.
- Who founded Pax Historia?
- Pax Historia was co-founded by Eli Bullock-Papa and Ryan Zhang, reportedly Virginia Tech freshman-year roommates who built the original prototype at a hackathon. The company went through Y Combinator's Winter 2026 (W26) batch, per an April 2, 2026 profile.
- How many users does Pax Historia have?
- Pax Historia reports roughly 35,000 daily active users and over 20 million rounds played as of April 2, 2026, per company materials and third-party coverage. It also processes an estimated 5 million AI requests per month through OpenRouter.
- Who is Pax Historia's biggest competitor?
- Myth-OS, a Steam-listed AI sandbox, directly overlaps Pax Historia's grand-strategy-plus-narrative pitch. AI Dungeon, the longest-running AI game-master platform, is a cautionary precedent, its Steam concurrent players fell to about 26 by April 18, 2026 after peaking near 500,000.
- Has Pax Historia raised venture funding?
- Total funding is reported between $500,000 and $1 million as of March 24, 2026, from accelerator and pre-seed sources including Y Combinator, Z Fellows, and Pace Capital. No priced institutional seed round or valuation has been publicly disclosed.
Timeline
- 2026-06-17Open-source clone 'Local-Pax-Historia' explicitly inspired by Pax Historia appears on GitHub
- 2026-05-30Second open-source clone 'Tommi-K/pax-historia' appears on GitHub
- 2026-04-18AI Dungeon Steam concurrent players decline to about 26, cited as cautionary precedent
- 2026-04-16Canopywave case study describes Pax Historia's shift to subscription-pass pricing (vendor-claimed)
- 2026-04-13BuiltIn lists Pax Historia at 6 total employees
- 2026-04-02Claude's Corner profile reports 35,000 DAU and over 20 million rounds played
- 2026-03-24YC W26 batch report notes Pax Historia's launch tweet got most-liked reaction in 199-company batch
- 2026-02-04Official Launch YC page for Pax Historia published
- 2025-12-27Founding Engineer (AI/ML) job posted, San Francisco Dogpatch, $150K to $240K plus equity
- 2025-08-18Pax Historia alpha version launches
Revision history
27 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Pax-Automata research doc (2025-08-17)
- Claude's Corner: Pax Historia, AI Rewrites History (2026-04-02)
- Founding Engineer (Full Stack Web) - Pax Historia (2026-04-13)
- YC W26 Batch Breakdown (2026-03-24)
- Founding Engineer (AI/ML) at Pax Historia (2025-12-27)
- Valuation, Funding Rounds & Stock Price (2026-03-24)
- Jobs at Pax Historia
- Best Alternatives for Immersive AI Roleplay (2026-04-18)
- Paradox Interactive
- francescodemunari/Local-Pax-Historia (2026-06-17)
- Steam Franchise: Paradox Interactive (2026-03-08)