We're entering the Build with Gemini XPRIZE — the $2,000,000, 90-day competition run by XPRIZE with Google and Devpost that launched at Google I/O on May 19, 2026 and closes for submissions on August 17.
This is not a "we made an AI demo" announcement. The contest doesn't reward demos. It rewards a real business that operates with AI — real users, real revenue, and proof that AI is doing real work inside the operation. That's a high bar, and we're saying so out loud on purpose. EvalLense exists to make evaluation honest; it would be strange to be dishonest about our own.
What we're building
We're entering with a new line of work, not a repackage of what already exists. Under the rules you can run a new project through an existing legal entity, but the business itself has to be new — created after the submission window opened. So that's what this is: a fresh, AI-operated service built on the methodology we've hardened across 500+ internal pitch-deck review runs, stood up as something people can actually use and pay for.
It's the natural extension of our thesis. We've argued from day one that the hard part of AI evaluation was never the model — it was the methodology around it, and the human who keeps the final call. The XPRIZE is, conveniently, a place to prove that a methodology-first approach can run as a live operation, not just a product page.
The bar we're grading ourselves against
The contest scores three things in equal weight, and we'd rather hold ourselves to them in the open:
| Criterion | What it demands | Our honest read |
|---|---|---|
| Business Viability | Real customers, revenue by month, a sustainable model | The hard one. Pre-revenue doesn't win this. We have weeks to earn arms-length traction, not just ship |
| AI-Native Operations | AI executing real work in production | Our comfort zone — but we have to show AI running operations, while keeping the final decision human |
| Category Impact | Moves a real category, can scale | We fit Professional Services Access and Entrepreneurship; impact has to be shown, not claimed |
We're clear-eyed about the odds. The prize ladder is steep — $500k for first, $200k for second, $100k each for third through fifth, plus fifteen $50k runner-ups and $50k category awards — and the field is enormous. A 90-day window is short, the revenue bar is real, and the judges grade evidence: they're not required to test the product. None of that is a reason to sit it out. It's a reason to build the kind of evidence trail we ask everyone else to keep.
Why we're doing this in public
There's a tidy irony in a judging company entering a judging contest, and we like it. The XPRIZE publishes its rubric, requires a paper trail, and scores operations over impressions — which is exactly the discipline we sell. Entering keeps us honest: it forces us to live inside our own standard, on someone else's clock, with the scoreboard public.
We'll share what we learn as we go — what the evidence requirements actually feel like to meet, where the methodology holds, and where it has to bend. If you run evaluations of your own and want to compare notes — or see how we score a batch of submissions before your next deadline — book a demo. The verdict stays yours. We just bring the evidence.
Sources
- Build with Gemini XPRIZE — official rules (prizes, dates, criteria, requirements) — https://xprize.devpost.com/rules
- Build with Gemini XPRIZE — overview — https://xprize.devpost.com/
