❓ Problem

Yesterday a college student posted a photo of a 500mph turbojet RC plane he had built in his dorm room. Within hours, Boom Supersonic's CEO put up $100k for the first amateur to break Mach 1, a stranger offered to co-sponsor another $50k, and someone else added $25k for a no-sonic-boom variant. A quarter-million-dollar prize purse, assembled by strangers, in an afternoon, over quote-tweets. Rules "coming next week."

That is the whole problem in one screenshot. The energy is real, the money is real, and there is nowhere to put any of it. The prize lives on a founder's timeline until it scrolls out of memory. There are no rules, no escrow, no way to enter, and crucially no way to prove the dorm kid's plane actually hit the number.

Kaggle industrialised exactly this energy for data science. It turned company-sponsored technical challenges into a platform, grew to over 23 million accounts, and sold to Google in 2017. But Kaggle only cracked the easy half of the problem. Software entries are free to submit and trivially verified: you score a submission against a hidden test set in milliseconds, and nobody has to trust anybody. Hardware breaks both properties. Entries cost real money, real workshop time and real risk, and the results cannot be auto-graded by a server. A CEO cannot fund the dorm kid on a whim because there is no mechanism to confirm the plane flew, to hold the money safely, or to back the build before he has already spent his own savings on it. Demand exists, talent exists, capital exists, and the three never connect. Here is the idea.

✅ Solution

Testbet: a competition platform where hard-tech companies put up cash prizes for physical, technical challenges, and the world's garage engineers race to win them.

• A sponsor posts the ask, the prize and the rules. "First amateur to break Mach 1 with a self-built aircraft, $100k, here is what counts as proof." Structured, adjudicated, permanent, not a tweet that scrolls away.
• Anyone can pile into the pool. Other companies, universities, institutions and individuals stack funding onto the same challenge, turning one CEO's $100k impulse into a $500k purse. The improvised co-sponsoring that happened by accident on X becomes a one-click feature.
• Challenges run multi-stage, with milestone funding. Teams unlock money as they clear technical checkpoints (a validated 700mph run, then a clean transonic pass, then the record attempt) so the best builders get backed along the way instead of financing a Mach 1 attempt out of their own pocket.
• Verification is the product, not an afterthought. Standardised sensor kits, tamper-evident telemetry, live test days and third-party judging turn "trust me, it hit the number" into an auditable result. This is the exact thing Kaggle got for free and hardware never had. Own it and you own the category.

The wedge is narrow on purpose: be the home for the one prize that is already boiling over on Twitter this week. The platform underneath, pooled prizes plus milestone escrow plus trusted verification, expands into every hard-tech challenge a company would rather crowdsource than solve in-house.

📊 Key Numbers

Market size

• Global R&D spend reached roughly $2.87 trillion in 2024. The world's top corporate spenders alone pour around $1.2 trillion a year into it, and industrial R&D makes up about 39% of the global total. This is the pool a challenge platform quietly siphons from: prize money is just R&D spend routed through a crowd instead of a payroll.
• The prize-challenge market is small but proven. HeroX, the for-profit XPRIZE spin-off, already runs sponsor-funded challenges from a few thousand dollars up to $2.5M on a pay-only-when-solved model, and XPRIZE itself has run grand prizes from the $10M Ansari to the $30M Google Lunar prize. None of them is a self-serve, verified, built-hardware platform.
• Hard-tech capital is flowing again: advanced hardware and semiconductors attracted nearly $4 billion of the largest 2024 rounds, with robotics and physical-AI names like Physical Intelligence raising $470M. Every one of those companies has narrow physical problems worth a prize.

ARR potential

• Take-rate model: at a HeroX-style 15-20% platform fee, $100M of prize volume flowing across the platform is a ~$15-20M revenue business before you touch verification services or recruiting.
• Layer on paid verification (sensor kits, adjudicated test days) at a few thousand dollars per serious entrant, plus enterprise hosting retainers, and a focused platform has a clean path to $10-25M ARR within a few years.
• The real ceiling is the Kaggle ceiling. Google did not buy Kaggle for its hosting fees, it bought the community and the talent map. A verified leaderboard of the world's best garage hardware engineers is a strategic asset a defence-tech, aerospace or robotics giant would pay nine figures to own.