The graveyard
Ideas that died of timing, not of being wrong. Most people read a dead company as a verdict. It is a date stamp.
//00 REPYH / An impact organisation by NUVC
The crazy ones are the only edge left.
AI made every reasonable idea cheap to build. What is left worth backing is the one that sounds insane right up until it is obvious.
//01 Is this you
You already know.
- 01You pitched it and watched a face do the polite thing.
- 02Your CV does not explain you, and it never will.
- 03You are rebuilding something the market already buried.
- 04You work in an industry that has never had a conference poster.
- 05You stopped trying to explain why you cannot leave it alone.
If none of that landed, you are fine without us. This is for the other kind.
//02 Why now
The filter selects for conformity.
CV, network, prior exit, warm intro — every filter the startup world runs on predicts one thing: whether you can execute. That is the part that just got cheap.
An outlier fails it by definition.
What did not get cheap is conviction that survives being laughed at. It has never been legible on paper, which is why the people who have it are still unfound. So we go where the filter is not pointed.
//03 Where we look
Three places crazy is still cheap.
Dirty jobs
Waste, freight, aged care, compliance, plant hire, the trades. Real revenue, real operators, no competition for attention.
Machines
Robots, rigs, sensors — anything that has to survive contact with the real world. Most people still price hardware as slow and expensive. It got cheap while everyone was watching the models.
AI is the why-now on all three, never a lane on its own. A lane that is only “AI” selects for the same legible founders every accelerator is already outbidding us for.
//04 What we put in
Tokens, compute and hardware.
Building got cheap, not free. The gap is where these ideas quietly stop, so that is the only place we spend.
Tokens
Inference credits on open-weights models. Enough to run something real, not a demo allowance.
Generative capacity
Video, image and audio credits, for teams whose product is the output rather than the model.
AI infrastructure
GPU time, hosting, deployment — the unglamorous half, where most of these actually die.
Robotic hardware
Kits, boards, sensors, machines. You cannot prototype a waste route in a browser tab.
Categories we source, not confirmed sponsors. Never a stake, a SAFE or a note — you keep every share of what you build.
//05 The charter
Six things we will never do.
- 01Take equity.
- 02Run a demo day.
- 03Sell early access.
- 04Give one VC better terms than another.
- 05Stand between you and a recruiter.
- 06Hide where people land, including with competitors.
We can hold these because we are not investing. NUVC takes no equity in anyone here, so nothing we publish has to protect a position.
//06 Plainly
What this is not.
- —Not a business. REPYH earns nothing. NUVC funds it, and its business is elsewhere.
- —Not an accelerator. No cohort, and no portfolio to graduate into.
- —Not a platform. No accounts, no app, nothing to log into.
- —Not a retention product. Success is where you land next, not how long you stay.
- —Not a hunt for the best founders. That field is crowded and already well served.
//07 Register
Put your name down.
No application, no pitch deck, no fee. A name and an email, and we come back to you when there is a date.
That did not go through — check the name and email, then try again.