Thomas Li

Two years into Chassis Autonomy we opened our own workshop and test centre. It was a genuine milestone — the moment we stopped borrowing other people’s facilities and could build and test our own hardware on our own terms.

When I spoke at the opening, I spent almost all of it on the people standing in the room. Every one of them had taken a real risk to join an early-stage company building something that did not yet exist, and the facility we were all admiring existed only because they had. Bricks and mortar are the easy part. They are also the part that gets celebrated.

I have always run teams on the same basis. Engineers are given the context and the authority to make their own decisions, because people who are trusted to decide things get better at deciding things. I stay reachable while they do it — and on the days I genuinely cannot stop, I say so and come back, rather than leaving someone to wait and guess.

Fail fast is a cliché, particularly in deep tech, but there are two layers to it. One is the product: build it, test it, break it, and improve it fast — spend the money on proving the design, not on engineering it to pass a programme review. The other is the person. I give engineers enough support to keep their heads above water and deliberately no more, because the mistakes they make themselves are the ones that stay learned. The two layers need each other: a team that has not been allowed to be wrong will not break its own hardware honestly, and hardware that never gets broken teaches nobody anything.

And I have worked in too many organisations where decisions simply do not get made. We’ll pick it up next time. Go and get more data. Nothing is ever refused and nothing ever moves. Where a call genuinely cannot be settled on precedent or evidence, I will find the one data point that tips the balance and make it — because a decision that turns out to be wrong has still moved the work forward, and you can fork back to the branch you left. Standing still is the only option that teaches you nothing.