The cost of being early

There is a real cost to being early, and the difficulty is that while you are in it you have no way of knowing you are in it. I have been early many times. Never by a decade, but by several years, repeatedly. Each time, another couple of years and I would have met the moment squarely.

Here is the inversion that took me too long to see. Being on the cutting edge feels fantastic. You are genuinely innovating, and you can tell. Which means the feeling of genuine innovation is itself a warning sign. If it feels like nobody has done this before, check whether what you actually have is a timing problem, because you still have to meet customers where they are, when they are.

The rough test I use now is about who else is in the room. If nobody else is building it, you are probably too early, or it is not the right thing. If the only people building it are entrenched incumbents, you are too late. If there are several young companies and no dominant incumbent, that is usually the moment.

What early actually costs is specific. It is harder to raise money, because you are asking investors to underwrite a market that does not exist yet. And it is harder to get customers, nearly always, because you are asking them to change before they have felt the reason to.

The hard question is how you tell too early from right idea, hold on. For a long time I thought that was the question, and it is not, because it presents a choice that is not real. It is not pivot or persevere. Find the version of the thesis that a customer will pay for today, and let that fund the wait. Do not dial back the vision. Dial back the offering until it clears today's willingness to pay. Willingness to pay is the only signal I know that reliably separates early from wrong.

Two examples I keep returning to. Amazon started with books, deliberately, because there are more titles in print than any physical shop could stock, which meant an online catalogue could offer something a store physically could not. That was a wedge, not the ambition. And Notion nearly died in 2015: Ivan Zhao laid off everyone except his co-founder Simon Last, they moved to Kyoto where they could live on far less, and rebuilt the product from scratch, cutting an ambitious block-based app-builder down to something people would actually use. The vision survived. The offering shrank until it fit.

My own version of early is a list. TinyOS was federated computing, and nobody would fund it, which is a reasonably ordinary idea now. In 2015 I split a convolutional net across a drone, a hub and the cloud, with staged confidence at each tier, because there was no other way to get the latency. At D-Watch we engineered privacy into a camera drone before the regulation existed, and then regulation arrived and killed the company anyway.

Financially, all of that was a bad bet. Personally it was a good one, and I do not say that to be gracious about it. The companies did not survive but the pattern recognition did. I knew what inference at the edge was going to need before it had a name, which is why I can architect a model-agnostic system in one pass now while people around me are discovering the requirement the expensive way. Things I learned in those years keep resurfacing in work I am doing this month.

So the trade is this. You pay in equity and years, and you buy expertise that arrives roughly five years ahead of the market. For an operator that is a good deal. For a founder it is brutal, which is exactly why the rule about dialling back the offering rather than the vision matters as much as it does.

“If it feels like you are doing something brand new, check whether you have a timing problem.”