Annoying Is Investable
Don’t be discouraged if what you produce initially is something other people dismiss as a toy. In fact, that’s a good sign.
I think people read this as “haha remember when computers were toys and now look at Apple” which is true but also not the interesting part.
There’s nothing more valuable than an unmet need that is just becoming fixable.
Become Fixable. This is such a ridiculously good way to think about technology.
There are an effectively infinite number of important problems in the world. Fusion would be nice. So would teleportation. I would like a robot that cleans my house, cooks my food, does my laundry and costs $4,000. I would also like a 200 Wh/kg electric airplane and a machine that prints defect-free single crystal turbine blades in my garage.
These are all enormous markets. Unfortunately the universe does not care about TAM. The interesting moment is when something changes and a problem moves from impossible to this is terrible but technically it works.
Maybe lasers got cheaper. Maybe compute got 1000x better. Perhaps some idiot figured out how to manufacture the thing without aligning 40 microscopic parts by hand. Possibly a material that used to delaminate after 20 cycles now survives 2,000. Maybe launch went from $50,000/kg to $2,000/kg. Suddenly you can suddenly buy the sensor for $80 instead of $8,000. Maybe the model that could barely identify a cat three years ago can now operate a computer.
Nothing magical happened to the problem. It just became slightly less impossible. And the first thing built immediately after something becomes slightly less impossible usually looks stupid.
You are looking at generation zero of a technology and comparing it against generation 900 of whatever it is supposed to replace. The incumbent has had fifty years of engineers removing every mundane little failure mode. There are factories producing millions of them. There are standards and suppliers and technicians and textbooks and weird German men who have spent their entire careers optimizing one bearing.
Then three guys from Stanford show up with a machine that costs 8x more, works 63% of the time and occasionally destroys itself.
Toy?
Correct. The mistake is thinking “toy” is a terminal state.
Humans are extremely good at comparing things and weirdly bad at comparing rates of change. If the existing machine works 99.99% of the time and your machine works 60% of the time, the existing machine is obviously better.
But what if yours worked 10% of the time eighteen months ago? Now I am interested.
A humanoid robot that successfully folds laundry 60% of the time is useless. I am not buying it. I would spend more time rescuing socks from the robot than folding them myself.
A humanoid that went from 10% to 30% to 60% in eighteen months is terrifying. Same robot. Same 60%. Completely different information.
This is maybe one of the recurring mistakes in deep tech investing. Everyone wants to know whether the thing is good. Obviously I want to know whether the thing is good. But sometimes I care much more about why it is possible now when it wasn’t possible five years ago.
Because “we built a better machine” is nice, but “We are sitting downstream of three curves that just crossed” is something else.
This is also why I think some of the best startup ideas initially sound unserious. If everyone agrees something is technically possible, economically sensible, urgently needed and ready for mass adoption, congratulations, McKinsey has already published a useless 94-page PDF about it and seventeen companies raised Series Bs last year.
You probably aren't early. Early looks much worse.
Early is a shitty robot doing a job nobody trusts robots to do, or a semiconductor process with horrific yield, a manufacturing technique that produces a beautiful part and then produces garbage four times in a row or a computer that hobbyists buy because apparently normal people have no reason to own a computer.
The toy is evidence that you might have arrived at the boundary. Not always, however. There is an important distinction here because otherwise you can use this argument to justify literally any garbage company.
Sometimes the robot sucks because robotics is hard and will continue to be hard, or perhaps your new battery chemistry has terrible cycle life because God hates you. It could also be that the manufacturing process you created has 40% yield because it is fundamentally a bad manufacturing process. Sometimes the toy is just a toy.
The question is what changed?
A technology becomes interesting when the bottleneck moves from physics to engineering. Costs fall, a process becomes repeatable, some previously exotic piece of equipment becomes commercially available. Something that needed a national lab can suddenly be done by eight sleep-deprived people in a warehouse in Bangalore.
That transition matters. Not impossible → solved. Impossible → annoying. Annoying is investable. You can hire around annoying, iterate around annoying, throw capital and 5,000 test cycles at annoying.
You cannot outwork physics.