Daedalus

/ 01 - Dispatch from the shop floorThe factory that quotes itself

It is a weekday in Karlsruhe and a CAD file - a bracket for a satellite, say, or a housing for a surgical tool - lands in Daedalus's inbox. Nobody flinches. Nobody walks the file over to a senior machinist for a squinting, coffee-fueled estimate. The software reads the geometry, picks the stock, plans the toolpath, finds a window on a CNC mill, prices the job, and emails the customer a quote. Hours later, the part is on a machine. Days later, it is in a crate.

This is the boring miracle Daedalus is trying to make ordinary. Precision manufacturing - the kind that makes the parts inside satellites, semiconductor fab tools, MRI machines and drone gimbals - has historically been the slowest, most artisanal corner of the industrial economy. Daedalus thinks that is a software problem. The company runs a single 50,000-square-foot factory full of off-the-shelf machines, and the trick - the only trick that matters - is the platform telling those machines what to do.

Caption: a sentence that would have gotten you laughed out of a German machine shop in 2014. It got Khosla to write a check in 2019.

/ 02 - The problemWhy nobody can get a small bracket made

Precision manufacturing in Europe runs on the Mittelstand: thousands of family-owned machine shops, most of them excellent, most of them booked solid, almost all of them dependent on a small number of people in their fifties who can look at a drawing and intuit how to make it. That intuition is the bottleneck. It is also retiring.

The result, if you have ever tried to source a small batch of tight-tolerance parts, is familiar. Quotes take weeks. Lead times stretch into months. Half the conversation is on the phone. The other half is in a PDF. There is no API for any of it. Meanwhile semiconductor equipment makers, defense primes and medtech startups all need more parts, faster, and to tighter specs than ever.

Caption: every contract manufacturer has a fax machine somewhere. Daedalus is the one without an excuse for it.

So there is a market gap big enough to drive a forklift through, and a labor shortage actively widening it. Most startups respond to this by becoming a marketplace - a thin layer of software on top of the existing shops, routing jobs and skimming. Daedalus made the more annoying decision. They went and built the shop.

/ 03 - The betAn OpenAI engineer walks into a machine shop

Jonas Schneider was OpenAI's first engineer. He co-founded the Robotics team, the one that famously taught a single robotic hand to solve a Rubik's cube. That project quietly killed itself in 2020, partly because the team concluded that the hardest part of getting machines to act on the world is not the model - it is everything around the model. The sensors. The integration. The grimy reality of a workpiece that is .03 mm out of where the simulator said it would be.

Schneider's read on that lesson was, in retrospect, the founding insight of Daedalus: the AI is the easy part. The hard part is owning the whole stack underneath it. So in 2019 he started a company whose product is not a model, not a robot, and not even, strictly speaking, software. It is a factory. The software just happens to be the thing that makes the factory worth running.

Caption: a Karlsruhe Institute of Technology alum returns home with a Y Combinator batch number and a strong opinion about toolpaths.

/ 04 - The productOne platform, end to end

The thing Daedalus sells, on paper, is parts. The thing they have actually built is the Manufacturing AI Platform - an end-to-end system that ingests a CAD file and pushes it through every step of the production lifecycle without anyone in a polo shirt walking it from desk to desk.

The platform learns. Every part finished is a row of training data: which strategy held tolerance, which tool wore out early, which fixture caused chatter. That dataset is the long-term moat. Daedalus's pitch to investors is not really about a single factory - it is that the second factory should be faster than the first, and the tenth should be faster than the second.

Caption: a comparison that flatters both companies and confuses exactly one CFO.

/ 05 - The proofCustomers, capital, calipers

The TechCrunch piece on the Series A makes the case the way investors prefer to hear it: $21M, NGP Capital leading, Addition and Khosla returning, total capital raised around $41M. The numbers are tidy. What is more interesting is the customer list - or rather, the part of it you can describe in public. Daedalus does not name names, for reasons familiar to anyone who has ever shipped a part with a NATO stock number on it. The sectors, though, are listed openly: semiconductors, energy, e-mobility, defense, pharma.

Chart: the only graph in industrial AI where the smallest bar is the most impressive number.

Investors call this kind of customer list "design partners." Operators call it "the people who have your CEO's mobile number." Both are correct.

/ 06 - The missionEuropean industrial resilience, one bracket at a time

If you read the daedalus.de site, the words "European industrial resilience" appear in the way other companies say "user growth." This is not an accident. The company's bet on Karlsruhe is also a bet on a continent that has spent the last three years rediscovering it has supply chains, that those supply chains run through places it does not control, and that the parts inside the equipment it relies on are made by people who are retiring.

Daedalus's framing - the founders' framing, the investors' framing, the framing on the press releases - is that precision manufacturing is a strategic capability, not a commodity. You cannot outsource the bottom of the stack and call yourself sovereign. The part where Daedalus is unusual, and slightly subversive, is in believing that the way to defend that capability is not to romanticize the Mittelstand but to give it an AI co-pilot.

Caption: a sentence engineered to land at exactly one Berlin policy conference per quarter.

/ 07 - MarginaliaThings they will not tell you on the front page

/ 08 - Why it mattersBack to the inbox

So return to the CAD file from the opening. In the old world, that bracket goes into a queue, somebody calls somebody, a quote arrives sometime next week, the part lands in six. The bracket is not the point. The bracket is every component in a fab tool, a satellite, a surgical robot. There are millions of brackets. The slow inbox is a tax on every physical thing Europe makes.

The Daedalus version is not magic and not yet finished. It is a software platform plugged into a normal-looking machine shop in Karlsruhe, run by a team that mostly does not look like a normal machine shop, owned by an engineer who has thought harder than most about why robots in the real world keep disappointing the people who built them. It quotes the bracket in minutes. It makes the bracket in days. It learns from the bracket so the next one is cheaper.

That is the entire pitch. It is not a moonshot, and they should be grateful for that. The bar is not "build something that has never existed." The bar is: take a process the world depends on, and make it boring in a new way. Daedalus is closer to that than almost anyone else trying.

Caption: the smartest thing the company does is refuse to be science fiction.