Josh Gruenstein on robotics as the ultimate systems problem
Tutor Intelligence deploys real robots into American factories and warehouses to earn the data that trains its models, treating robotics as one compounding system rather than a stack of separate bets.
The Ultimate Systems Problem
Gruenstein treats robotics as one system spanning models, hardware, software, supply chain, customers, and product, so drop any single layer and you lose most of the trade-offs you need to make progress.
robotics is not I think any one layer of the stack it's not just about intelligence it's not just about hardware it's not just about the customer.
Start From the Robot's Job
Rather than build a model and look for a use, Tutor deployed real robots into paying jobs first and let the data and the models follow from the work.
we started the company with this thesis that if we could just start deploying robots, we could build uh the label data sets that help us train these models and deliver this value to the American consumer.
The Flywheel Is the Asset
Deploying robots produces data, the data trains better models, better models do more and cheaper tasks, and that funds more robots, so the loop itself, not any one model, is what holds long-term value.
ultimately we think the thing that has kind of the long-standing generational value in a business is the flywheel and loops themselves.
Earn the Data, Don't Harvest It
Because Tutor's data comes from robots customers pay to do real tasks, it is economically valuable by construction, unlike data scraped from crews sent out hoping it turns out useful.
if you have robots in real customer sites that customers are paying for, for those robots to do that task, you know for sure that that data is economically valuable.
An Almost Empty Market
The US deployed only about 38,000 industrial robots in 2025 and nine of ten facilities have none, so the hard part is not beating rivals but convincing buyers who have never owned a robot.
We called them up and we're saying, "Hey, have you ever considered a robot for 16 bucks an hour?" And they haven't, right?
Robots by the Hour
Tutor sells the work a robot does divided by its cost, pricing it like hourly labor with no capital outlay, which flips robotics from a big upfront purchase to an ongoing service.
we have this framework of the value of a robot is you take how much work a robot can do and you divide it by its cost and that's it, right? Everything else is is commentary.
A Tutor in the Loop
When the model cannot yet do a task, a human teleoperator, a tutor, steps in to teach it, and whether to remove that human is treated as a cost calculation, not a matter of principle.
then it becomes a unit economics question, right? Like how much engineering do you have to do to get that last nine uh out of your models versus using tutors?
When It Works, It's a Machine
On the factory floor workers end up naming the robots and growing attached because the machine takes the backbreaking work off them, and the clearest sign a deployment succeeded is that the novelty wears off and the robot becomes ordinary equipment.
a robot is a robot unless it works in which case it's a machine.