DuetRight Labs · Our moonshot

A robot crew for every small business.

Robots are hard. Putting one to work in your business should be easy.

We're a Seattle general contractor, and we want to do for robots what we already do for renovations: find the right task, build the station, install it, teach the robot and keep it running, for a monthly fee.

Tell us your task

SeattleCo-founded by Angelo and Stan550+ completed jobs in the last two years100+ five-star reviewsLabs since September 16, 2026 · robot moonshot since September 26

Last updated September 28, 2026. Status: day one. We've bought our robot arm kit and installed its software. The arm isn't built or running yet, we haven't held our first small-business conversation, and there's no paying robot customer.

Watch

DuetRight Labs in two minutes

Stan walks through the idea: what we'd do for your business, how you'd pay, and where we honestly are today.

Earlier this year, AI researcher Alex Wissner-Gross wrote that the coming age of AI has plenty of people writing about it. Then he named what it's missing:
“…what it has always needed is a general contractor.”
Alex Wissner-Gross, “Solve Everything,” The Innermost Loop, February 14, 2026

In plain terms: robots need someone local who shows up, sets them up and keeps them working. That's what a general contractor does, and it's our trade.

The idea

Robot arms got cheap, and AI got good at learning tasks. Small businesses still have no one to put them to work.

Three things are true right now. Put them together and you get a gap that a general contractor can fill.

The gap

Big companies have automation engineers. Most small businesses don't. Think of the coffee shop, the bakery, the print shop: short on hands, with nobody on staff to set up a robot. The companies that install and maintain robots for a monthly fee mostly serve factories and big chains.

The moment

Open-source robot arms are now cheap enough for a hobbyist to build at home. New AI models learn a task from demonstration: a person walks the arm through the job dozens of times, and a model is trained on those recordings. Making it reliable is the hard part, and that's the real work.

The missing piece

Somebody has to show up, build it, install it, make it work and keep it running. That's a general contractor's job, and it's what we already are.

How it works

Like a renovation, but for robots

It's the playbook we run on every bathroom: look first, plan it, build it, install it right, then stand behind it.

Where we are: this is the model we're building. It's not for sale yet. We're proving it on our own bench first, then with one paying small business.

Assess

We visit your business and watch how a day really runs. Then we pick the one task worth handing to a robot first: something repetitive that eats hours, where it's easy to check whether the job got done right.

Design and make

We pick the robot that fits the task and design the station around it in Autodesk Fusion. Grippers, fixtures, brackets and mounts come off our Bambu P1S 3D printer, and the tougher metal parts will be cut on our 5-axis CNC, the InfiMaker K1, once it arrives. Every station is built to fit your space, not a factory floor.

Install

Power, mounting, safety and network. That's trades work, and trades work is what we know. We aim for stations that plug into an existing outlet. If one needs new wiring, an electrical contractor does it, with a permit.

Program

We teach the robot your task by demonstration and let AI do the heavy lifting. Before it goes to work for you, it has to prove itself on our test bench against a bar we set before the test starts.

Maintain

A monthly fee covers the service, fixes and updates. You can own the robot, or have it included in the fee; we're testing which works better for small shops. When your process changes, we teach the robot the new version. Once the first robot runs reliably, we find the next task. That's how one robot becomes a crew.

The bigger idea

Digital labor and physical labor

Why this matters to you: today you pay for software by the seat and for work by the hour. We think small businesses will soon buy the finished work instead, delivered by AI and robots.

Digital labor

AI agents

AI agents handle the screen work: drafting, follow-up, scheduling, paperwork. We already use them inside DuetRight, with a person signing off on anything that goes out.

Physical labor

Robots

Robots handle the hands-on work: moving, cutting, sorting, packing, wiping. This half is earlier and harder, and it's the half Labs is for.

Our thesis

DuetRight Labs helps small businesses move from buying software and hiring every hour of labor to buying outcomes delivered by AI and robotics. We start with one robot doing one task, and we keep going until every small business that wants a robot crew has one.

We don't build the robots. We make them work.

The hardware battle isn't ours

Figure, Tesla, Unitree, Apptronik and plenty of others are racing to build better robots. Let them fight that fight. Every robot they improve is one more we can put to work. We're not affiliated with any of them.

We own the last mile

The last mile is making a robot useful, reliable and worth the money inside an ordinary small business: picking the task, building the station, installing it, teaching it, keeping it running and upgrading it as the technology gets better. That's contracting, and contracting is our trade.

Where we are: this is a thesis, not a track record. No robot is working for a customer yet.

Two ways to pay

Own the robot, or pay for robot hours

Why this matters to you: you shouldn't need a big upfront bill or an engineer on staff to put a robot to work. We're testing two simple ways to buy, and owners like you will pick the winner.

Option A

You own it, we run it

You buy or finance the robot. We assess the job, set up and install the station, program the robot and keep it running on a monthly service plan: visits, fixes, updates, and re-teaching when your work changes.

For owners who want to own the equipment
You pay for the hardware once and for the care monthly

Option B

Robot hours

We own or lease the robot. You pay a monthly base fee plus the productive hours the robot actually works. There's no big upfront bill, and we carry the hardware.

For owners who want to try a robot without buying one
You pay for hours worked, not hours sitting idle

The math we're testing

$50–80

Per hour of human labor

A typical range for what one hour of work fully costs a business once wages, taxes, benefits and training are in. It's a common range, not our survey. Your own number is the one that counts.

Testing

Per productive robot hour

This is what we're finding out. It has to come in meaningfully lower, run after run, or there's no deal.

$21.30

Seattle minimum wage, 2026

Per hour, before taxes and benefits (City of Seattle).

Robots for sale today

We set a price after the bench proves the robot, not before.

The opportunity is the gap. If an hour of work fully costs you $50 to $80, and a robot can reliably do that same hour for meaningfully less, the difference stays in your business, and a fair share of it pays for our work.

Nothing is for sale yet. We ask about both options in every conversation, and on December 1 we pick based on what owners tell us and what our bench proves. Either way, the price follows the job: hours, tasks and response time. If the price changes, the job changes with it.

The proving ground

DuetRight is our first proving ground

Why this matters to you: we won't ask you to try a robot we haven't run in our own business first. By the time a workflow reaches your shop, it has a track record and a cost number.

Start with real work

We pick jobs our own crews do every week, where we already know exactly what done right looks like.

Put robots on it, shop first

We start with controlled tasks in our own shop: material prep, cutting, kitting (gathering the parts and supplies for each job) and fabrication. Work in client homes and businesses comes only after the shop tasks prove out.

Collect the data

Every run gets logged: how long it took, whether it passed, and whether a person had to step in. Any human touch counts as a miss.

Learn and improve

We fix the station, the fixtures and the teaching, then run it again. Every fix carries over to the next install.

Calculate the return

Hours freed, cost per productive robot hour and mistakes avoided, measured against the real cost of doing the job by hand. If the number doesn't work, the workflow doesn't ship.

Package what works

A proven task becomes a repeatable package: the station design, the parts list, the trained skill, the install checklist and the service plan.

Bring it to other businesses

Then we take that package to other small businesses with the same kind of work, and the loop starts again on the next task.

Every workflow earns a return-on-investment number before we sell it. Where we are: the loop hasn't started. No shop task is running yet, and the robot's first practice task, writing on cards with a pen, starts in October.

The map

One trunk, six branches

Everything in Labs hangs off one sentence: a robot crew for every small business.

A robot crew for every small business

Assess

Visit the business
Find the one task
Set the reliability bar
Find the price

Make

Pick the robot for the task
Autodesk Fusion, with AI
Bambu P1S 3D printer
InfiMaker K1 5-axis CNC

Install

Power and mounting
Risk assessment
Network
Electrical contractor where required

Program

Teach by demonstration
Open-source robot software
Prove it on the bench
Reteach when the task changes

Maintain

One monthly fee
Fixes and updates
Regular check-ins
Reliability tracked over time

Community

Meet Seattle robotics builders
DuetRight Robot Challenge (idea)
Robots for Main Street prize idea
Share what we learn

The lab

What's on the bench

Why this matters to you: every robot we'd put in your business gets built and tested here first. It's a small setup today. Here's what we have and what comes next.

Bought

SO-101 robot arms

The SO-101 is a small, low-cost robot arm anyone can build from free, public plans. We teach it by showing it: someone moves one arm by hand, and a second arm copies the motion and learns from it. We've bought the parts, but it isn't built or running yet; assembly starts in October. It's our practice arm. For a customer, we pick whatever robot fits the task, often an arm you can buy off the shelf, and build the station around it.

On hand

Bambu Lab P1S 3D printer

Our 3D printer is already in the shop. It will print the plastic parts: the arm's frame, the gripper fingers, and the holders and brackets. If a part doesn't fit, we change the design and print it again.

Picked, backing on Kickstarter

A CNC: the InfiMaker K1

Once it arrives, this machine will cut the metal parts that plastic can't handle: brackets, mounts, holders and gripper parts, in aluminum and even steel. It's an enclosed desktop machine that cuts from almost any angle, so complex parts take fewer steps, and it changes its own cutting tools. Parts we design on the computer can go straight to it. We're backing it on Kickstarter, with delivery estimated for February 2027. Until then, we print parts or have them made by a local machine shop.

Setting up

Autodesk Fusion, with AI

We design every station on the computer in Autodesk Fusion, a 3D design program. Since April 2026, AI assistants can work inside Fusion too, so AI can help us draw parts and make changes. A person checks every design before anything gets made.

Free and open

Open-source robot software

LeRobot is free software for teaching a robot by showing it: it records the demonstrations, trains the robot on them and runs the skill. It's built to work with the arm we're assembling.

Planned

The test bench

Every robot has to prove its task here before it goes to a customer. We set the bar before the test starts, for example 95 good runs out of 100, and we don't move it afterward.

How we chose the CNC

Our pick

InfiMaker K1, about $5,699 (Kickstarter late pledge)

According to the maker, it cuts from almost any angle, is fully enclosed, changes its own tools and is made for metal, including steel. Delivery estimated February 2027.

Considered

Makera Carvera Air, about $2,199

Enclosed and available now. Cuts aluminum and brass, from the top only. It was our first pick, until we decided robot parts need more angles and steel.

Considered

Carbide 3D Nomad 3, about $2,800

Very precise, but it only fits small parts.

Considered

Shapeoko 5 Pro, about $2,200–3,300

A large, open cutting table, priced by size. Great for big flat parts, less suited to metal.

Prices are the makers' listed prices as of September 27, 2026, and can change. A Kickstarter pledge isn't a store purchase, and delivery dates can slip. We haven't bought a CNC yet. We're not affiliated with InfiMaker, Makera, Carbide 3D or Autodesk.

Tracker

Where we are: day one

These are real numbers, zeros included, and we'll update them as they change. Last updated September 28, 2026.

Bought

Robot arm hardware

Robot arm kit bought and its software installed. Not built or running yet.

Tasks taught

Planned first practice task: writing with a pen on cards

Small businesses talked to

First lead: a local coffee shop whose owner makes every drink himself. We haven't talked yet.

Paid pilots

Goal: one signed pilot by December 1

DoneAug–Sep 2026Bought our first SO-101 robot arm hardware
DoneSep 16, 2026DuetRight Labs started
DoneSep 24–25Angelo and Stan attended Moonshots LIVE in Los Angeles
DoneSep 2026Found our first lead: a local coffee shop (we haven't talked with the owner yet)
DoneSep 26Went public with Labs as a moonshot: a robot crew for every small business

What's next is in the 90-day plan below.

Build log

What happened, and when

Why this matters to you: this is the honest record. Every step, decision and miss goes here as it happens, newest first.

October 3, 2026 · Bench update

Getting the bench ready

We have both SO-101 arms. The follower arm is assembled; the leader arm still needs assembly. Physical commissioning is still ahead.

On our M4 Mac, we’ve tested the software foundation for the next stage: LeRobot’s minimal setup, GPU and video checks, and the C development workflow.

The LaserPecker LP2 Plus is connected to its software over USB. We’re working on the first card artwork and test workflow. The engraved result and its alignment on the card still need to be checked on the physical machine.

PageSep 27, 2026Added a two-minute video at the top of this page. Stan walks through the idea, how you'd pay, and where we honestly are on day one.
PageSep 27, 2026Turned the map into a real mind map: our one sentence in the middle, the six parts of the offer around it in the order we do them, with lines drawing out as you scroll. Jev rated it clearer than the old grid. On phones it stays a simple list.
PageSep 27, 2026Added gentle motion to this page: sections rise into view, the mind map fans out, the log writes itself in, and a bar at the top shows how far you've read. Jev tested ten ideas; we kept the five that make the page easier to follow and dropped the gimmicks. It stays still if your device asks for less motion.
DecisionSep 27, 2026New CNC pick: the InfiMaker K1, a desktop 5-axis CNC that cuts metal, including steel. We're backing it as a Kickstarter late pledge, with delivery estimated for February 2027.
BuildSep 27, 2026Confirmed our robot arm kit (the motors and electronics; we print the plastic parts ourselves), downloaded the files to print those parts, and installed the free robot software on our computer.
AISep 27, 2026Jev (TypeSafe) now helps run Labs: it reviews this page for claims we can't back up, scores small-business conversations on our four task criteria, and sorts incoming “Tell us your task” messages. Contact details are stripped before anything is sent.
DecisionSep 27, 2026CNC pick: the Makera Carvera Air. We'll buy it only if our first customer task needs machined parts.
PageSep 27, 2026Published the full plan: two ways to pay, DuetRight as the proving ground, the first-customer plan, and how we'll keep installs safe.
ReviewSep 26, 2026A second AI (Codex) reviewed the plan independently. We moved customer conversations ahead of equipment purchases and corrected our conversation count to zero.
DecisionSep 26, 2026The moonshot: a robot crew for every small business. Labs went public on this page.
LearningSep 24–25, 2026Angelo and Stan spent two days at Moonshots LIVE in Los Angeles. Takeaways: monkey first, reliability is the bar, and robots need a general contractor.
StartSep 16, 2026DuetRight Labs started.
BuildAug–Sep 2026Bought our first SO-101 robot arm hardware.

We add to this log every time something happens. Tools named here are ones we use. We're not affiliated with their makers.

The first customer

How we'll find our first customer

Why this matters to you: if you run a small business, this is the conversation we want to have with you. We come to listen, not to pitch.

0 of 10

Conversations so far

First up: a local coffee shop. We haven't sat down with the owner yet.

By October 16

Five structured conversations, coffee shop first.

By October 26

A provisional first task, picked from those first five.

By November 30

Ten conversations across different kinds of small business.

The plain questions we'll ask

Walk me through a normal day. Where do you run out of hands?
What work do you do over and over? How many times a day, and how long each time?
How many hours a week does it take, and what do those hours cost you?
What does a mistake cost you when that job goes wrong?
If a machine did it reliably, would you pay for that? What would it be worth a month?
Would you rather own the robot and have us run it, or pay for robot hours?
What would make you say no to a robot here?

How we score a task

Each task gets a score from 1 to 5 on four things, with one line of evidence for each. We pick the top score with nothing below 3.

Repetitive: done the same way, many times a day
Safe: light, slow, no food contact, people clear of the station
Easy to check: a clear pass or fail on every run
Valuable: frees real hours every week, or wins sales you're losing now

About 30 minutes at your shop (our estimate). We record only with your permission. Tasks that touch food or drink come later, once a simpler task is paying, because they bring extra health rules.

The 90-day plan

September 28 to December 27, 2026

Why this matters to you: these are dates you can hold us to. The moonshot is a robot crew for every small business, and these 90 days work backward from its first step: one robot doing one real task for one paying small business.

By October 16

Start talking and building

Five structured small-business conversations, coffee shop first
SO-101 assembled and moving under a person's control (target: October 9)
Fusion and its AI connector set up

By October 26

Pick a first task

A provisional first customer task, picked from the first five conversations
First metal parts designed in Fusion for the InfiMaker K1 (delivery estimated February 2027)
Insurance, electrical and risk-assessment answers lined up (target: October 23)

By October 31

First real test

Our standard robot station designed in Fusion
First 100-run bench test of the practice task, with the number posted here, whatever it is

November

Find the first customer

Ten conversations in all by November 30
Build the chosen task on our bench and run its 100-run test
Test both ways to pay with real owners

December 1

Go or no-go

Go if we have a signed paid pilot and the robot clears a bench bar we set in advance, for example 95 good runs out of 100, with any human touch counted as a miss
If nobody will pay, we change the task or the kind of business. The moonshot stays.

Dec 2–27

Line up what's next

If it's a go: plan the install and build the customer's station
If not: start the next round of conversations and put the next task on the bench
Update this page with real numbers either way

Q1 2027

Run the pilot, if it's a go

Install in the customer's space
Run it against success measures written down before the install
Post the real numbers here

Every week

Build in public

What the robot did, what broke and what we learned
The tracker above, updated as numbers change

Safety and installs

Safe first, then fast

Why this matters to you: a robot in your shop has to be safe for your people and your customers, and it has to be installed right. Here's how we'll do it. Nothing is installed yet, so everything below is our plan.

Plan

A written risk assessment

Every station gets a written, task-by-task risk assessment before it goes to work, and a new one whenever the task, the station or the space changes. It covers installing, teaching and servicing the robot too, because that's when many robot accidents happen.

Plan

An emergency stop

Every station gets a red emergency stop within easy reach. After it's pressed, the robot stays stopped until someone resets it by hand.

Plan

Plug-in stations

Where we can, stations plug into an existing outlet, so unplugging cuts power to the robot. When a task needs new wiring, a licensed electrical contractor does the work and pulls the permit. We don't do electrical work ourselves.

Plan

Robot safety standards

We'll work from the current US industrial robot safety standard, ANSI/A3 R15.06-2025, which adopts ISO 10218 and absorbs the older collaborative-robot guidance in ISO/TS 15066. For service robots in shops and restaurants, UL 3300 applies. We won't claim a station meets a standard until the work is done.

Plan

Food-contact rules

If a task touches food, drink or a food-contact surface, it follows Washington's food code, with certified food equipment wherever possible. Before we design anything that touches food, we check the specific task with Public Health – Seattle & King County.

Before the first install

Insurance reviewed

Our insurance broker reviews coverage for robot installs and service before the first install. We won't say a robot job is covered until that's confirmed in writing.

Also planned: cameras watch only the robot's work area, with no audio, and your staff know they're there. The robot runs on its own network, never your point-of-sale system.

This is our plan, not legal advice or a certification. OSHA doesn't certify robots, and nothing we build will be called UL listed or NSF certified unless it actually holds that listing.

The moonshot test

Does it pass the moonshot test?

X, the moonshot factory (formerly Google X), looks for three things together: a huge problem, a radical solution and breakthrough technology. Here's how we stack up.

Our score: yes

Huge problem

Small businesses are short on hands, and automation is priced and staffed for factories. A U.S. Census Bureau working paper found that only about 2% of U.S. firms used robots in 2016–2018, mostly larger ones. Firms said robots didn't seem to fit their work or cost too much to set up.

Our score: yes

Radical solution

A neighborhood general contractor for robots, paid monthly. Robot subscriptions already exist for factories and big chains. Our bet is that a local trades crew can do it for small shops: we assess, build, install, program and keep the robot running, whether the shop owns it or we include it.

Our score: borrowed

Breakthrough technology

The breakthrough is real, and it isn't ours: robots that learn a task from demonstration, built by research labs and open-source teams. Our part is putting it to work, with parts we design in Fusion and make on our own 3D printer, with a 5-axis CNC coming next.

Monkey first

Astro Teller's rule: if the plan is a monkey reciting Shakespeare on a pedestal, don't start by building the pedestal. Do the hard part first.

Our monkey is this question: can we make one robot do one real task reliably for one paying small business? Until the answer is yes, the challenge and the prize idea stay on paper.

Double down if

A small business signs a paid pilot
The robot clears the bar we set in advance on our bench
The owner says they'd keep paying monthly
A second business asks for the same thing

Change course if

Nobody will pay for the task we picked: change the task
The robot can't clear the bar: narrow the task or rebuild the station
The price doesn't work for a real shop: change the kind of business

Verdict: the problem is huge, the solution is new, and the technology is ready to borrow. Now the monkey has to talk: one robot, one real task, one paying small business.

The prize angle

Hard to solve, easy to check

Why this matters to you: challenges bring smart people to hard problems, and the problem we care about is robots that work reliably in real shops like yours. At Moonshots LIVE we sat in on a session about how XPRIZE designs its prizes, and Peter Diamandis urged the audience to run competitions in their own communities. So we're drafting two. Both are on paper for now, with no dates, venues or sponsors yet.

The DuetRight Robot Challenge (Seattle)

Teams would program an open-source robot arm to do a real small-business task. The winner wouldn't be the flashiest demo. It would be the robot that does the job right, run after run, with nobody touching it.

A real task from a real shop
Judged on reliability: good runs out of many tries
Hard to solve, easy to check
Built for Seattle's robotics community

Why judge on reliability? At the first Amazon robot-picking contest, held in Seattle in 2015, half the teams scored zero. The organizers found the hard part wasn't any one piece, it was getting everything to work together. That's the kind of problem a general contractor solves. Seattle also has the people for it: in 2025 it hosted one of the local events of a worldwide robot-building hackathon, using the same kind of arm we're building.

Robots for Main Street (a prize idea)

This would be a prize for robots doing real jobs inside real small businesses, where the shop isn't changed to suit the robot and the cost targets are ones a real shop could afford.

Real businesses, not labs
The shop stays as it is and the robot fits in
Scored on reliability, not a one-time demo
A cost a real shop can afford

We plan to submit it if and when XPRIZE opens its next public call for prize ideas. XPRIZE hasn't reviewed or endorsed it.

Moonshots LIVE, Los Angeles

What we brought home

Why this matters to you: these ideas decide how we treat your business, starting with proving it works before we ask you to pay. Angelo and Stan spent September 24 and 25, 2026 at Moonshots LIVE in Los Angeles. Three ideas stuck.

Astro Teller, X

Monkey first: do the hardest part first. Drop an idea early if the basic numbers don't work. Be bold and humble at the same time.

For us, the hard part is one robot doing one task reliably for one paying business. That comes first, and we'll set the December 1 pass mark before the first test on the bench.

Salim Ismail, OpenExO

Keep a person accountable for what the AI does. Don't lose your people. Start new ventures outside the core business, where they can move fast.

Labs runs as its own mission, next to the renovation business, and our renovation crews keep doing what they do. Every robot we install will have a person at DuetRight who answers for it.

Peter Diamandis

Say your one sentence out loud. Run competitions in your community and in your company.

Our sentence is “a robot crew for every small business.” And we're drafting a Seattle robot challenge.

These lessons are paraphrased from our own notes. Any mistakes are ours, not the speakers'.

Follow along

We're about to start building this in public

Nothing's posted yet. Once the first arm is out of the box, we'll share the assembly, the bench runs, the misses and the wins. If you run a small business, it's the easiest way to see whether a robot could help yours before you ever talk to us.

YouTube X LinkedIn

How we run Labs

Our rules, and your turn

A few rules keep a moonshot honest. Then comes the part where you come in.

Honest numbers. The tracker shows what's real, zeros included.
Hard part first. We prove the robot works before we build anything that only looks good.
The bar comes before the test. We decide what reliable means before the robot runs.
A person answers for every robot. AI does the heavy lifting, and a person signs off.
Safety before speed. Every station gets a documented risk assessment before it goes to work, and a new one whenever it changes.
Your monthly fee is set by the job the robot does. If the price changes, the job changes with it.
Renovations keep running. Our renovation business keeps its full focus, and Labs runs as its own mission.

Got a task you'd hand a robot?

This page is us asking for help. If you run a small business, tell us the one task you'd hand a robot tomorrow. We'll listen, and if it's a fit, we'll come see it.

If you build, program or tinker with robots in Seattle, we'd like to meet you.

Planning a bathroom or kitchen renovation? We still do those, and we'd love to hear about it.

Tell us your task Build robots with us Start Your Project

Frameworks referenced: X's moonshot test (huge problem, radical solution, breakthrough technology), Astro Teller's “monkey first” and XPRIZE-style prize design. Speaker lessons are paraphrased from our notes on talks at Moonshots LIVE (Los Angeles, September 24–25, 2026). The quote near the top of this page is from Alex Wissner-Gross, “Solve Everything,” The Innermost Loop, February 14, 2026. The 2% robot-use figure is from a U.S. Census Bureau working paper (CES 22-12R, 2022) covering 2016–2018. DuetRight Labs is not affiliated with, sponsored by or endorsed by XPRIZE, X (formerly Google X), Moonshots, or any company or speaker named on this page. The DuetRight Robot Challenge and Robots for Main Street are ideas, not scheduled events or submitted entries. Product names belong to their owners. Last updated September 26, 2026.