Skip to content
QwestLab
Saints first look

QwestLab / Signal 01 · Overview

Real work.Seen first-person.Taught to machines.

Signal 01 is QwestLab’s first camera fleet: twenty first-person cameras, worn by paid contributors while they do ordinary work. The footage trains the AI systems now learning to handle physical tasks.

The people who back the fleet are not the people holding the cameras. Nothing is open yet — this page explains the idea and asks whether you would take part.

What egocentric video isWould you take part?
First-person view of a contributor's hands washing a white plate with a sponge under a running tap.
Washing dishesAccepted clip · the contributor's own point of view
  • 20 cameras · one first fleet
  • Contributors do the recording, and are paid for it
  • A temperature check · nothing is charged

What egocentric video is

The view from your own eyes, while your hands do something.

Ego means self. Egocentric video is footage from a person's own point of view: what they see while they work. Nobody films them from across the room. There is no camera operator and no performance. Just their hands, doing what they already do.

Overhead view of someone else's hands chopping spring onions on a wooden board, filmed from across the counter.

Third-person

Someone films you

The camera is across the room. The viewpoint belongs to the operator, not to the person working. Useful for many things, but it is not what a machine will see when it does the task itself.

First-person view of a contributor's hands smoothing a folded navy T-shirt on a bed.
Folding laundryAccepted clip · both hands visible

First-person

Egocentric

The camera sees what the worker sees. Both hands stay in the frame; that is the whole technique. It is the viewpoint a machine will eventually have, so it is the viewpoint it learns from.

A person in an apron wearing a phone on a head strap at eye level, looking down at a kitchen counter.
Head mount. A strap holds the camera at eye level. The view is the wearer’s.
A person wearing a phone in a chest harness while drilling into a plank in a workshop.
Chest mount. The same job, worn lower, with the hands centred.

Not new, and not a gimmick

Research groups and companies have collected first-person footage for years. It is an established category of training data. What Signal 01 adds is a paid, reviewed workforce doing it at fleet scale, with dedicated cameras rather than a phone on a strap.

How AI uses it

A robot cannot learn to fold a shirt from a manual.

It learns the way people learn a physical skill: by watching it done, many times, up close. First-person footage shows the hands, the object and the result together, from the viewpoint a machine will eventually have.

First-person view of two hands holding an empty glass mug over a dark marble kitchen counter beside a gas hob.
Reaching for a mugFirst-person · the viewpoint a robot arm will have
  1. 01

    Robots that handle objects

    Humanoids and robot arms learn everyday manipulation from demonstrations: washing, folding, sorting, restocking. Thousands of real examples teach what a good grip, a safe motion and a finished task look like.

  2. 02

    Assistants that see what you see

    Wearable assistants on glasses and headsets have to recognise what the wearer is doing and what comes next. They are trained on the same viewpoint they will operate from.

  3. 03

    Models that understand physical work

    Vision-language models learn to describe and check a task: is the plate clean, is the shirt folded, was a step skipped. Reviewed first-person footage is the teaching material.

  4. 04

    Why real beats staged

    A staged, camera-ready version teaches the wrong thing. Ordinary kitchens, ordinary lighting and ordinary hesitation are the point, so footage is recorded by people doing their real tasks.

From a pair of hands to a trained model

  1. 01Wear

    A contributor puts on the camera.

  2. 02Record

    They do an approved task the way they always do it.

  3. 03Review

    A person checks the clip: framing, both hands visible, no faces or private details.

  4. 04Accept

    Only accepted footage counts, measured in hours.

  5. 05Train

    A commercial buyer feeds those hours to the models and robots above.

Contributors control the camera; there is no hidden recording. Faces and private details are screened before anything reaches a buyer, and a contributor’s name and address are never handed over with the footage.

How big it is

The data behind physical AI is a market of its own.

Every model on this page is trained on collected, reviewed, labelled data. That is a bought-and-sold market with published estimates. Two of them frame Signal 01: the market its footage is sold into, and the robotics demand that is pulling on it.

  1. AI data collection and labelling

    The market Signal 01's footage is sold into

    +28.4% a year

    $17.1Bby 2030

    from $6.3B in 2026

    Source · Grand View Research

  2. Embodied AI

    Robots that learn from demonstration: the demand behind the buyers

    +39.7% a year

    $67.6Bby 2033

    from $6.5B in 2026

    Source · Grand View Research

  3. Where the supply comes from

    The first contributor pool: QwestLab was built inside Kumu, for Kumu's community first, starting in the Philippines.

    100,000+

    monthly active Kumu users

    Source · QwestLab contributor guide

Third-party estimates for the categories QwestLab sells into, quoted from QwestLab's market sizing of August 2026. They describe the market, not Signal 01's revenue. The embodied-AI figure is evidence that robotics demand is real; it is not part of any Signal 01 arithmetic.

Where Signal 01 fits

Twenty cameras, one first fleet.

Signal 01 funds and deploys QwestLab's first twenty dedicated first-person cameras. Contributors from the Kumu community wear them through approved tasks. Every clip is reviewed by a person. Accepted hours are sold to a commercial buyer that already pays for this kind of footage. QwestLab runs recruiting, deployment, review and payouts. The people who back the fleet do none of that work.

Four young creators in a warmly lit room, smiling at their phones, one filming with a phone on a small tripod and ring light.
Kumu's creator communityWhere QwestLab recruits its first contributors
  1. What it is01 / 03
    • QwestLab's first fleet: twenty dedicated first-person cameras, bought once and put to paid work.
    • A capped, one-time structure. It does not grow with the company.
    • An early look for the Saints community, before anything opens more widely.
  2. What it is not02 / 03
    • Not live. The fleet is planned, not deployed, and no number on this page is a result.
    • Not a job. Backing a camera does not mean recording anything yourself.
    • Not something the company needs. QwestLab scales on its own capital either way.
    • Not an offer and not advice. This page carries no pricing or wallet information.
  3. What comes next03 / 03
    • A decision, once we know whether anyone wants it. The temperature check below is how that gets answered.
    • Photographs of the production cameras, once QwestLab has approval to show them.
    • Questions in the meantime: ask in the QwestLab channel.
Conceptual top-down rendering of twenty identical camera units arranged in a grid, each labelled POD01-C01 to C20 and marked planned.

Pod 01 · 20 units · state: planned

A conceptual visualization of the twenty-unit pod, not the production camera. Every unit starts as planned; nothing on this page pretends the fleet is already running.

Who does what

Funding a camera is not a job.

Holders fund the cameras. Contributors operate them. The two are never the same people. If you back the first fleet, you are not signing up for shifts, tasks or footage. The recording is done by paid contributors who chose that work.

  1. Contributors01 / 03

    Wear the cameras and do the work

    Qualified people from the Kumu community, recording approved everyday tasks from their own point of view. They are paid per accepted hour for that work.

    Paid per accepted hour

  2. QwestLab02 / 03

    Runs the program

    Buys and assigns the cameras, recruits and trains contributors, reviews every clip, sells accepted hours to the buyer, and handles payouts.

    Keeps half of net operating profit

  3. Holders03 / 03

    Fund the fleet, and nothing else

    A unit funds hardware. It is not a job, a shift or an assignment. Holders never wear a camera, never record, never handle footage and are never asked to do any work.

    Share the other half, split across units

Saints of LA: Eternal — the collection's characters gathered in golden cloudscape above the Los Angeles skyline.
Saints of LA

First look · no obligation

Why the Saints get this first.

Saints first

Signal 01 goes to the Saints community first, so the people who have been here early get a look before anyone else.

No obligation

Nobody has to take part. There is no allocation to hold on to, nothing to defend and no penalty for sitting this one out.

The company scales either way

QwestLab is scaling either way. The company's growth is funded on its own capital and does not depend on this pod — Pod 1 exists so the community can be early, not because the roadmap needs it.

What a unit could earn

Modelled from stated assumptions. Not results.

Nothing has been earned yet, because no camera has been deployed yet. What follows is arithmetic on figures QwestLab has published: what an accepted hour pays, what it costs, and what would be left for 280 units if the cameras were busy for a given number of hours. Change the hours and every figure moves.

Where one accepted hour goes

Paid for one accepted hour
$10

What a commercial buyer pays for an hour of footage it accepts.

To the contributor
− $5

The person who wore the camera and did the work.

QwestLab operating cost
− $1.25

Review, support, payment fees, hardware reserve. An estimate, not finalised.

Net operating profit
$3.75

Split 50/50 between QwestLab and the holder pool.

Only accepted hours pay. Footage that fails review earns nothing, and a camera nobody wears earns nothing at all.

Three assumptions, all unproven

  1. Conservative

    24 hrs / camera / month

    $3.21per unit / month

    Accepted hours
    480 / mo
    Holder pool
    $900 / mo
    Across
    280 units

    What QwestLab is planning around. QwestLab communicates publicly around this case. It is a planning assumption, not a result.

  2. Expected

    77 hrs / camera / month

    $10.31per unit / month

    Accepted hours
    1,540 / mo
    Holder pool
    $2,887.50 / mo
    Across
    280 units

    Quoted by the commercial buyer, not by QwestLab: 3.5 hours per day, weekdays. Unverified at QwestLab's actual operating level — treat it as unproven upside, not an expectation.

  3. Maximum

    88 hrs / camera / month

    $11.79per unit / month

    Accepted hours
    1,760 / mo
    Holder pool
    $3,300 / mo
    Across
    280 units

    Quoted by the commercial buyer, not by QwestLab: 4 hours per day, weekdays. Unverified at QwestLab's actual operating level — treat it as unproven upside, not an expectation.

Shown in dollars per unit rather than as a percentage or a yield, deliberately. A rate implies a schedule nobody can promise.

Run it forward.

Pick a number of units, an assumption and a horizon. The curve is what identical months would add up to; the dashed line is what you paid; the marked month is where they would first meet. Then read the small print underneath, because no month will actually be identical.

Units · 1 to 20

Assumption · accepted hours per camera per month

Horizon

Outlay
$30
1 × $30
Per month, modelled
$3.21
$3.21 per unit
After 24 months
$77
identical months, before any costs to you
Outlay covered by
Month 10
within this horizon
1 unit at $30 is an outlay of $30. At 24 accepted hours per camera per month the modelled pool would pay $3.21 a month for the position, $77.14 over 24 months. The cumulative total would first cover the outlay in month 10.$25$50$75$100StartM12M24Outlay $30Month 10 · outlay covered
Cumulative modelled payout for the position, month by month, against what it cost. Assumes every month is identical, payouts arrive monthly and on time, all 20 cameras stay in service and the buyer keeps buying. None of those is guaranteed, and the first months of a new fleet are usually the weakest.

Month by month · cumulative

  1. M1$3
  2. M2$6
  3. M3$10
  4. M4$13
  5. M5$16
  6. M6$19
  7. M7$23
  8. M8$26
  9. M9$29
  10. M10$32
  11. M11$35
  12. M12$39
  13. M13$42
  14. M14$45
  15. M15$48
  16. M16$51
  17. M17$55
  18. M18$58
  19. M19$61
  20. M20$64
  21. M21$68
  22. M22$71
  23. M23$74
  24. M24$77

Before the outlay is coveredThe month it is first coveredAfter

Illustrative model only. Not historical results, promised returns or guarantees.

Figures are before any tax, transfer fees or payout costs on your side, and after QwestLab’s estimated operating cost, which is not finalised. Utilisation, acceptance rates, camera lifespan and payment timing can all differ materially from the assumption you chose, and the modelled month a position is covered can move by many months or never arrive. A unit participates only for the operating life of the original 20 cameras.

Read this next to the figures above

  • These are projections from stated assumptions, not historical performance. No camera has been deployed and nothing has been earned.
  • Utilisation is the whole risk. Cameras can sit unworn, and whether any minimum number of hours is guaranteed is unconfirmed.
  • The $1.25 operating estimate is not finalised. A higher real cost reduces every figure here.
  • A unit’s participation ends with the operating life of the original 20 cameras, and lost or retired cameras are not replaced.

The full risk register, the assumption model and the terms live on the Signal 01 site, which opens if this temperature check says there is appetite.

Temperature check

Would you actually take part?

Signal 01 is not open, and this is not a sale. Before QwestLab builds any of it, the Saints get asked a plain question: at $30 a unit, how many would you want, from one to 20? Marking a number reserves nothing, charges nothing and commits you to nothing. It tells us whether there is anything here worth doing.

At the published price

$30

1 × $30

Nothing is charged, nothing is reserved, and no wallet is involved.

Unit price
$30
Units in the pod
280
Funds
20 cameras · $8,400 in total

What happens with your answer. It goes to QwestLab and nowhere else. If enough people want it, the terms get a legal pass and the community hears first. If they do not, Signal 01 does not happen and nothing is lost.

What it is not. Not a payment, not a whitelist spot, not a queue position and not a promise from either side. No wallet is connected and nothing here touches a chain.

Figures shown are the published structure, not a result: 280 units at $30 would fund 20 cameras and the first month of contributor float. Every projection of what those cameras might earn lives on the full Signal 01 site, with its risks.