ProgrammableMarketing/humanarchive.ai (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 6
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 3
Generated Nov 9, 2026
  1. Programmable
  2. Playbooks
  3. Human Archive
Human Archive logo

growth playbook · Nov 9, 2026

humanarchive.ai (opens in a new tab)

Human Archive growth playbook

Diagnosis

Human Archive raised an $8.2 million seed, was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day", and is scaling a large collection machine: 100+ full-time staff, thousands of rigs, annotators in the thousands and a new San Francisco study. The collection side is moving fast; the selling side and the contributor experience are not keeping up.

The website is seven sentences long with no catalogue, no sitemap and no description, so labs can't find or shortlist the data without a form. The only public sample is RGB-only, while the pitch is depth, touch and motion. The contributor app has 7 ratings averaging 3.3, and the reviews complain about failed uploads, pending reviews and payouts.

This playbook starts with a public, indexable catalogue, a contributor page and clear payout rules. It then adds a multi-sensor sample, buyer datasheets, resumable uploads and review tracking, and finishes with bets on published post-training results, a U.S. services-for-data pilot and annotation as a service.

Growth Score

6/10

Biggest Opportunity

Labs are searching for exactly the multi-sensor egocentric data Human Archive collects. Public specs, a multi-sensor sample and published results would turn that demand into inbound requests instead of warm introductions.

Biggest Risk

Consent and privacy scrutiny around collecting egocentric data in India: large home-services platforms declined to take part, and the IT ministry is reportedly reviewing the sector. Meanwhile, contributor reviews that call the app a scam can undercut both supply and buyer trust.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Put the OTS catalogue on humanarchive.ai: a public spec page for every collection, so labs can find and shortlist the data without filling in a form

    Write the HA-Ego spec sheet from the numbers already on the public sample listing and publish it at /catalog/ha-ego.

    High impact First result in 1 month
  2. 2

    Add a contributor path to the homepage: "Get paid for everyday tasks", with the App Store link and how pay works

    Add an App Store badge and a one-line contributor pitch to the homepage.

    High impact First result in 2 weeks
  3. 3

    Publish review times, the minimum withdrawal and payout methods in Help, before contributors guess

    Agree the review-time target the annotation team can meet today, and publish it.

    High impact First result in 2 weeks

02

Your idea portfolio

23 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 7 ideas are marked Unexpected: plays you are unlikely to have tried.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 3Big bets · 9Fill-ins · 6Reconsider · 5
Effort →

Showing 23 of 23 ideas

This week 6

Quick wins with a clear first action.

  • 01

    The homepage offers an "OTS catalog", but the only way to see it is to leave a name and email and wait for "our researcher dashboard and samples". Human Archive's strongest public proof (500 hours from 4,651 people, 259 task categories, 45 environment types) appears only on third-party dataset listings, never on its own site. Search engines can't find anything either: there is no sitemap, no robots.txt and no meta description, and a search for site:humanarchive.ai returns only unrelated "human archive" projects. Meanwhile, Digital Divide Data publishes its frame rate, pose data, hand skeleton and LeRobot v2 format openly. A researcher comparing vendors can shortlist DDD without a call, but not Human Archive.

    High impactMedium effort First result in 1 month CEO

    First action

    Write the HA-Ego spec sheet from the numbers already on the public sample listing and publish it at /catalog/ha-ego.

    How to measure it

    Catalogue sample requests from labs

    Now
    –
    Target
    15 a month
    By
    60 days

    Steps

    1. 1

      Draft spec sheets

      One per collection: hours, contributors, tasks, environments, countries, sensors, sample rates, formats, licence.

      CEO 3 days

    2. 2

      Add the consent summary

      How contributors consent, how faces are blurred and what buyers receive.

      Co-founder, Operations 2 days

    3. 3

      Build the pages

      A /catalog index plus one page per collection, linked from the homepage.

      Member of Technical Staff 3 days

    4. 4

      Make it indexable

      sitemap.xml, robots.txt, meta descriptions and schema.org Dataset markup; submit to Search Console.

      Member of Technical Staff 1 day

    5. 5

      Link the samples

      Each page links to its gated sample and explains how access requests work.

      CTO 1 day

    6. 6

      Sharpen the form

      Ask for modality, hours, task types and deadline, and route answers to the CEO.

      CEO 1 day

    7. 7

      Measure

      Search impressions, catalogue views and qualified requests each week.

      CEO weekly

    First result
    1 month
    Cost
    Staff time only
    Owner
    CEO

    Smallest test

    Publish the HA-Ego spec page with a sitemap, and send that link instead of the form in the next ten lab conversations.

    Why it works

    Build /catalog with one page per collection: egocentric RGB video, the multi-sensor rig (depth, tactile gloves, motion capture, wrist cameras) and Voice Exchange conversations. Each page lists hours, unique contributors, task categories, environment types, countries, sensors and sample rates, delivery formats, licence terms, a short consent and anonymization summary, and a link to the matching sample. Keep pricing and the researcher dashboard behind the form, but make the form ask useful questions (modality, hours, task types, deadline). Add a sitemap, robots.txt, meta descriptions and schema.org Dataset markup so search engines and dataset search can index each page. Link the catalogue from the homepage and from every launch post.

    Lab buyers shortlist on specs before they talk to a vendor. Public spec pages turn anonymous research visits into qualified requests, give the CEO one link to send instead of a deck, and make the most-requested company from YC W26 findable by researchers searching for egocentric data.

    Ready to use

    ChecklistCatalogue page launch list

    [ ] /catalog index linked from the homepage
    [ ] One page per collection: hours, unique contributors, task categories, environments, countries
    [ ] Sensors, sample rates and delivery formats
    [ ] Licence terms and consent/anonymization summary
    [ ] Link to the gated sample and how access works
    [ ] sitemap.xml and robots.txt
    [ ] Meta description and share image on every page
    [ ] schema.org Dataset markup
    [ ] Request form: modality, hours, tasks, deadline

    Risks and how to handle them

    • Rivals copy the spec format

      Specs are table stakes; the hours and sensor mix are the moat.

    • Publishing too much detail

      Publish counts and formats; keep pricing and customer-specific collections private.

    Evidence

    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
    • Your websiteContact page: "Get in touch" and "Drop your details and we'll send over our researcher dashboard and samples." The only fields are name, email and notes.Source
    • Dataset listingHA-Ego-Samples, Human Archive's public sample: "500 hours of footage from 4,651 unique individuals", 9,343 clips across 259 task categories and 45 environment types; RGB video only (MP4); Apache-2.0 licence; access by request on Hugging Face; 163 downloads shown on the listing.Source
    • Your websiteCrawl check (November 9, 2026): /sitemap.xml and /robots.txt both return 404, and a search for site:humanarchive.ai returned no pages from the domain, only unrelated projects with "human archive" in their names. The domain was registered in October 2025.Source
    • Competitor researchDigital Divide Data's egocentric dataset page publishes its specs: rectified stereo video above 1080p at 60 fps, per-frame 6DoF head pose, a 21 to 26-joint hand skeleton per hand, IMU, delivery in HDF5 and LeRobot v2, and the capture headsets used, with "Access Dataset" buttons.Source
  • 02

    Much of Human Archive's reach comes from links shared on X and in press coverage, and the homepage itself points to an X post. Yet the homepage has no meta description, no social share image and no favicon, so shared links show as bare text and search results fall back to scraped copy. The page is already fast (mobile score 99, and real visitors pass Core Web Vitals), so polish is the only gap.

    Medium impactLow effort First result in 1 week Member of Technical Staff

    First action

    Write the homepage meta description and upload a 1200x630 share image.

    How to measure it

    Homepage search click-through

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Write the description

      Positioning plus one proof number, under 160 characters.

      CEO 1 hour

    2. 2

      Make the card

      1200x630 image with the rig and the headline numbers.

      Co-founder, Hardware half a day

    3. 3

      Add tags and favicon

      Open Graph, X card and favicon on every page.

      Member of Technical Staff 2 hours

    4. 4

      Fix link text

      Replace "click here" with descriptive labels.

      Member of Technical Staff 1 hour

    5. 5

      Add a main landmark

      Clears the one accessibility flag.

      Member of Technical Staff 30 minutes

    6. 6

      Check previews

      Test on X, LinkedIn, Slack and iMessage.

      Member of Technical Staff 1 hour

    First result
    1 week
    Cost
    $0
    Owner
    Member of Technical Staff

    Smallest test

    Ship the description and card today and repost the launch link.

    Why it works

    Add a meta description built around the positioning line and one proof number, an Open Graph and X card image showing the capture rig with "500 hours from 4,651 people", and a favicon. Replace "click here" links with descriptive text ("See the catalogue", "Book 30 minutes") and wrap the content in a main landmark. Do the same on the contact, help and legal pages.

    A clear preview card raises click-through on every share at no ongoing cost, and descriptive link text helps both search engines and screen readers.

    Risks and how to handle them

    • Platforms cache old previews

      Refresh the cache with each platform's card validator.

    Evidence

    • PageSpeedGoogle PageSpeed (November 9, 2026): mobile performance 99, desktop 83, and real visitors pass Core Web Vitals (LCP 795 ms, INP 74 ms, CLS 0). SEO score 82: "Document does not have a meta description" and "Links do not have descriptive text". The homepage also has no social share image or favicon, and accessibility flags "Document does not have a main landmark".Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
  • 03

    The iPhone app's subtitle is "Get paid for everyday tasks", and the stated plan is a platform that lets anyone take part and earn. But humanarchive.ai never mentions or links the app, so anyone who hears about Human Archive and wants to contribute hits a page written only for AI labs. Mindrift, a rival, has a dedicated page that states its rate estimate and payout schedule up front.

    High impactLow effort First result in 2 weeks Co founder

    First action

    Add an App Store badge and a one-line contributor pitch to the homepage.

    How to measure it

    Installs via the contributor page

    Now
    –
    Target
    1,000 a month
    By
    90 days

    Steps

    1. 1

      Write the page

      Tasks, what's recorded, pay, eligibility and timing.

      Co-founder, Operations 1 day

    2. 2

      Add the badge

      App Store badge, smart app banner and campaign link.

      Member of Technical Staff 2 hours

    3. 3

      Link it

      From the homepage, Help and the rules pages.

      Member of Technical Staff 1 hour

    4. 4

      Add an FAQ

      Reuse the payout answers from Help (idea 4).

      Co-founder, Operations half a day

    5. 5

      Promote

      Pin the page in launch posts and partner channels.

      CEO ongoing

    6. 6

      Track

      Installs and first approved clip, by source.

      Member of Technical Staff weekly

    First result
    2 weeks
    Cost
    $0
    Owner
    Co founder, Operations

    Smallest test

    Add the badge to the homepage this week and count installs from the campaign link.

    Why it works

    Add a "Contribute" link and a /contribute page covering the tasks (video and voice), what is and isn't recorded, how pay works (payout shown on each task, paid after approval), eligibility (18+, regions where voice isn't offered, iPhone only), review and payout timing, and an App Store badge with a smart app banner. Track installs from the page with an App Store campaign link.

    Contributors decide in seconds whether an earning app is legitimate. A clear page on the company's own domain answers that question and turns press and word-of-mouth traffic into installs.

    Risks and how to handle them

    • More low-quality submissions

      Show the common rejection reasons on the page before install.

    Evidence

    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
    • App StoreHuman Archive iPhone app, "Get paid for everyday tasks": 3.3 out of 5 from 7 ratings; last update 1.2.0 on August 5. Reviews say uploads often fail at 50% or 80% and force re-recording (asking for uploads that run in the background); that later submissions stayed pending after 20 English script readings; and, in another, that reviews stop near the minimum withdrawal threshold. One calls it a "Useful way to make money". No Android app was found.Source
    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • Competitor researchMindrift's egocentric contributor page, "Get paid to record what you do every day", states an estimated $10 an hour in the US, pay per accepted task, and bi-weekly withdrawals through Payoneer up front.Source
  • 04

    Help and Terms say contributors are paid only after "final approval" and that approvals can be reversed, but no page gives a review time, a minimum withdrawal or the payout methods (the privacy policy is the only place Stripe and Circle are named). That silence is exactly what the negative App Store reviews fill: submissions "stayed pending", and reviews allegedly stop near the withdrawal threshold. Mindrift states its payout cadence and method up front.

    High impactLow effort First result in 2 weeks Co founder

    First action

    Agree the review-time target the annotation team can meet today, and publish it.

    How to measure it

    First-withdrawal rate among new contributors

    Now
    –
    Target
    +25%
    By
    90 days

    Steps

    1. 1

      Measure today

      Median and 90th-percentile time from submission to decision.

      Strategy & Project Lead (Data Operations) 1 day

    2. 2

      Set the target

      A window the team can meet 90% of the time.

      Co-founder, Operations 1 day

    3. 3

      Write the section

      Review time, minimum, methods, reversals, appeals.

      Co-founder, Operations 1 day

    4. 4

      Put it in the app

      Link from the balance screen.

      Member of Technical Staff 2 days

    5. 5

      Publish the score

      Monthly share of submissions decided on time.

      Strategy & Project Lead (Data Operations) monthly

    6. 6

      Review complaints

      Tag support email by payout topic.

      Co-founder, Operations weekly

    First result
    2 weeks
    Cost
    $0
    Owner
    Co founder, Operations

    Smallest test

    Add three answers to Help this week: review time, minimum withdrawal, payout methods.

    Why it works

    Add a "Getting paid" section to Help and the app: a target review window (for example, 90% of submissions decided within 5 days), the minimum withdrawal and why it exists, the payout methods by country, when and why approvals are reversed, and how to appeal. Publish monthly the share of submissions decided within the target.

    Clear payout rules prevent "is this a scam?" doubts and the reviews they produce, and they cut support email to [email protected].

    Risks and how to handle them

    • Missing the published target

      Start with a window you already meet, then tighten it.

    Evidence

    • Your websiteHelp, Terms and Privacy pages: approved submissions earn the payout shown on each task, and "No payout is earned until a submission has received final approval under our review process"; approvals can be reversed. None of the pages states review times, a minimum withdrawal or contributor payout methods (the privacy policy names Stripe and Circle as payout processors). Voice tasks are not available in Illinois, Texas, Washington or Oregon.Source
    • App StoreHuman Archive iPhone app, "Get paid for everyday tasks": 3.3 out of 5 from 7 ratings; last update 1.2.0 on August 5. Reviews say uploads often fail at 50% or 80% and force re-recording (asking for uploads that run in the background); that later submissions stayed pending after 20 English script readings; and, in another, that reviews stop near the minimum withdrawal threshold. One calls it a "Useful way to make money". No Android app was found.Source
    • Competitor researchMindrift's egocentric contributor page, "Get paid to record what you do every day", states an estimated $10 an hour in the US, pay per accepted task, and bi-weekly withdrawals through Payoneer up front.Source
  • 05

    Human Archive posted a part-time Research Study Coordinator role on November 3 to run a study in its SF office and "Source and schedule 1000s of unique data collectors". The site has nowhere for a would-be participant to land: no study page, no booking link and no eligibility rules.

    Medium impactLow effort First result in 2 weeks Research Study Coordinator

    First action

    Publish the page with a booking calendar for next week's sessions.

    How to measure it

    SF sessions completed

    Now
    –
    Target
    200 a week
    By
    60 days

    Steps

    1. 1

      Define the session

      Length, pay, what's recorded, eligibility.

      Research Study Coordinator 1 day

    2. 2

      Build the page

      Booking calendar with daily capacity and reminders.

      Member of Technical Staff 2 days

    3. 3

      Consent preview

      Show the consent text before booking.

      Co-founder, Operations 1 day

    4. 4

      Seed channels

      Campus job boards and local community boards.

      Research Study Coordinator ongoing

    5. 5

      Add referrals

      A "bring a friend" link after each session.

      Research Study Coordinator 1 day

    6. 6

      Track

      Sign-ups, show-ups and approved data by source.

      Research Study Coordinator weekly

    First result
    2 weeks
    Cost
    Participant pay plus staff time
    Owner
    Research Study Coordinator

    Smallest test

    Open 50 slots on the page and see how fast they fill.

    Why it works

    Publish /sf-study with what a session involves, how long it takes, the pay per session, the location, eligibility (18+), what is recorded and how consent works, and a self-serve booking calendar with daily capacity. Add a "bring a friend" link with the same booking flow. Post it on campus job boards and local community boards, and track sign-ups and show-ups by source.

    Self-serve booking turns one coordinator into a pipeline that scales to thousands, and the show-up data shows which channels deliver reliable participants.

    Risks and how to handle them

    • No-shows

      Text reminders and a waitlist that fills cancellations.

    Evidence

    • HiringResearch Study Coordinator (Operations, part-time, San Francisco, posted November 3, 2026): "$25-50/hour" for 20 hours a week to run a study in the SF office, set up capture environments and "Source and schedule 1000s of unique data collectors".Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
  • 06

    The app has just 7 ratings, averaging 3.3. Of the four visible reviews, three are about failed uploads, pending reviews or payouts, and two call the company a scam. With so few ratings, every new one moves the average, and prospective contributors read these reviews before installing.

    Medium impactLow effort First result in 1 month Co founder

    First action

    Reply to the three negative reviews this week.

    How to measure it

    Average rating and rating count

    Now
    –
    Target
    4.3+ from 50+ ratings
    By
    90 days

    Steps

    1. 1

      Reply now

      Answer each negative review with facts and a fix.

      Co-founder, Operations 1 hour

    2. 2

      Pick the moment

      First successful withdrawal.

      Co-founder, Operations 1 day

    3. 3

      Add the prompt

      Apple's native review request at that event.

      Member of Technical Staff 1 day

    4. 4

      Exclude bad moments

      Never after a rejection or failed upload.

      Member of Technical Staff half a day

    5. 5

      Ship

      In the next app release.

      Member of Technical Staff 1 week

    6. 6

      Monitor

      Rating, count and new review themes.

      Co-founder, Operations weekly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder, Operations

    Smallest test

    Reply to the existing reviews and watch for updated ratings.

    Why it works

    Reply publicly to each review with the fix or the facts (review times, payout rules from idea 4). Add Apple's in-app rating prompt, triggered only right after a contributor's first successful withdrawal, never after a rejection. Track the rating weekly.

    Contributors who were just paid are the most credible answer to a scam claim, and Apple limits how often the prompt can appear, so timing it to the best moment matters.

    Risks and how to handle them

    • Ratings rise while problems remain

      Pair with ideas 4, 7 and 8 so the causes are fixed too.

    Builds on

    Evidence

    • App StoreHuman Archive iPhone app, "Get paid for everyday tasks": 3.3 out of 5 from 7 ratings; last update 1.2.0 on August 5. Reviews say uploads often fail at 50% or 80% and force re-recording (asking for uploads that run in the background); that later submissions stayed pending after 20 English script readings; and, in another, that reviews stop near the minimum withdrawal threshold. One calls it a "Useful way to make money". No Android app was found.Source
    • Your websiteHelp, Terms and Privacy pages: approved submissions earn the payout shown on each task, and "No payout is earned until a submission has received final approval under our review process"; approvals can be reversed. None of the pages states review times, a minimum withdrawal or contributor payout methods (the privacy policy names Stripe and Circle as payout processors). Voice tasks are not available in Illinois, Texas, Washington or Oregon.Source

This quarter 11

The core plays for the next three months.

  • 07

    An App Store review says uploads "often fail at 50% or 80%", forcing re-recording and wasting the contributor's mobile data, and asks for uploads that keep going in the background while recording continues. Video tasks also carry motion data, so files are large, and collection runs across many cities and countries on mixed networks.

    High impactMedium effort First result in 6 weeks Member of Technical Staff

    First action

    Instrument the current uploader to log the failure rate and the percentage reached when uploads fail.

    How to measure it

    Uploads completed on the first attempt

    Now
    –
    Target
    98%
    By
    1 quarter

    Steps

    1. 1

      Instrument

      Failure rate, percentage reached, network type, file size.

      Member of Technical Staff 2 days

    2. 2

      Chunk uploads

      Resumable multipart uploads to storage.

      Member of Technical Staff 2 weeks

    3. 3

      Background transfer

      Uploads continue while the app is closed or recording.

      Member of Technical Staff 1 week

    4. 4

      Queue and Wi-Fi option

      Retry with backoff; optional Wi-Fi only.

      Member of Technical Staff 1 week

    5. 5

      Show status

      Per-clip state in the app.

      Member of Technical Staff 3 days

    6. 6

      Roll out

      Staged release, comparing failure rates.

      Member of Technical Staff 2 weeks

    First result
    6 weeks
    Cost
    Engineering time
    Owner
    Member of Technical Staff

    Smallest test

    Add retry-from-last-chunk for clips over a size threshold and compare failure rates for two weeks.

    Why it works

    Switch to chunked, resumable uploads that use background transfer, so a dropped connection resumes from the last chunk instead of restarting. Queue clips with an optional "Wi-Fi only" setting, keep each clip on the device until the server confirms it, and show per-clip status (queued, uploading, received, in review). Log failure points to find the worst networks and file sizes.

    Every failed upload costs a contributor's time and Human Archive's data, and it is the most concrete complaint in the reviews. Fixing it lifts both supply and the rating.

    Risks and how to handle them

    • Device storage fills while clips wait

      Warn at a storage threshold and pause new recording.

    Evidence

    • App StoreHuman Archive iPhone app, "Get paid for everyday tasks": 3.3 out of 5 from 7 ratings; last update 1.2.0 on August 5. Reviews say uploads often fail at 50% or 80% and force re-recording (asking for uploads that run in the background); that later submissions stayed pending after 20 English script readings; and, in another, that reviews stop near the minimum withdrawal threshold. One calls it a "Useful way to make money". No Android app was found.Source
    • Your websiteContributor rules: video tasks record first-person video and motion (IMU) data with no audio. Voice Exchange offers solo speaking or script reading, a one-time reading assessment per language, and live two-party conversations between contributors. Contributors who join through a distributor network are paid by that network.Source
    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
  • 08

    The worst reviews describe silence: submissions that "stayed pending", and suspicion that reviews stop near the withdrawal threshold. Human Archive is scaling annotation "from hundreds to thousands of annotators" against throughput and quality targets, so it already measures the review queue internally; contributors just can't see it.

    High impactMedium effort First result in 1 month Strategy & Project Lead

    First action

    Add review age to the annotation dashboard and list submissions older than 7 days.

    How to measure it

    Share of submissions decided within the published window

    Now
    –
    Target
    95%
    By
    1 quarter

    Steps

    1. 1

      Measure aging

      Count and age of pending contributor submissions.

      Strategy & Project Lead (Data Operations) 2 days

    2. 2

      Clear the backlog

      Oldest first, with a one-off push.

      Strategy & Project Lead (Data Operations) 2 weeks

    3. 3

      Add states

      Status and expected date per submission in the app.

      Member of Technical Staff 2 weeks

    4. 4

      Notify

      Push notification on every decision.

      Member of Technical Staff 3 days

    5. 5

      Alert on aging

      Internal alert when items pass the target.

      Member of Technical Staff 2 days

    6. 6

      Review monthly

      On-time rate next to throughput and quality.

      Strategy & Project Lead (Data Operations) monthly

    First result
    1 month
    Cost
    Engineering and reviewer time
    Owner
    Strategy & Project Lead (Data Operations)

    Smallest test

    Clear every submission older than 7 days and add the status label first.

    Why it works

    Give every submission a visible state (received, in review, approved, rejected with reason) and an expected decision date based on current queue speed. Set an internal service level for contributor reviews (oldest-first, with alerts when items age past the target), and send a push notification when a decision is made. Put contributor review time on the annotation team's dashboard next to throughput and quality.

    Uncertainty, not rejection, turns contributors into critics. A visible queue with a date makes a wait feel fair, and an aging alert stops submissions from being forgotten.

    Risks and how to handle them

    • Dates slip in busy weeks

      Show a range, and update it live from queue speed.

    Builds on

    Evidence

    • App StoreHuman Archive iPhone app, "Get paid for everyday tasks": 3.3 out of 5 from 7 ratings; last update 1.2.0 on August 5. Reviews say uploads often fail at 50% or 80% and force re-recording (asking for uploads that run in the background); that later submissions stayed pending after 20 English script readings; and, in another, that reviews stop near the minimum withdrawal threshold. One calls it a "Useful way to make money". No Android app was found.Source
    • Your websiteHelp, Terms and Privacy pages: approved submissions earn the payout shown on each task, and "No payout is earned until a submission has received final approval under our review process"; approvals can be reversed. None of the pages states review times, a minimum withdrawal or contributor payout methods (the privacy policy names Stripe and Circle as payout processors). Voice tasks are not available in Illinois, Texas, Washington or Oregon.Source
    • HiringAnnotation and ingest: a Strategy & Project Lead (Data Operations, Bangalore) to "scale from hundreds to thousands of annotators" against throughput, quality and utilisation targets; a Member of Technical Staff for the internal platform, with "ingest stations moving terabytes" and "annotators working through millions of items"; and a Dark Store & Inventory Operations Lead for "an internal mini dark store that will manage cameras, power banks, charging stations".Source
  • 09

    Human Archive's case to labs is synchronized RGB-D, tactile force, motion capture and wrist cameras, which its lead investor says no one else has combined at scale. Yet the only public sample, HA-Ego-Samples, is RGB video in MP4. A researcher who downloads it sees the same kind of data many egocentric vendors sell, and never touches the part that's unique.

    High impactMedium effort First result in 2 months CTO

    First action

    Pick five tasks already captured with the full rig, and check synchronization and anonymization.

    How to measure it

    Multi-sensor sample requests to pilot

    Now
    –
    Target
    5 pilots
    By
    2 quarters

    Steps

    1. 1

      Select clips

      Five tasks, all streams present and in sync.

      CTO 1 week

    2. 2

      Anonymize and check

      Faces blurred; consent covers public release.

      Co-founder, Operations 1 week

    3. 3

      Document

      Schema, calibration, sample rates, coordinate frames.

      Research Engineer 1 week

    4. 4

      Build the viewer

      A notebook that plays all streams together.

      Machine Learning Engineer 1 week

    5. 5

      Publish

      Gated release, linked from the catalogue.

      CTO 2 days

    6. 6

      Follow up

      The CEO contacts every lab that requests access.

      CEO ongoing

    First result
    2 months
    Cost
    Staff time and hosting
    Owner
    CTO

    Smallest test

    Share a two-hour multi-sensor slice privately with three labs and record their reactions.

    Why it works

    Publish a gated multi-sensor sample: 10 to 20 hours across a handful of tasks, with every stream time-synced, plus calibration files, a schema document and a notebook that plays video, depth, hand pose and glove force side by side. Release it with a short technical write-up and link it from the catalogue (idea 1).

    The sample shapes a buyer's idea of what the company sells. Letting researchers handle the synchronized streams shows the differentiator directly and leads to higher-value requests.

    Risks and how to handle them

    • Rivals study the format

      Release a small slice; the advantage is scale and hardware.

    Builds on

    Evidence

    • Dataset listingHA-Ego-Samples, Human Archive's public sample: "500 hours of footage from 4,651 unique individuals", 9,343 clips across 259 task categories and 45 environment types; RGB video only (MP4); Apache-2.0 licence; access by request on Hugging Face; 163 downloads shown on the listing.Source
    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
  • 10

    Digital Divide Data advertises delivery in HDF5 and LeRobot v2, so a lab can load it into an existing training pipeline on day one. Human Archive's own Machine Learning Engineer works on "VLA training, robotics post-training", so the internal loaders already exist; buyers just can't see or use them.

    Medium impactMedium effort First result in 1 month Machine Learning Engineer

    First action

    Ask the last five labs that requested samples which format they train on.

    How to measure it

    Buyer time to first batch

    Now
    –
    Target
    Under 1 day
    By
    1 quarter

    Steps

    1. 1

      Survey buyers

      Formats and frameworks in use.

      CEO 1 week

    2. 2

      Write converters

      HDF5 and LeRobot first.

      Machine Learning Engineer 2 weeks

    3. 3

      Map sensors

      Depth, glove force and pose in the schema.

      Research Engineer 1 week

    4. 4

      Write the quickstart

      Load, visualize and train in one page.

      Machine Learning Engineer 3 days

    5. 5

      List on the catalogue

      Formats per collection.

      Member of Technical Staff 1 day

    6. 6

      Measure

      Time to first batch in each pilot.

      CTO per pilot

    First result
    1 month
    Cost
    Engineering time
    Owner
    Machine Learning Engineer

    Smallest test

    Convert the public sample to one format and send it to two labs.

    Why it works

    Publish converters and a quickstart for the formats labs ask for most, starting with HDF5 and LeRobot, covering the multi-sensor streams (depth, glove force, hand pose). State the supported formats on every catalogue page and measure how long a new buyer takes to load a first training batch.

    Integration effort is a hidden cost in every data purchase. Removing it shortens pilots and makes Human Archive the easier vendor to say yes to.

    Risks and how to handle them

    • Formats change

      Version the converters and pin schema versions.

    Evidence

    • Competitor researchDigital Divide Data's egocentric dataset page publishes its specs: rectified stereo video above 1080p at 60 fps, per-frame 6DoF head pose, a 21 to 26-joint hand skeleton per hand, IMU, delivery in HDF5 and LeRobot v2, and the capture headsets used, with "Access Dataset" buttons.Source
    • HiringHardware and models: firmware, mechanical and electrical engineers for "our wearable capture hardware" (China), a Camera Systems and Calibration Engineer to "own the entire optics and calibration stack", a Head of Engineering in San Francisco, and a Machine Learning Engineer working on "multimodal perception, VLA training, robotics post-training".Source
  • 11

    Consent is the question hanging over egocentric data collection: large home-services platforms declined to take part, and the IT ministry is reportedly reviewing consent practices across the sector. Human Archive says its contracts comply with India's DPDP Act, that data is anonymized and that faces are blurred, and its investor describes QA and anonymization pipelines. None of that is written down where a buyer's legal team can read it.

    High impactMedium effort First result in 2 months Co founder

    First action

    Draft the datasheet for HA-Ego from the existing consent notices and contracts.

    How to measure it

    Median days to signature

    Now
    –
    Target
    -30%
    By
    2 quarters

    Steps

    1. 1

      Collect the facts

      Consent notices, contracts, anonymization steps per collection.

      Co-founder, Operations 1 week

    2. 2

      Draft the template

      Provenance, consent, anonymization, licence, retention.

      Co-founder, Operations 3 days

    3. 3

      Legal review

      Outside counsel in India and the U.S.

      CEO 2 weeks

    4. 4

      Publish

      On each catalogue page.

      Member of Technical Staff 1 day

    5. 5

      Use in deals

      Attach to every proposal.

      CEO ongoing

    6. 6

      Update

      With each new collection.

      Co-founder, Operations per release

    First result
    2 months
    Cost
    Legal review fees
    Owner
    Co founder, Operations

    Smallest test

    Send a draft datasheet to one lab currently in review and ask whether it answers their questions.

    Why it works

    Write a standard datasheet for each collection: where and when it was captured, how contributors and site owners consented (with the notice text), what was blurred or removed, retention and deletion rules, licence and sublicence terms, and an audit contact. Offer it on the catalogue page and in every proposal, and review it with outside counsel before publishing.

    Legal and compliance review is often the slowest step in a data purchase. A ready datasheet shortens it and sets Human Archive apart on trust, which is where it faces the most scrutiny.

    Risks and how to handle them

    • Publishing invites critics

      Publish only what counsel has checked, and fix gaps before publishing.

    Builds on

    Evidence

    • PressTechCrunch also reported that large Indian home-services platforms declined to take part, and that India's IT ministry is reportedly examining the consent practices of startups that collect egocentric data through home-service workers. Human Archive says its contracts comply with India's DPDP Act, that all data is anonymized and that faces are blurred.Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
  • 12

    The only buyer form asks for name, email and notes. Human Archive can collect across homes, hotels, restaurants, agriculture, manufacturing, construction and retail in 30+ countries, but the form gives a lab no way to say what it needs, so every enquiry starts with a scoping call. Claru, by contrast, leads with proof numbers and fast custom delivery.

    High impactLow effort First result in 1 month CEO

    First action

    Turn the notes field into five questions this week.

    How to measure it

    Briefs that reach a proposal

    Now
    –
    Target
    50%
    By
    1 quarter

    Steps

    1. 1

      Write the questions

      Modality, environments, tasks, hours, countries, deadline, format.

      CEO 1 day

    2. 2

      Build the form

      Replace the contact page form.

      Member of Technical Staff 2 days

    3. 3

      Auto-reply

      Feasibility note and matching sample.

      Member of Technical Staff 2 days

    4. 4

      Capacity check

      Operations confirms sites and rigs.

      Co-founder, Operations per brief

    5. 5

      Send proposals

      Within two working days.

      CEO per brief

    6. 6

      Review

      Brief-to-proposal and proposal-to-pilot rates.

      CEO monthly

    First result
    1 month
    Cost
    $0
    Owner
    CEO

    Smallest test

    Add three optional fields and compare reply quality.

    Why it works

    Build a short brief: modality (video, multi-sensor, voice), environments, task list, hours, countries, deadline, delivery format and budget band. Answer automatically with a feasibility note and a matching sample, and have the CEO send a scoped proposal within two working days.

    Structured briefs qualify buyers, shorten the first call and let the operations team estimate capacity before the deal is signed.

    Risks and how to handle them

    • Longer form, fewer enquiries

      Keep only name and email required; the rest optional.

    Evidence

    • Your websiteContact page: "Get in touch" and "Drop your details and we'll send over our researcher dashboard and samples." The only fields are name, email and notes.Source
    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
    • Competitor researchClaru's homepage, "Training data for physical AI.", puts proof numbers up front: "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships". It has no public catalogue or pricing.Source
  • 13

    Human Archive adds data every day: thousands of rigs in the field, ingest stations "moving terabytes" and annotators "working through millions of items". A lab that licensed data last quarter has no way of knowing what has been added since, so repeat purchases depend on someone remembering to ask.

    Medium impactLow effort First result in 1 month CTO

    First action

    Pull last month's added hours by environment and task from the ingest platform.

    How to measure it

    Buyers making a second purchase

    Now
    –
    Target
    40%
    By
    2 quarters

    Steps

    1. 1

      Pull the numbers

      Hours by environment, task and sensor.

      Member of Technical Staff 1 day

    2. 2

      Write the note

      One page with charts and new samples.

      CTO half a day

    3. 3

      Build the list

      Buyers plus catalogue sign-ups.

      CEO 1 day

    4. 4

      Add refresh pricing

      For existing licences.

      CEO 1 week

    5. 5

      Add voting

      Labs vote on the next tasks.

      CTO 1 day

    6. 6

      Send monthly

      Same day each month.

      CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    CTO

    Smallest test

    Email one release note to current buyers and count replies.

    Why it works

    Publish a monthly release note to buyers and catalogue subscribers: hours added by environment and task, new sensors or countries, quality metrics (rejection rate, sync accuracy) and new samples. Offer existing buyers refresh pricing on the new hours, and invite labs to vote on what to collect next.

    In a data business, freshness drives repeat purchases the way a changelog drives software upgrades. Votes also tell operations where demand is before it commits rigs.

    Risks and how to handle them

    • Revealing capacity to rivals

      Send the detailed note to buyers only; publish a summary.

    Builds on

    Evidence

    • HiringAnnotation and ingest: a Strategy & Project Lead (Data Operations, Bangalore) to "scale from hundreds to thousands of annotators" against throughput, quality and utilisation targets; a Member of Technical Staff for the internal platform, with "ingest stations moving terabytes" and "annotators working through millions of items"; and a Dark Store & Inventory Operations Lead for "an internal mini dark store that will manage cameras, power banks, charging stations".Source
    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
  • 14

    Human Archive is hiring a partnerships intern to "build relationships with the most ambitious, hands-on student organizations in the country" for campus data collection, sponsorships and hiring, and says "This is not a traditional campus ambassador program." Its founders come from university robotics labs, and its public sample is free to use under Apache-2.0.

    Medium impactMedium effort First result in 6 weeks Founder's Office Intern

    First action

    Shortlist ten clubs with workshop access and pitch five.

    How to measure it

    Approved campus hours

    Now
    –
    Target
    500 hours
    By
    1 quarter

    Steps

    1. 1

      Shortlist clubs

      Robotics, maker and engineering clubs with workshops.

      Founder's Office Intern, Partnerships 1 week

    2. 2

      Design the package

      Kit, pay per approved hour, data access, interviews.

      Co-founder, Operations 3 days

    3. 3

      Consent and rules

      Campus permissions and contributor rules.

      Co-founder, Operations 1 week

    4. 4

      Launch five

      Kits delivered, first sessions run.

      Founder's Office Intern, Partnerships 1 month

    5. 5

      Track

      Approved hours, quality and candidates per club.

      Founder's Office Intern, Partnerships weekly

    6. 6

      Scale

      Expand the best-performing format.

      Co-founder, Operations next quarter

    First result
    6 weeks
    Cost
    Kits and contributor pay
    Owner
    Founder's Office Intern, Partnerships

    Smallest test

    Run one club for one month.

    Why it works

    Offer robotics and maker clubs a package: capture rigs, pay for approved hours recorded in their workshops and labs (assembly, soldering, machining, cooking in hostels), early access to samples for their own projects and competitions, and a fast-track interview for members. Start with five clubs, measure approved hours and hires per club, then scale the best format.

    Clubs bring skilled hands, varied environments and a hiring pipeline in one relationship, and members who use the data become its advocates when they join labs.

    Risks and how to handle them

    • Kit loss or damage

      Track kits like the dark store does, with a named club owner.

    Evidence

    • HiringFounder's Office Intern, Partnerships (posted August 12, 2026): "build relationships with the most ambitious, hands-on student organizations in the country" for campus data collection, sponsorships and hiring pipelines; "This is not a traditional campus ambassador program."Source
    • Dataset listingHA-Ego-Samples, Human Archive's public sample: "500 hours of footage from 4,651 unique individuals", 9,343 clips across 259 task categories and 45 environment types; RGB video only (MP4); Apache-2.0 licence; access by request on Hugging Face; 163 downloads shown on the listing.Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
  • 15

    Human Archive works with "500+ industry partners", and its operations managers deal directly with "owners of factories, resorts and restaurants". The Founder's Office is charged with sourcing and structuring site partnerships. Yet there is nothing on the site a site owner can read or forward to a partner before saying yes.

    Medium impactLow effort First result in 1 month Founding Operations Manager

    First action

    List the ten questions site owners ask most often in first meetings.

    How to measure it

    Sites signed via the page or referrals

    Now
    –
    Target
    20 a month
    By
    1 quarter

    Steps

    1. 1

      Collect questions

      From operations managers' first meetings.

      Founding Operations Manager 3 days

    2. 2

      Write the page

      Day plan, privacy, benefits, form.

      Founding Operations Manager 3 days

    3. 3

      Translate

      Hindi first, then other local languages where sites are.

      Founding Operations Manager 1 week

    4. 4

      Publish

      At /sites, linked from the homepage.

      Member of Technical Staff 1 day

    5. 5

      Partner referrals

      A payment for each new site introduced.

      Co-founder, Operations 1 week

    6. 6

      Track

      Leads, signed sites and approved hours per site.

      Founding Operations Manager monthly

    First result
    1 month
    Cost
    Translation and referral payments
    Owner
    Founding Operations Manager

    Smallest test

    Send the draft page to ten prospective sites and compare sign-up rates.

    Why it works

    Publish /sites in English and Hindi: what a collection day looks like, team size, how workers and guests are kept out of frame or blurred, insurance and equipment, payment or service benefits, and a short form for site details (type, size, city, available hours). Give existing partners a referral payment for introducing other sites.

    Site access is the bottleneck for field collection. A page that answers the owner's first ten questions shortens every site negotiation and lets partners refer each other.

    Risks and how to handle them

    • Unsuitable sites apply

      Ask about site type, hours and size on the form and screen before calling.

    Evidence

    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
  • 16

    The stated plan is a platform that lets anyone take part and earn, and some contributors already arrive through distributor networks that pay them. Rival Claru advertises 34,000+ contributors in 70+ countries. Referral rewards tied to approval bring in more contributors without paying for low-quality sign-ups.

    Medium impactLow effort First result in 1 month Co founder

    First action

    Set the bonus and the approval threshold that unlocks it.

    How to measure it

    Referred contributors with approved clips

    Now
    –
    Target
    30% of new contributors
    By
    1 quarter

    Steps

    1. 1

      Set the rules

      Bonus, approval threshold, monthly cap.

      Co-founder, Operations 2 days

    2. 2

      Build links

      Referral codes and attribution in the app.

      Member of Technical Staff 1 week

    3. 3

      Show status

      Invite progress for the referrer.

      Member of Technical Staff 3 days

    4. 4

      Fraud checks

      Duplicate devices and payout accounts.

      Member of Technical Staff 3 days

    5. 5

      Launch

      To contributors with approved history first.

      Co-founder, Operations 1 day

    6. 6

      Review

      Cost per approved contributor by channel.

      Co-founder, Operations monthly

    First result
    1 month
    Cost
    Referral bonuses
    Owner
    Co founder, Operations

    Smallest test

    Offer referrals to the top 100 contributors for one month.

    Why it works

    Add a referral link in the app. Both people earn a bonus once the new contributor's first few clips are approved, with a monthly cap per referrer. Show the referrer the status of each invite, and pause referrers whose invitees mostly get rejected.

    Contributors who already get paid are the most credible recruiters, and tying the reward to approval protects quality and cost.

    Risks and how to handle them

    • Referral fraud

      Pay only after approval, and check devices and payout accounts.

    Builds on

    Evidence

    • Your websiteContributor rules: video tasks record first-person video and motion (IMU) data with no audio. Voice Exchange offers solo speaking or script reading, a one-time reading assessment per language, and live two-party conversations between contributors. Contributors who join through a distributor network are paid by that network.Source
    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • Competitor researchClaru's homepage, "Training data for physical AI.", puts proof numbers up front: "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships". It has no public catalogue or pricing.Source
  • 17

    The homepage offers "a robot deployment" next to the catalogue but never says what one is. Human Archive already runs the hard parts: field teams that set up sites, capture and upload every day, plus a model team doing "robotics post-training". Without a defined offer, every deployment conversation starts from zero.

    High impactMedium effort First result in 1 quarter Head of Engineering

    First action

    Write the scope from the last deployment conversation and set a price range.

    How to measure it

    Signed deployment pilots

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Define scope

      Site, tasks, weeks, deliverables.

      Head of Engineering 1 week

    2. 2

      Price it

      Range based on site and task count.

      CEO 3 days

    3. 3

      Pick sites

      Two partner sites ready to host.

      Founding Operations Manager 2 weeks

    4. 4

      Publish

      /deployments page with the brief form.

      Member of Technical Staff 2 days

    5. 5

      Sell

      Offer to labs already sampling data.

      CEO ongoing

    6. 6

      Report

      Success-rate report at the end of each pilot.

      Machine Learning Engineer per pilot

    First result
    1 quarter
    Cost
    Field team and engineering time, covered by pilot fees
    Owner
    Head of Engineering

    Smallest test

    Offer the pilot to three labs that already have samples.

    Why it works

    Define a 6 to 8-week pilot: one partner site, the customer's robot, Human Archive's field team capturing human demonstrations of the target tasks, post-training on that data, and a final success-rate report. Publish the scope, deliverables, timeline and a price range on a /deployments page, with the brief from idea 12.

    A fixed scope makes the deployment a product labs can budget for, and every pilot produces task data that can also be added to the catalogue (subject to the customer's terms).

    Risks and how to handle them

    • Robots vary widely

      Start with one or two supported robot types.

    Builds on

    Evidence

    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
    • HiringHardware and models: firmware, mechanical and electrical engineers for "our wearable capture hardware" (China), a Camera Systems and Calibration Engineer to "own the entire optics and calibration stack", a Head of Engineering in San Francisco, and a Machine Learning Engineer working on "multimodal perception, VLA training, robotics post-training".Source

Big bets 3

Larger plays that can change the trajectory.

  • 18

    Human Archive plans to fine-tune models on its own data and test them on robots "to demonstrate data quality", and its Machine Learning Engineer works on VLA training and robotics post-training. Every data vendor claims quality; a published, reproducible result that shows Human Archive's multi-sensor data raising robot success rates would be proof no competitor page currently offers.

    High impactHigh effort First result in 2 quarters CTO

    First action

    Choose the three tasks and the baseline policy.

    How to measure it

    Enquiries referencing the results

    Now
    –
    Target
    20
    By
    2 quarters after publication

    Steps

    1. 1

      Pick tasks

      Three tasks with clear success criteria.

      CTO 1 week

    2. 2

      Set the baseline

      An open policy and a fixed evaluation protocol.

      Machine Learning Engineer 2 weeks

    3. 3

      Train variants

      Without, with RGB, with multi-sensor.

      Machine Learning Engineer 6 weeks

    4. 4

      Evaluate on robots

      Repeated real-world trials.

      Research Engineer 4 weeks

    5. 5

      Write it up

      Report, data slice and protocol.

      CTO 3 weeks

    6. 6

      Distribute

      Venues, catalogue pages and sales follow-up.

      CEO ongoing

    First result
    2 quarters
    Cost
    Compute, robot time and ML staff for two quarters
    Owner
    CTO

    Smallest test

    Run one task internally and share the result privately with two labs.

    Why it works

    Pick three manipulation tasks and one open policy baseline. Train with and without Human Archive data (RGB-only versus multi-sensor), measure success on real robots, and publish a technical report with ablations, the evaluation protocol and the data slice used. Present it at robotics venues and link it from every catalogue page.

    Labs buy data that moves their metrics. A public ablation that separates the value of touch and depth from video alone justifies premium pricing for the multi-sensor collections.

    Risks and how to handle them

    • Small or negative effect

      Publish honestly and use it to steer what to collect next.

    Builds on

    Evidence

    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • HiringHardware and models: firmware, mechanical and electrical engineers for "our wearable capture hardware" (China), a Camera Systems and Calibration Engineer to "own the entire optics and calibration stack", a Head of Engineering in San Francisco, and a Machine Learning Engineer working on "multimodal perception, VLA training, robotics post-training".Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
  • 19

    Human Archive plans early U.S. pilots that offer services such as cleaning or cooking in exchange for data collection, and it has just hired a San Francisco study coordinator to schedule "1000s of unique data collectors". U.S. homes and kitchens are environments labs want, and the booking, consent and capture setup for the SF study can be reused.

    High impactHigh effort First result in 1 quarter Co founder

    First action

    Price one session type and recruit five professionals to wear rigs.

    How to measure it

    U.S. home hours approved

    Now
    –
    Target
    1,000 hours
    By
    2 quarters

    Steps

    1. 1

      Design the offer

      Service, discount and a no-recording option.

      Co-founder, Operations 2 weeks

    2. 2

      Consent and legal

      Household and worker consent; state rules.

      CEO 3 weeks

    3. 3

      Recruit providers

      Trained cleaners and cooks, paid properly.

      Research Study Coordinator 3 weeks

    4. 4

      Open bookings

      On the SF study booking system.

      Member of Technical Staff 1 week

    5. 5

      Run sessions

      Capped pilot with daily quality checks.

      Research Study Coordinator 2 months

    6. 6

      Compare economics

      Cost and quality per hour across sources.

      Co-founder, Operations monthly

    First result
    1 quarter
    Cost
    Service-provider pay, discounts and insurance
    Owner
    Co founder, Operations

    Smallest test

    Twenty sessions in one neighbourhood over two weeks.

    Why it works

    Offer San Francisco households a discounted or free cleaning or cooking session, done by a trained, paid professional wearing the capture rig, with explicit opt-in consent from everyone in the home and a full-price option with no recording. Use the SF study's booking page (idea 5) and consent flow, cap the pilot at a few hundred sessions, and measure cost per approved hour against India and the in-office study.

    It brings in varied, real U.S. home environments at a known cost per hour, and it tests the model the company plans to expand.

    Risks and how to handle them

    • Privacy concerns in homes

      Opt-in only, a no-recording option at full price, and blurring before anything leaves the device.

    Builds on

    Evidence

    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • HiringResearch Study Coordinator (Operations, part-time, San Francisco, posted November 3, 2026): "$25-50/hour" for 20 hours a week to run a study in the SF office, set up capture environments and "Source and schedule 1000s of unique data collectors".Source
  • 20

    Human Archive is building annotation to scale "from hundreds to thousands of annotators", with its own platform for human review over "millions of items". Its investor lists annotation and benchmarking as part of the pipeline. Many robot teams record their own teleoperation or egocentric footage and still need it labelled, and annotation sold on its own is a second revenue line using staff and tools already being hired.

    High impactHigh effort First result in 1 quarter Strategy & Project Lead

    First action

    Ask three labs in conversation whether they have unlabelled footage and what they pay for labelling today.

    How to measure it

    Monthly annotation revenue

    Now
    –
    Target
    First 3 paying customers
    By
    2 quarters

    Steps

    1. 1

      Validate demand

      Three lab conversations about label needs.

      CEO 2 weeks

    2. 2

      Define products

      Label types, accuracy and turnaround.

      Strategy & Project Lead (Data Operations) 2 weeks

    3. 3

      Secure the pipeline

      Customer data isolated in the platform.

      Member of Technical Staff 1 month

    4. 4

      Train a team

      A dedicated pod with QA leads.

      Strategy & Project Lead (Data Operations) 1 month

    5. 5

      Pilot

      One paid pilot with clear quality targets.

      CEO 6 weeks

    6. 6

      Price and scale

      Per labelled hour, with volume tiers.

      CEO next quarter

    First result
    1 quarter
    Cost
    Annotator pod and platform work
    Owner
    Strategy & Project Lead (Data Operations)

    Smallest test

    Label 50 hours of one lab's footage as a paid pilot.

    Why it works

    Offer annotation for customers' own footage (action segmentation, hand-object labels, task success, quality checks) with clear turnaround and accuracy targets. Start with buyers already in conversation, price per labelled hour of footage, and use spare annotator capacity so the core collection pipeline isn't slowed.

    It brings revenue from labs that aren't ready to buy data yet, keeps the annotator team busy, and teaches Human Archive which labels labs value most.

    Risks and how to handle them

    • Distracts from the core dataset

      Run it as a separate pod with its own capacity.

    Evidence

    • HiringAnnotation and ingest: a Strategy & Project Lead (Data Operations, Bangalore) to "scale from hundreds to thousands of annotators" against throughput, quality and utilisation targets; a Member of Technical Staff for the internal platform, with "ingest stations moving terabytes" and "annotators working through millions of items"; and a Dark Store & Inventory Operations Lead for "an internal mini dark store that will manage cameras, power banks, charging stations".Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
    • Competitor researchClaru's homepage, "Training data for physical AI.", puts proof numbers up front: "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships". It has no public catalogue or pricing.Source

Wildcards 3

Unexpected experiments, each with a cheap way to test it.

  • 21

    The app already runs Voice Exchange: solo speaking and script reading, a reading assessment for each language, and live two-party conversations between contributors. The privacy policy says voice data is licensed separately. Yet the homepage never mentions voice, so a whole product line that's already being collected has no storefront.

    Medium impactMedium effort First result in 2 months CEO

    First action

    Count approved voice hours by language and type.

    How to measure it

    Paid voice licences

    Now
    –
    Target
    2
    By
    2 quarters

    Steps

    1. 1

      Inventory

      Approved hours by language and task type.

      Strategy & Project Lead (Data Operations) 3 days

    2. 2

      Check consent

      Confirm licensing terms cover the buyers targeted.

      Co-founder, Operations 1 week

    3. 3

      Build samples

      Short consented clips per language.

      Machine Learning Engineer 1 week

    4. 4

      Catalogue page

      Voice section with specs and samples.

      Member of Technical Staff 3 days

    5. 5

      Outreach

      Ten speech-focused teams.

      CEO 1 month

    6. 6

      Decide

      Invest more or keep it opportunistic.

      CEO end of quarter

    First result
    2 months
    Cost
    Staff time only
    Owner
    CEO

    Smallest test

    Add a voice row to the catalogue and count requests.

    Why it works

    Add a voice collection to the catalogue (idea 1): hours per language, conversation versus read speech, speaker counts, recording conditions, consent terms (audio only, no voiceprints) and a short sample per language. Offer custom languages and topics through the brief (idea 12), and test demand with speech and conversational-AI teams.

    The supply exists and the consent flows are built, so a catalogue page costs little. It may open buyers beyond robotics.

    Risks and how to handle them

    • Distracts from robotics focus

      Cap the effort at one page and ten conversations until a deal closes.

    Builds on

    Evidence

    • Your websiteContributor rules: video tasks record first-person video and motion (IMU) data with no audio. Voice Exchange offers solo speaking or script reading, a one-time reading assessment per language, and live two-party conversations between contributors. Contributors who join through a distributor network are paid by that network.Source
    • Your websiteHelp, Terms and Privacy pages: approved submissions earn the payout shown on each task, and "No payout is earned until a submission has received final approval under our review process"; approvals can be reversed. None of the pages states review times, a minimum withdrawal or contributor payout methods (the privacy policy names Stripe and Circle as payout processors). Voice tasks are not available in Illinois, Texas, Washington or Oregon.Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
  • 22

    Human Archive already collects in manufacturing, construction, agriculture and kitchens, and builds tactile gloves and wrist cameras that capture fine hand control. Most collection today comes from service workers doing everyday tasks. The dexterous, skilled work robots struggle with most is captured less often, and skilled workers expect higher pay.

    Medium impactMedium effort First result in 1 quarter Founding Operations Manager

    First action

    Ask current buyers which skilled tasks they would pay a premium for.

    How to measure it

    Skilled-tier hours approved

    Now
    –
    Target
    300 hours
    By
    2 quarters

    Steps

    1. 1

      Ask buyers

      Which trades and tasks matter most.

      CEO 2 weeks

    2. 2

      Define the tier

      Trades, test tasks and pay rates.

      Founding Operations Manager 1 week

    3. 3

      Recruit

      Trade schools, workshops and partner sites.

      Founding Operations Manager 1 month

    4. 4

      Verify skill

      A paid test task reviewed by an expert.

      Strategy & Project Lead (Data Operations) ongoing

    5. 5

      Capture with the full rig

      Gloves, wrist cameras and depth.

      Founding Operations Manager 2 months

    6. 6

      Package

      A premium catalogue collection.

      CTO 2 weeks

    First result
    1 quarter
    Cost
    Higher contributor pay and rig time
    Owner
    Founding Operations Manager

    Smallest test

    Twenty hours from one trade at one partner workshop.

    Why it works

    Create a higher-paid task tier for verified trades (professional cooks, mechanics, electricians, assembly technicians) recorded with the full glove and wrist-camera rig. Recruit through trade schools, workshops and partner sites, verify skill with a short test task, and offer the resulting collection as a premium catalogue line.

    Skilled-task data is scarcer and worth more to labs working on dexterous manipulation, and a premium tier gives the multi-sensor rig its most valuable use.

    Risks and how to handle them

    • Workplace permissions

      Record only with the employer's written agreement and keep coworkers out of frame.

    Evidence

    • HiringOperations Manager posts (India, August to September 2026): "100+ full-time team members" collecting data "from 100,000+ contributors and 500+ industry partners" in homes, hotels, restaurants, agriculture, manufacturing, construction and retail "in 30+ countries", with "thousands of custom collection rigs, gloves and wrist cameras". Site contacts include "owners of factories, resorts and restaurants".Source
    • PressTechCrunch (May 26, 2026): $8.2 million from Wing Venture Capital, NVP Capital and Y Combinator. Workers in India's home-services, hotel and restaurant sectors wear camera caps, with more than 1,000 active headsets. A multi-sensor rig adds tactile gloves, a full-body motion-capture suit and wrist cameras, synchronizing RGB-D, motion and tactile force. Plans: Southeast Asia and the U.S., a platform that lets anyone take part and earn, early U.S. pilots offering services such as cleaning or cooking in exchange for data collection, and fine-tuning models on its own data to demonstrate data quality.Source
    • HiringHardware and models: firmware, mechanical and electrical engineers for "our wearable capture hardware" (China), a Camera Systems and Calibration Engineer to "own the entire optics and calibration stack", a Head of Engineering in San Francisco, and a Machine Learning Engineer working on "multimodal perception, VLA training, robotics post-training".Source
  • 23

    HA-Ego-Samples is free to use under Apache-2.0, with 259 task categories and 45 environment types, but the listing shows only 163 downloads. The founding team comes from Stanford, Berkeley, NYU, UPenn and JHU robotics labs, the academic community that turns datasets into standards through benchmarks and citations.

    Medium impactMedium effort First result in 2 quarters Machine Learning Engineer

    First action

    Pick the benchmark tasks and train one baseline on the sample.

    How to measure it

    Teams submitting results

    Now
    –
    Target
    25
    By
    2 quarters

    Steps

    1. 1

      Pick tasks

      Two or three with clear metrics.

      Machine Learning Engineer 1 week

    2. 2

      Train baselines

      Simple, reproducible models.

      Machine Learning Engineer 3 weeks

    3. 3

      Build the leaderboard

      Evaluation script and results page on the site.

      Member of Technical Staff 2 weeks

    4. 4

      Find a venue

      A workshop challenge slot.

      CTO 1 month

    5. 5

      Promote

      To robotics labs and university groups.

      Founder's Office Intern, Partnerships ongoing

    6. 6

      Convert

      Offer top teams full-data access for research.

      CEO after the challenge

    First result
    2 quarters
    Cost
    Staff time and prizes
    Owner
    Machine Learning Engineer

    Smallest test

    Publish one baseline and a leaderboard page, and share them with five university labs.

    Why it works

    Define two or three benchmark tasks on the sample (for example, action recognition across environments, or hand-object interaction detection), publish baselines, an evaluation server or script and a leaderboard on humanarchive.ai, and run it as a challenge at a robotics or vision workshop. Ask entrants to cite the dataset.

    Benchmarks become the default data students and labs train and test on. Every paper that uses one is a citation and a warm lead, at the cost of a sample already given away.

    Risks and how to handle them

    • Low participation

      Partner with an existing workshop and offer compute credits or prizes.

    Evidence

    • Dataset listingHA-Ego-Samples, Human Archive's public sample: "500 hours of footage from 4,651 unique individuals", 9,343 clips across 259 task categories and 45 environment types; RGB video only (MP4); Apache-2.0 licence; access by request on Hugging Face; 163 downloads shown on the listing.Source
    • Your websiteHomepage (checked November 9, 2026), seven sentences in total: "Human Archive is a data and deployments infrastructure company for Physical AI." "Human Archive is currently working alongside leading AI labs to scale human video/sensor data and deploy robotics." "If you're interested in our OTS catalog or a robot deployment, click here or request time here." The links go to a contact form, a 30-minute booking page, an X post and the legal pages. There is no catalogue page, no link to the public sample dataset, no link to the iPhone app and no careers link.Source
    • PressWing VC (May 26, 2026): Human Archive was "the most requested startup coming out of Y Combinator (YC) W26 Demo Day"; buyers are "frontier robotics labs and humanoid foundation model teams"; the pitch is "the Common Crawl for embodied AI"; modalities include egocentric RGB video, audio, stereo depth, tactile force, motion capture and wrist cameras, with QA, anonymization, annotation and benchmarking pipelines before delivery.Source
More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • Ketone-IQ growth strategyDrinkable ketone shots, with and without caffeine, sold as "brain fuel" for focus and energy, direct on ketone.com by subscription and in…21 growth ideas
  • Let's Hang! growth strategyAn event platform from Vyral LLC: event pages, RSVPs, ticket sales, reminders, text blasts and check-in, with payments by Venmo, Zelle…23 growth ideas
  • Owl Labs growth strategyMakes the Meeting Owl family of 360° video conferencing cameras with built-in microphones and speakers, plus the Owl Bar, Whiteboard Owl…22 growth ideas
  • Payhip growth strategyAn e-commerce platform for creators to sell digital downloads, courses, coaching, memberships and physical products from their own store…21 growth ideas
  • WHOOP growth strategyA screenless wearable and membership that tracks strain, sleep and recovery, with tiered memberships (One, Peak, Life), apparel…21 growth ideas
  • Ditto growth strategyAn edge sync platform: an offline-first mobile database with peer-to-peer mesh sync and CRDT conflict resolution (Edge SDK, Cloud, Server…22 growth ideas
  • Kojo growth strategyMaterials management and procurement software for commercial construction: field ordering from a phone, purchasing, warehouse and tool…20 growth ideas
  • Closure Intelligence growth strategyClosure builds an AI "digital analyst" for law enforcement: investigators upload case evidence and Closure transcribes and searches jail…22 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 9, 2026, and cites them throughout. Human Archive did not commission, review or endorse it, and Programmable is not affiliated with Human Archive. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how Human Archive runs its business.

Something wrong, or do you represent Human Archive? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 306 ideas

    Public catalogue and site basics, a contributor page with payout rules, the San Francisco study sign-up page and App Store replies.

  2. Days 31 to 6011 ideas

    Resumable uploads and review status for contributors; a multi-sensor sample, standard formats, datasheets, a collection brief and monthly release notes for buyers; campus, site-partner and referral programmes; a packaged deployment pilot.

  3. Days 61 to 906 ideas

    Commit to the bets: published post-training results, the U.S. services-for-data pilot and annotation as a service; test the voice catalogue, a skilled-hands tier and an open benchmark.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value synchronized multi-sensor data. The most cited problem: labs: shortlisting data vendors without public specs or samples.

  • Real-world environments: homes, hotels, restaurants, factories
  • Labs: consent and provenance questions from their legal teams
  • In their words: “egocentric data”

Market

Competitive landscape

4 competitors compared: Claru, Digital Divide Data, Mindrift and Teleoperation data collection.

  • Gap to exploit: Public, comparable specs for multi-sensor egocentric data
  • Claru: No public catalogue or pricing
  • Digital Divide Data: Office-style environments; collection less real-world

Hiring signals

Human Archive has 22 open roles, and they show where the money is going: collection at scale, not selling.

  • Research Study Coordinator (Operations, Part-time): Source and schedule thousands of collectors in San Francisco. Ideas 5, 19.
  • Operations Manager (on-site): 100,000+ contributors, 500+ partners, 30+ countries. Ideas 7, 15, 22.
  • Strategy & Project Lead (Data Operations): Annotation from hundreds to thousands of people. Ideas 8, 20.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks firstNew accounts get 100,000 free enrichment credits with code BETA100K (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence