ProgrammableMarketing/tryasimov.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 · 9
  • This quarter · 9
  • Big bets · 4
  • Wildcards · 2
Generated Nov 10, 2026
  1. Programmable
  2. Playbooks
  3. Asimov
Asimov logo

growth playbook · Nov 10, 2026

tryasimov.ai (opens in a new tab)

Asimov growth playbook

Diagnosis

Asimov is a three-person YC Winter 2026 company that says the largest robotics labs are already its customers and that it delivers thousands of hours of data a day. Its supply side is well built: a detailed collector application, a contributor agreement, a privacy policy and an app support page. The buyer side of tryasimov.ai is not.

The "Request Data", "Contact Us" and "Partner with Us" buttons are not links, the page has no email or form, "Explore Data" is blank without scripts, and there is no sample, format list or dataset size. Meanwhile a rival publishes an "Asimov alternatives" page that lists its own formats and says Asimov's are not specified.

This playbook first gives buyers a working path and a spec sheet, fixes speed and the search description, and makes one privacy promise everywhere. Then it scales collector supply with the pay ramp, referrals and a demand board, and moves on to pilots, provenance datasheets, a request portal and the teleoperation line the site already advertises.

Growth Score

6/10

Biggest Opportunity

Every robotics buyer who reaches tryasimov.ai today finds no way to ask for data. A working request form, a spec sheet and a viewable sample turn the YC traction claims into inbound pipeline.

Biggest Risk

Better-funded rivals with published numbers and formats (Claru lists 34,000+ contributors and 100+ active datasets) are already writing comparison pages about Asimov, and different privacy wording across the job post, the policy and the cleaning page could slow buyers' compliance reviews.

01

Start here

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

  1. 1

    Make "Request Data" work: a short request form that replies with a sample and a spec sheet

    Link "Request Data" to a mailto with a prefilled subject today, then replace it with the form.

    High impact First result in 1 week
  2. 2

    Publish a one-page dataset spec sheet: formats, annotation schema, QA and turnaround

    List the formats and fields Asimov actually delivers to current customers.

    High impact First result in 2 weeks
  3. 3

    One privacy promise everywhere: the job post, the collector page and the policy should say the same thing

    Put the three current statements side by side and decide which is accurate.

    High impact First result in 2 weeks

02

Your idea portfolio

24 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.

  • 9 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 · 4Big bets · 7Fill-ins · 6Reconsider · 7
Effort →

Showing 24 of 24 ideas

This week 9

Quick wins with a clear first action.

  • 01

    On tryasimov.ai the three buttons a robotics buyer would press, "Request Data", "Contact Us" and "Partner with Us", plus the footer "Contact", are not links in the page as served, and the page has no email address, form or calendar link (checked twice on November 10, 2026). The only buyer contact Asimov publishes is a founders' email on its YC launch post. So a lab engineer who arrives from YC, a referral or search, reads "clean, annotated datasets ready for training" and clicks "Request Data" gets nothing, while Claru's homepage offers "Book a Call" and first delivery "in days, not months". The site's supply side, by contrast, has a full collector application. The demand side, which pays the bills, has no path at all.

    High impactLow effort First result in 1 week Founder

    First action

    Link "Request Data" to a mailto with a prefilled subject today, then replace it with the form.

    How to measure it

    Qualified buyer requests from the site

    Now
    –
    Target
    10 a month
    By
    60 days

    Steps

    1. 1

      Stop the dead end today

      Point all four buyer buttons to a mailto with a prefilled subject so nothing is lost while the form is built.

      Founder (engineering) 1 hour

    2. 2

      Write the six questions

      Organisation, embodiment, tasks and environments, hours, format, timeline. Keep the form under a minute.

      Founder (go-to-market) 2 hours

    3. 3

      Build /request

      A server-rendered form with spam protection, sending to the founders' inbox and a shared channel.

      Founder (engineering) 1 day

    4. 4

      Auto-reply with proof

      An instant reply with the spec sheet, one sample clip and a calendar link.

      Founder (go-to-market) half a day

    5. 5

      Split partners from buyers

      "Partner with Us" goes to a business form so labs and restaurants are routed differently.

      Founder (engineering) 2 hours

    6. 6

      Check without scripts

      Confirm the links and footer email appear in the raw page and in link previews.

      Founder (engineering) 1 hour

    7. 7

      Track and reply fast

      Log source, request size and time to first reply; aim for a same-day answer.

      Founder (go-to-market) weekly

    First result
    1 week
    Cost
    $0; about two days of founder time
    Owner
    Founder (go to market)

    Smallest test

    Link "Request Data" to a prefilled email today and count the messages over two weeks.

    Why it works

    Turn "Request Data" and "Contact Us" into links to a /request page with a six-field form: organisation, robot or embodiment, tasks and environments, hours needed, preferred format, and timeline. Route submissions to the founders' inbox and a shared channel, and auto-reply within minutes with a spec sheet (idea 2) and a link to a short sample clip. Point "Partner with Us" to a separate business form (idea 13) so restaurants and warehouses don't land in the lab queue. Add a plain email link in the footer as a fallback, and check that the links work with scripts turned off, so AI assistants and link previews that read the raw page also find a route.

    Inbound buyers of custom data are few and valuable. Each lost lab inquiry is a lost contract, and a form that asks about the robot, tasks and hours qualifies the lead and shortens the first call.

    Ready to use

    CopyRequest Data form fields

    1. Organisation and your role
    2. Robot or embodiment (humanoid, arm, mobile manipulator, other)
    3. Tasks and environments (kitchen, cleaning, office, warehouse, other)
    4. Hours needed (under 50, 50-500, 500-5,000, 5,000+)
    5. Preferred format and annotations (video only, + pose, + depth, + segmentation)
    6. Timeline (pilot this month, this quarter, exploring)
    
    Auto-reply: thanks + spec sheet + 30-second sample clip + calendar link.

    Risks and how to handle them

    • Low-quality or spam requests

      Require organisation and hours needed, and add a simple bot check.

    • Founders can't answer quickly

      The auto-reply carries the spec sheet and sample, so buyers get value before the call.

    Evidence

    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
    • Competitor researchClaru homepage: "Training data for physical AI"; "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships"; a "Book a Call" button and first delivery "in days, not months".Source
  • 02

    Asimov's homepage lists annotation types but no file format, dataset size, sample or turnaround. A rival's "Asimov (YC W26) alternatives" page lists its own formats (WebDataset, HDF5, RLDS, with manifests, checksums and datasheets) and pilots in under a week, and says Asimov's formats are not specified. A buyer comparing the two only sees numbers for the rival.

    High impactLow effort First result in 2 weeks Founder

    First action

    List the formats and fields Asimov actually delivers to current customers.

    How to measure it

    Spec-to-request rate

    Now
    –
    Target
    15% of spec viewers
    By
    60 days

    Steps

    1. 1

      List what ships

      Formats, fields, frame rates and metadata from recent deliveries.

      Founder (engineering) 1 day

    2. 2

      Write the QA rules

      Turn the contributor agreement's rejection reasons into buyer-facing quality guarantees.

      QA lead half a day

    3. 3

      Add a sample

      One 30-second clip with its annotation files, from a consenting collector.

      Founder (engineering) 1 day

    4. 4

      Publish /data-spec

      HTML page and downloadable PDF, dated.

      Founder (engineering) 1 day

    5. 5

      Link it

      From the homepage buyer section and the request auto-reply.

      Founder (go-to-market) 1 hour

    6. 6

      Update quarterly

      Refresh when formats or annotations change.

      Founder (engineering) quarterly

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Founder (engineering)

    Smallest test

    Send the spec as a PDF to the next three buyer conversations and ask what is missing.

    Why it works

    Publish /data-spec (HTML plus a PDF): capture rig (phone model classes, head mount, ARKit pose, LiDAR/depth where supported), the annotation schema, output formats, the QA rules the contributor agreement already sets (hands visible, camera angle, minimum length, no scripted or duplicate activity), privacy handling, typical pilot turnaround and minimum order. Link it from the homepage and the request auto-reply.

    Robotics engineers decide whether data fits their pipeline from formats and schema. A spec sheet answers that before the call and takes the "not specified" line away from rivals.

    Risks and how to handle them

    • Revealing too much to rivals

      Publish formats and quality rules, not tooling or collector sources.

    Evidence

    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Competitor researchClaru publishes an "Asimov (YC W26) alternatives" comparison page. It lists Claru's delivery formats (WebDataset, HDF5, RLDS, with manifests, checksums and datasheets) and pilot datasets in under a week; for Asimov, delivery formats are not specified.Source
    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
  • 03

    The footer's "Explore Data" link goes to /explore, which shows only the word "Asimov" to any reader that doesn't run scripts: no datasets, hours, environments or way to request data. Search crawlers, link previews and AI assistants often read pages that way, and the homepage's own 3D viewer reads "Loading 3D viewer..." in the same view. The one page meant to show the product shows nothing to them.

    Medium impactMedium effort First result in 1 month Founder

    First action

    Add a static summary block and a request link above the viewer on /explore.

    How to measure it

    Explore-to-request rate

    Now
    –
    Target
    10%
    By
    60 days

    Steps

    1. 1

      Pick what to show

      Environments, task types and counts you are happy to publish.

      Founder (go-to-market) half a day

    2. 2

      Render on the server

      Use the framework's server rendering for the summary and grid.

      Founder (engineering) 3 days

    3. 3

      Add poster images

      Light, compressed stills for each sample clip.

      Founder (engineering) 1 day

    4. 4

      Link to the request form

      A "Request this data" button on each card, passing the environment to the form.

      Founder (engineering) half a day

    5. 5

      Title and description

      Unique meta tags and a share image for /explore.

      Founder (go-to-market) 1 hour

    6. 6

      Check readers

      Verify the page with scripts off and in a link-preview tester.

      Founder (engineering) 1 hour

    First result
    1 month
    Cost
    Engineering time
    Owner
    Founder (engineering)

    Smallest test

    Add one paragraph and a request link in plain HTML to /explore today.

    Why it works

    Server-render the essentials of /explore: a short intro, a grid of environments and task types with clip counts or hours, poster images for sample clips, and a "Request this data" button linked to the form from idea 1. Keep the interactive viewer as an enhancement on top. Give the page its own title and description.

    Pages that can be read without scripts can be indexed, previewed when shared in Slack or email, and summarised by assistants. When a lab engineer pastes the link to a colleague, the preview then shows data, not a blank page.

    Risks and how to handle them

    • Publishing counts that change daily

      Show rounded figures and a last-updated date.

    Builds on

    Evidence

    • Your websiteThe "Explore Data" page (/explore), linked from the footer, shows only the word "Asimov" to a reader that does not run scripts: no datasets, hours, environments or way to request data.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
  • 04

    Real visitors fail Google's Core Web Vitals on loading: the main content appears at 2.6 s for the slowest quarter of visits, against a 2.5 s target. Responsiveness and layout stability pass. In Google's phone lab test the main content takes 21 s, and the page downloads 5.9 MB, 4.0 MB of it images, with about 3.9 MB saveable by better image delivery. Most of the homepage images are photos of people and partner meetings.

    Medium impactLow effort First result in 1 month Founder

    First action

    Convert the five largest homepage images and re-run PageSpeed.

    How to measure it

    Field LCP p75

    Now
    –
    Target
    under 2.5 s (passing)
    By
    45 days

    Steps

    1. 1

      List the heavy files

      Rank homepage images by size from the PageSpeed report.

      Founder (engineering) 1 hour

    2. 2

      Use the image component

      Modern formats, sizes per screen width, width and height set.

      Founder (engineering) half a day

    3. 3

      Lazy-load below the fold

      Everything after the first screen, including partner photos.

      Founder (engineering) 2 hours

    4. 4

      Defer the 3D viewer

      Load it on scroll or on click, with a still image first.

      Founder (engineering) half a day

    5. 5

      Re-test

      PageSpeed on mobile and desktop.

      Founder (engineering) 1 hour

    6. 6

      Watch real visitors

      Core Web Vitals report over the next 28 days.

      Founder (engineering) monthly

    First result
    1 month (real-visitor data updates over 28 days)
    Cost
    $0
    Owner
    Founder (engineering)

    Smallest test

    Compress the single largest image and compare the lab score.

    Why it works

    Serve the homepage photos through the framework's image component in modern formats at display size, lazy-load everything below the first screen, and load the 3D viewer only when it scrolls into view. Re-test on PageSpeed and track the real-visitor numbers in Search Console.

    Collectors apply from phones, often on mobile data in many countries, and lab buyers judge polish fast. A page 10 times lighter loads faster on both and passes the test that counts for search.

    Risks and how to handle them

    • Photos look softer

      Check quality at 2x on a phone before shipping.

    Evidence

    • PageSpeedGoogle PageSpeed (November 10, 2026), homepage on mobile: performance 65 (desktop 80). Real visitors fail Core Web Vitals on Largest Contentful Paint (75th percentile 2.6 s against a 2.5 s target; responsiveness and layout stability pass). In the lab test the main content takes 21.0 s; the page downloads 5.9 MB, 4.0 MB of it images, and better image delivery would save about 3.9 MB.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
  • 05

    The description Google and link previews show says "End-to-end data and teleoperation services for embodied AI. From egocentric capture to remote operation." The page itself argues that teleoperation gives "thousands of hours at best, usually in controlled labs" and sells human activity capture instead, and no page describes a teleoperation service. Searchers are promised something the visit doesn't deliver, and the pitch against teleoperation is muddied before the click.

    Medium impactLow effort First result in 2 weeks Founder

    First action

    Replace the homepage description today.

    How to measure it

    Homepage search CTR

    Now
    –
    Target
    +15%
    By
    60 days

    Steps

    1. 1

      Draft three options

      Each under 155 characters with the core offer and one proof point.

      Founder (go-to-market) 1 hour

    2. 2

      Ship the homepage one

      Update the metadata in the app.

      Founder (engineering) 30 minutes

    3. 3

      Write the others

      Collectors, cleaning and explore pages.

      Founder (go-to-market) 1 hour

    4. 4

      Check previews

      Paste links into Slack, X and LinkedIn preview tools.

      Founder (go-to-market) 30 minutes

    5. 5

      Request re-crawl

      Submit the pages in Search Console.

      Founder (engineering) 15 minutes

    6. 6

      Measure

      Compare click-through after four weeks.

      Founder (go-to-market) monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Founder (go to market)

    Smallest test

    Change the homepage description today and check the preview.

    Why it works

    Rewrite the homepage description around the real offer, for example: "Egocentric video of real people doing real tasks in homes and workplaces, annotated with 3D pose, depth and segmentation, for robotics teams. Backed by Y Combinator." Add distinct descriptions for /collectors, /cleaning and /explore. If teleoperation is a real plan, give it its own page later (idea 21).

    A description that matches the page lifts click-through and keeps AI summaries of Asimov accurate.

    Risks and how to handle them

    • Teleoperation is a planned product

      Mention it on its own page once it is live, not in the homepage description.

    Evidence

    • Your websiteThe homepage description used by search results and link previews says "End-to-end data and teleoperation services for embodied AI. From egocentric capture to remote operation.", while the page argues that teleoperation gives "thousands of hours at best, usually in controlled labs" and sells human activity capture; no page describes a teleoperation or remote-operation service.Source
    • Your websiteSite scan (November 10, 2026): no blog, case studies, pricing page, social share image or links to company social profiles; the domain was registered in November 2025.Source
  • 06

    tryasimov.ai has no social share image and no links to company social profiles. Asimov's growth runs on links being passed around: YC launch posts, lab engineers sharing in Slack, collectors sharing the job with friends. Each share currently shows a bare preview, and visitors can't check the team's profiles for credibility.

    Low impactLow effort First result in 1 week Founder

    First action

    Design one 1200x630 image and add it to the homepage metadata.

    How to measure it

    Referral and social sessions

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Design three images

      Homepage, collectors and cleaning.

      Founder (go-to-market) half a day

    2. 2

      Add the tags

      Open Graph and Twitter card tags per page.

      Founder (engineering) 1 hour

    3. 3

      Add profile links

      LinkedIn, X and YC in the footer.

      Founder (engineering) 30 minutes

    4. 4

      Test previews

      In Slack, LinkedIn and messaging apps.

      Founder (go-to-market) 30 minutes

    5. 5

      Refresh caches

      Use each network's preview debugger.

      Founder (go-to-market) 15 minutes

    6. 6

      Track

      Tag shared links by channel.

      Founder (go-to-market) monthly

    First result
    1 week
    Cost
    $0
    Owner
    Founder (go to market)

    Smallest test

    Add the homepage image and compare previews before and after.

    Why it works

    Create a share image for the homepage (headline plus a frame from a real capture) and separate ones for /collectors ("Get paid to record everyday tasks") and /cleaning. Add footer links to the company's LinkedIn and X profiles, and the YC page.

    Rich previews get more clicks when shared, and visible company profiles reassure both buyers and collectors that a young company is real.

    Risks and how to handle them

    • A capture frame shows private details

      Use a frame already cleared under the privacy rules.

    Evidence

    • Your websiteSite scan (November 10, 2026): no blog, case studies, pricing page, social share image or links to company social profiles; the domain was registered in November 2025.Source
    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
  • 07

    Asimov describes privacy three ways. The Data Collection Specialist post says no audio is collected and faces and personal information are automatically blurred. The Privacy Policy says microphone audio is collected when an assignment needs narration or the collector opts in, and doesn't mention face blurring. The collector page asks collectors to keep faces, screens and documents out of frame. A collector, or a lab's compliance reviewer, who reads all three gets different answers.

    High impactLow effort First result in 2 weeks Founder

    First action

    Put the three current statements side by side and decide which is accurate.

    How to measure it

    Privacy-related support and buyer questions

    Now
    –
    Target
    -50%
    By
    90 days

    Steps

    1. 1

      Compare statements

      Job post, collector page, policy, app onboarding.

      Founder (engineering) 2 hours

    2. 2

      Confirm the pipeline

      What audio, blurring and review steps really run today.

      Founder (engineering) half a day

    3. 3

      Write /privacy-practices

      Plain-language page for collectors and buyers.

      Founder (go-to-market) 1 day

    4. 4

      Legal check

      Have counsel align the policy with the page.

      Outside counsel 1 week

    5. 5

      Update everywhere

      Job posts, collector page, app and spec sheet.

      Strategic Operations Intern 1 day

    6. 6

      Review each release

      Recheck when the app changes what it captures.

      Founder (engineering) per release

    First result
    2 weeks
    Cost
    Staff time plus a short legal review
    Owner
    Founder (engineering)

    Smallest test

    Fix the job post wording to match the policy this week.

    Why it works

    Write one canonical "How we protect privacy" page covering audio (when and how it is opted into), on-device review before upload, automated blurring if it exists and what it covers, what collectors must avoid, and what buyers receive. Update the job post and collector page to link to it with identical wording, and update the policy if blurring is a real step.

    Consistent privacy promises build collector trust and remove a question that slows down buyers' legal and ethics reviews, which can decide large data contracts.

    Risks and how to handle them

    • Stronger promises than the pipeline delivers

      Only state steps that run on every upload today.

    Evidence

    • HiringData Collection Specialist (YC job board, remote, "anywhere in the world"): "$5-15/hr base pay" rising "up to $30/hr after your first 5 hours collected", plus volume bonuses and unlimited hours; the post says no audio is collected and that faces and personal information are automatically blurred.Source
    • Your websitePrivacy Policy (updated June 20, 2026): covers "approved adult collectors" using the Asimov Capture app; collects first-person video, microphone audio only when an assignment needs narration or the collector opts in, and ARKit pose and LiDAR/depth where supported; accepted data may be "sold, licensed, sublicensed, shared"; it does not mention face blurring and does not address cleaning clients or partner businesses.Source
    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source
  • 08

    The homepage says "Our cleaners wear our kits", and Asimov cleans apartments for over 100 founders, investors and students in the Bay Area each weekend. The /cleaning booking page shows "Privacy reviewed" and "Avoid faces & personal info" badges but never mentions the kits or how footage from a cleaning is used, and the Privacy Policy covers collectors, not cleaning clients. Saying it plainly protects the brand and turns the cleaning business into a documented, consented data source.

    Medium impactLow effort First result in 2 weeks Founder

    First action

    Draft the four-sentence disclosure and add it above the booking form.

    How to measure it

    Consented cleaning bookings

    Now
    –
    Target
    100% of recorded cleanings
    By
    30 days

    Steps

    1. 1

      Write the disclosure

      What is recorded, what is avoided, how footage is used.

      Founder (go-to-market) 2 hours

    2. 2

      Add the choice

      Consent checkbox and room-skip notes on the form.

      Founder (engineering) half a day

    3. 3

      Legal review

      Counsel checks the wording and the policy section.

      Outside counsel 1 week

    4. 4

      Brief cleaners

      A script for explaining the kit at the door.

      Strategic Operations Intern 1 day

    5. 5

      Tag the data

      Mark every cleaning session with its consent record.

      Founder (engineering) 2 days

    6. 6

      Ask existing clients

      Email current clients the new terms and confirm consent.

      Strategic Operations Intern 1 week

    First result
    2 weeks
    Cost
    Staff time plus a short legal review
    Owner
    Founder (go to market)

    Smallest test

    Add the disclosure text and watch the booking rate for two weeks.

    Why it works

    Add a short "How recording works" block to /cleaning: what the cleaner's kit records, what is avoided, how clients can ask for areas to be skipped, and what the discount reflects. Add a consent checkbox (with a no-recording option at standard price, if that suits the business) to the booking form, and a client section in the privacy materials.

    Clients who choose recording knowingly are happier and refer friends. Labs buying home data can be told every home in the set gave consent, which supports idea 19.

    Risks and how to handle them

    • Fewer bookings

      Keep the free first cleaning and test a no-recording price.

    Builds on

    Evidence

    • Your websiteCleaning page (/cleaning): "Bay Area Only"; "Your first cleaning is completely free with no commitment"; "Starting at $10/hr" after the first free 3-hour cleaning; badges "Privacy reviewed" and "Avoid faces & personal info". It does not mention the recording kits or how footage from a cleaning is used.Source
    • Your websiteHomepage, businesses and cleaning: "We pay your workers' salaries while they do their normal jobs wearing lightweight recording equipment", with dataset revenue share "For larger partnerships"; "We also own a growing cleaning company, offering free and reduced-price cleanings. Our cleaners wear our kits"; it cleans apartments for "over 100 founders, investors, and students" in the Bay Area each weekend.Source
    • Your websitePrivacy Policy (updated June 20, 2026): covers "approved adult collectors" using the Asimov Capture app; collects first-person video, microphone audio only when an assignment needs narration or the collector opts in, and ARKit pose and LiDAR/depth where supported; accepted data may be "sold, licensed, sublicensed, shared"; it does not mention face blurring and does not address cleaning clients or partner businesses.Source
  • 09

    The YC job post offers "$5-15/hr base pay" rising "up to $30/hr after your first 5 hours collected", plus volume bonuses. The /collectors page, where most applicants land, shows only "$5-15/hour" and makes applicants tick that rates are typically $5-15. The strongest reason to finish the first five hours is missing exactly where people decide.

    Medium impactLow effort First result in 2 weeks Founder

    First action

    Add one line under the pay figure: "Up to $30/hr after your first 5 approved hours."

    How to measure it

    5-hour completion rate

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Confirm the rules

      When the higher rate applies and to which tasks.

      Founder (go-to-market) 1 hour

    2. 2

      Add the ladder

      Three rows on /collectors.

      Founder (engineering) 2 hours

    3. 3

      Update the checkbox

      Match the expectations text to the ladder.

      Founder (engineering) 30 minutes

    4. 4

      Show progress in the app

      Accepted minutes toward five hours.

      Founder (engineering) 3 days

    5. 5

      Email at milestones

      At 1, 3 and 5 approved hours.

      Strategic Operations Intern 1 day

    6. 6

      Measure

      Applications and 5-hour completion, before and after.

      Strategic Operations Intern weekly

    First result
    2 weeks
    Cost
    $0 beyond the pay already offered
    Owner
    Founder (go to market)

    Smallest test

    Add the one line today and compare a week of applications.

    Why it works

    Add a pay ladder to /collectors: base rate, the rate after five accepted hours, and how volume bonuses work, with the caveat that only approved time is paid. Repeat it in the app's first-run screen and the acceptance email. Test whether applications and first-week hours rise.

    A visible next step raises sign-ups and pushes new collectors through the first five hours, where most drop-off happens in paid crowd work.

    Risks and how to handle them

    • Collectors expect $30 from day one

      Show the ladder with clear conditions and the approved-time rule.

    Evidence

    • HiringData Collection Specialist (YC job board, remote, "anywhere in the world"): "$5-15/hr base pay" rising "up to $30/hr after your first 5 hours collected", plus volume bonuses and unlimited hours; the post says no audio is collected and that faces and personal information are automatically blurred.Source
    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source

This quarter 9

The core plays for the next three months.

  • 10

    Collectors are paid only after QA accepts a session, and the Contributor Agreement lists the common rejection reasons: "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", duplicate or scripted activity. These live in a legal page. The collector page has no FAQ, and support is one inbox. Each rejected session costs the collector unpaid time and Asimov QA time.

    High impactMedium effort First result in 1 month QA lead

    First action

    Pull last month's rejections by reason and pick the top three.

    How to measure it

    Accepted share of uploaded minutes (new collectors)

    Now
    –
    Target
    +20 points
    By
    90 days

    Steps

    1. 1

      Rank rejection reasons

      Last 30 days of QA decisions.

      QA lead 1 day

    2. 2

      Make the checklist

      Good and bad frames for each top reason.

      QA lead 2 days

    3. 3

      Record the video

      90 seconds, shown before the first assignment.

      Strategic Operations Intern 2 days

    4. 4

      Add pre-upload checks

      Hands-visible, angle and length warnings in the app.

      Founder (engineering) 2 weeks

    5. 5

      Explain each rejection

      Reason plus one fix tip in the app and email.

      Founder (engineering) 3 days

    6. 6

      Publish a FAQ

      Top 15 support questions on /support.

      Strategic Operations Intern 2 days

    7. 7

      Track

      Acceptance by cohort each week.

      QA lead weekly

    First result
    1 month
    Cost
    About three weeks of staff and engineering time
    Owner
    QA lead

    Smallest test

    Send the checklist to the next 50 new collectors and compare their acceptance rate.

    Why it works

    Make a 90-second onboarding video and a one-screen checklist from the rejection reasons, with good and bad example frames. Add live in-app checks for the measurable ones (hands in frame, angle, minimum length) before upload, and send each rejection with its reason and a fix tip. Publish a collector FAQ on /support.

    Teaching quality up front raises the acceptance rate, which lifts collector earnings and retention and lowers QA cost per accepted hour.

    Risks and how to handle them

    • In-app checks give false warnings

      Warn, don't block, until the checks are tuned.

    Evidence

    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
    • Your websiteSupport page: help for the Asimov Capture app (recordings, uploads, review status, payouts, account deletion) through a single support email inbox; no FAQ or help articles.Source
    • HiringStrategic Operations Intern (YC job board, Berkeley or remote US, $20-50 an hour, about 20 hours a week): manages contractors "across countries and time zones", equipment shipments and inventory, onboarding, monitoring contractor activity and following up on issues, and tracking data in spreadsheets and internal tools.Source
  • 11

    The collector application already has an optional referral code field, but neither the collector page nor the Contributor Agreement describes a referral program: no reward, no rules, no way for a collector to get a code. Crowd collection grows fastest through people who already do the work recruiting friends in the same city.

    High impactLow effort First result in 1 month Strategic Operations Intern

    First action

    Decide the reward and the five-hour condition.

    How to measure it

    Referral share of activated collectors

    Now
    –
    Target
    30%
    By
    90 days

    Steps

    1. 1

      Set the reward

      A bonus paid at five approved hours for both people.

      Founder (go-to-market) 1 day

    2. 2

      Generate codes

      One code and share link per active collector.

      Founder (engineering) 3 days

    3. 3

      Write the rules

      On /collectors and in the agreement.

      Outside counsel 1 week

    4. 4

      Add fraud checks

      Device and payout-account matching.

      Founder (engineering) 3 days

    5. 5

      Launch to top collectors

      Message the most reliable 10%.

      Strategic Operations Intern 1 day

    6. 6

      Steer by demand

      Extra bonus for needed cities and phone models.

      Strategic Operations Intern ongoing

    7. 7

      Measure

      Referred cohort acceptance and retention.

      Strategic Operations Intern monthly

    First result
    1 month
    Cost
    Referral bonuses, paid only on approved hours
    Owner
    Strategic Operations Intern

    Smallest test

    Give codes to 20 top collectors for two weeks.

    Why it works

    Give every active collector a code and link in the app. Pay a bonus when a referred collector reaches five approved hours (tying the reward to quality, not sign-ups), and a smaller one to the new collector. Publish the rules on /collectors and in the agreement, and target referral pushes at the cities and phone models the demand board (idea 12) needs.

    Referred collectors are more likely to finish onboarding, because a friend has shown them how, and paying on approved hours keeps fraud and low-quality sign-ups down.

    Risks and how to handle them

    • Referral fraud

      Pay only after five accepted hours and check for duplicate devices and payout accounts.

    Builds on

    Evidence

    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source
    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
  • 12

    Applicants are matched on country, city, exact phone model, hours and task demand, and participation is limited to an "allowed launch country/region". But no page lists the allowed countries or what's in demand, so people outside the current need apply and wait, while the cities and phones Asimov actually needs don't hear about it.

    Medium impactMedium effort First result in 1 month Strategic Operations Intern

    First action

    List this month's open countries and the most-needed phone models.

    How to measure it

    7-day match rate

    Now
    –
    Target
    +30%
    By
    90 days

    Steps

    1. 1

      Pull demand

      Open assignments by country, phone and task.

      Strategic Operations Intern 1 day

    2. 2

      Design the board

      Open, limited and waitlist states.

      Founder (go-to-market) 1 day

    3. 3

      Build it

      A server-rendered block on /collectors.

      Founder (engineering) 3 days

    4. 4

      Update weekly

      From the assignment backlog.

      Strategic Operations Intern weekly

    5. 5

      Aim outreach

      Referral bonuses and local posts for open cells.

      Strategic Operations Intern weekly

    6. 6

      Measure

      Match rate and time to first assignment.

      Strategic Operations Intern monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Strategic Operations Intern

    Smallest test

    Post a static list of open countries and phones for two weeks.

    Why it works

    Add a demand board to /collectors: open countries, the phone models in demand (for example, ones with depth sensors), the task types needed this month, and a "waitlist" state elsewhere. Feed it from the assignment backlog weekly, and use it to aim referral bonuses and local posts.

    Telling people where demand is raises the share of applications that can be activated, and it cuts wasted review time for the operations team.

    Risks and how to handle them

    • Revealing customer priorities

      Show categories, never which lab needs them.

    Evidence

    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source
    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
    • HiringStrategic Operations Intern (YC job board, Berkeley or remote US, $20-50 an hour, about 20 hours a week): manages contractors "across countries and time zones", equipment shipments and inventory, onboarding, monitoring contractor activity and following up on issues, and tracking data in spreadsheets and internal tools.Source
  • 13

    Asimov's most unusual offer is "We pay your workers' salaries while they do their normal jobs wearing lightweight recording equipment", with dataset revenue share for larger partnerships, and the YC launch post invites restaurant and warehouse owners to get in touch. On the site the offer is one paragraph, and its "Partner with Us" button is not a link. An owner who would gladly have payroll covered has nothing to read and nowhere to apply.

    High impactMedium effort First result in 6 weeks Founder

    First action

    Write the restaurant version of the page with a simple coverage table.

    How to measure it

    Business partner applications

    Now
    –
    Target
    15 a month
    By
    90 days

    Steps

    1. 1

      Define the pilot

      Length, staff count, kit, payment terms.

      Founder (go-to-market) 2 days

    2. 2

      Build the estimator

      Headcount x shifts x task fit, shown as a range.

      Founder (engineering) 3 days

    3. 3

      Write vertical sections

      Restaurants, hotels, cleaning, warehouses.

      Founder (go-to-market) 3 days

    4. 4

      Staff FAQ

      Consent, privacy, what changes at work.

      Founder (go-to-market) 1 day

    5. 5

      Partner form

      Wire "Partner with Us" to it.

      Founder (engineering) half a day

    6. 6

      Local outreach

      Bay Area restaurants and hotels, starting with cleaning clients' networks.

      Strategic Operations Intern ongoing

    7. 7

      Publish a story

      First pilot partner, with permission.

      Founder (go-to-market) after first pilot

    First result
    6 weeks
    Cost
    Staff time; pilot payroll funded from data sales
    Owner
    Founder (go to market)

    Smallest test

    Pitch five local restaurants with a one-page PDF version.

    Why it works

    Build /partners with sections for restaurants, hotels, cleaning companies and warehouses: how a pilot works, what staff wear, what is never recorded, how coverage is calculated (an estimator from headcount, shifts and task types) and a partner application. Add a short FAQ for staff, and one partner story once a pilot runs.

    Covered payroll is a concrete, easy-to-explain benefit for thin-margin businesses, and each partner adds many hours of real-workplace data that individuals at home can't produce.

    Risks and how to handle them

    • Staff feel pressured to record

      Individual opt-in, with no penalty for declining.

    Builds on

    Evidence

    • Your websiteHomepage, businesses and cleaning: "We pay your workers' salaries while they do their normal jobs wearing lightweight recording equipment", with dataset revenue share "For larger partnerships"; "We also own a growing cleaning company, offering free and reduced-price cleanings. Our cleaners wear our kits"; it cleans apartments for "over 100 founders, investors, and students" in the Bay Area each weekend.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
  • 14

    Asimov publishes no prices, minimums or turnaround, so every buyer starts with an open-ended conversation. Claru promises first delivery "in days, not months" and pilots in under a week. A smaller lab or a new team at a large one needs a small, low-risk first purchase it can approve without a long negotiation.

    High impactMedium effort First result in 6 weeks Founder

    First action

    Price a pilot from the cost of recent deliveries.

    How to measure it

    Paid pilots

    Now
    –
    Target
    5 pilots, 40% converting
    By
    2 quarters

    Steps

    1. 1

      Cost it

      Collector pay, QA and annotation per accepted hour.

      Founder (engineering) 1 day

    2. 2

      Fix the scope

      Hours, environment, annotations, delivery days.

      Founder (go-to-market) 1 day

    3. 3

      Set the credit

      Pilot fee credited to a larger order.

      Founder (go-to-market) 1 hour

    4. 4

      Pre-stock supply

      Keep a bank of collectors ready for pilot environments.

      Strategic Operations Intern 2 weeks

    5. 5

      Publish the offer

      On the spec sheet and request form.

      Founder (engineering) half a day

    6. 6

      Review each pilot

      A results call with the buyer at delivery.

      Founder (go-to-market) per pilot

    First result
    6 weeks
    Cost
    Collection costs, covered by the pilot fee
    Owner
    Founder (go to market)

    Smallest test

    Offer the pilot to the next three buyer conversations.

    Why it works

    Package a pilot: a set number of hours in one environment (for example, kitchen tasks), the standard annotations, delivery in a stated number of days, at a fixed price credited against a larger order. List it on the spec sheet and request form as "Start with a pilot".

    A fixed price and date turn "let's talk" into a purchase order, and pilots that land well grow into the large custom collections.

    Risks and how to handle them

    • Missing the promised date

      Only offer environments with ready collector supply.

    Builds on

    Evidence

    • Competitor researchClaru homepage: "Training data for physical AI"; "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships"; a "Book a Call" button and first delivery "in days, not months".Source
    • Competitor researchClaru publishes an "Asimov (YC W26) alternatives" comparison page. It lists Claru's delivery formats (WebDataset, HDF5, RLDS, with manifests, checksums and datasheets) and pilot datasets in under a week; for Asimov, delivery formats are not specified.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
  • 15

    Claru already publishes an "Asimov (YC W26) alternatives" page that frames Asimov as specialised, lists Claru's formats and datasheets, and says Asimov's formats aren't specified. Right now, a buyer searching Asimov's name plus "alternatives" or "vs" reads the rival's version, and tryasimov.ai has no pages beyond the core five.

    Medium impactLow effort First result in 1 month Founder

    First action

    Draft the table from the spec sheet facts.

    How to measure it

    Compare-page requests

    Now
    –
    Target
    5 a month
    By
    90 days

    Steps

    1. 1

      Pick the frame

      Approaches first (teleoperation, broad vendors), named rivals only where fair.

      Founder (go-to-market) half a day

    2. 2

      Check the facts

      Each rival claim against its own public pages.

      Founder (go-to-market) 1 day

    3. 3

      Write the page

      Table plus when Asimov is not the right fit.

      Founder (go-to-market) 2 days

    4. 4

      Legal review

      Claims and trademarks.

      Outside counsel 3 days

    5. 5

      Publish and link

      From the spec sheet and footer.

      Founder (engineering) 1 hour

    6. 6

      Refresh

      Quarterly, with the date shown.

      Founder (go-to-market) quarterly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Founder (go to market)

    Smallest test

    Publish the approach comparison (no named rivals) first.

    Why it works

    Publish /compare with a fair table: Asimov vs in-house teleoperation vs broad data vendors, on environments (homes and real workplaces), capture (phone with pose and depth), annotations, formats, consent, turnaround and pilot terms. Keep claims factual and dated, and link it from the spec sheet.

    Comparison searches come from buyers close to a decision, and a page on your own domain usually ranks for your own name.

    Risks and how to handle them

    • Outdated claims about rivals

      Date the page and cite each rival's own site.

    Builds on

    Evidence

    • Competitor researchClaru publishes an "Asimov (YC W26) alternatives" comparison page. It lists Claru's delivery formats (WebDataset, HDF5, RLDS, with manifests, checksums and datasheets) and pilot datasets in under a week; for Asimov, delivery formats are not specified.Source
    • Competitor researchClaru homepage: "Training data for physical AI"; "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships"; a "Book a Call" button and first delivery "in days, not months".Source
  • 16

    Asimov's YC page says "The largest robotics labs in the world are already our customers" and that it delivers "thousands of hours a day to leading labs", but tryasimov.ai has no case studies or customer logos from buyers; the only logos are its own. A claim without a name is easy for rivals with published partnership counts to outshine.

    High impactMedium effort First result in 1 quarter Founder

    First action

    Ask the two happiest customers whether they would approve a short story.

    How to measure it

    Request rate on pages with the story

    Now
    –
    Target
    +25%
    By
    2 quarters

    Steps

    1. 1

      Pick candidates

      Customers with repeat orders.

      Founder (go-to-market) 1 day

    2. 2

      Ask

      Offer named, anonymised or logo-only options.

      Founder (go-to-market) 1 week

    3. 3

      Interview

      30 minutes on need, delivery and result.

      Founder (go-to-market) 1 day

    4. 4

      Write and approve

      One page, reviewed by the customer.

      Founder (go-to-market) 2 weeks

    5. 5

      Publish

      Homepage, spec sheet and request auto-reply.

      Founder (engineering) 1 day

    6. 6

      Repeat

      One story a quarter.

      Founder (go-to-market) quarterly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Founder (go to market)

    Smallest test

    Add one approved quote to the homepage.

    Why it works

    Ask the most satisfied customer for an approved story, even anonymised by type ("a humanoid lab") if a name isn't possible: what they needed, environments and hours delivered, turnaround, and one result they can share. Publish it on the homepage buyer section and the spec sheet, with a quote.

    One credible reference does more for a young data vendor than any feature list, because buyers' biggest worry is whether the data has worked for someone like them.

    Risks and how to handle them

    • Labs won't be named

      Use an anonymised story with real numbers.

    Evidence

    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
    • Your websiteSite scan (November 10, 2026): no blog, case studies, pricing page, social share image or links to company social profiles; the domain was registered in November 2025.Source
  • 17

    Asimov says it delivers "thousands of hours a day" and collects "across 3 continents", but the site shows no number about its own data. Claru's homepage leads with counts: contributors, countries, active datasets. Buyers compare numbers, and Asimov's are only in a YC post.

    Medium impactMedium effort First result in 1 month Founder

    First action

    Define the five figures and how each is counted.

    How to measure it

    Homepage request conversion

    Now
    –
    Target
    +20%
    By
    90 days

    Steps

    1. 1

      Define figures

      Accepted hours, countries, sites, tasks, collectors.

      Founder (engineering) 1 day

    2. 2

      Automate

      Monthly query from the QA database.

      Founder (engineering) 3 days

    3. 3

      Add the band

      Homepage and /explore, server-rendered.

      Founder (engineering) 1 day

    4. 4

      Write the note

      Three lines and a chart each month.

      Founder (go-to-market) monthly

    5. 5

      Share

      LinkedIn, X and to customers.

      Founder (go-to-market) monthly

    6. 6

      Audit

      Spot-check figures each quarter.

      QA lead quarterly

    First result
    1 month
    Cost
    Engineering time
    Owner
    Founder (engineering)

    Smallest test

    Add one number (accepted hours to date) to the homepage.

    Why it works

    Add a small, server-rendered stats band to the homepage and /explore, updated monthly from the pipeline: accepted hours to date, hours this month, countries, distinct homes and workplaces, and the task mix. Post a short monthly note with the figures on LinkedIn and X.

    Regular, specific numbers make the scale claim credible, give buyers something to cite internally, and create a monthly reason to post.

    Risks and how to handle them

    • Numbers look small next to rivals

      Lead with what rivals lack: distinct homes and workplaces.

    Builds on

    Evidence

    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Competitor researchClaru homepage: "Training data for physical AI"; "34,000+ contributors in 70+ countries", "100+ active datasets", "5+ frontier lab partnerships"; a "Book a Call" button and first delivery "in days, not months".Source
  • 18

    The Strategic Operations Intern post describes managing contractors "across countries and time zones", shipping equipment, onboarding, monitoring activity and following up on issues, all tracked "in spreadsheets and internal tools". The application already captures structured data (country, phone, hours, mount, Wi-Fi). As collectors scale, spreadsheet follow-ups become the limit on growth.

    Medium impactMedium effort First result in 1 month Strategic Operations Intern

    First action

    Define the seven stages and map current collectors to them.

    How to measure it

    Applicants becoming active collectors

    Now
    –
    Target
    +30%
    By
    90 days

    Steps

    1. 1

      Define stages

      Seven stages with exit criteria.

      Strategic Operations Intern 1 day

    2. 2

      Choose a tool

      One the team already uses, or a simple CRM.

      Founder (go-to-market) 1 day

    3. 3

      Pipe in applications

      Form fields create records automatically.

      Founder (engineering) 2 days

    4. 4

      Sync app events

      First upload and approved hours update the stage.

      Founder (engineering) 1 week

    5. 5

      Add alerts and templates

      Stuck for 7 days triggers a message.

      Strategic Operations Intern 3 days

    6. 6

      Weekly report

      Stage conversion and time in stage.

      Strategic Operations Intern weekly

    First result
    1 month
    Cost
    Low-cost CRM seat plus setup time
    Owner
    Strategic Operations Intern

    Smallest test

    Map last month's applicants to the stages in a single sheet first.

    Why it works

    Set up a lightweight pipeline (a CRM or ticketing tool) with stages: applied, matched, onboarded, first upload, five approved hours, active, lapsed. Pipe application fields in automatically, add alerts for collectors stuck at a stage, and template messages per stage. Report conversion between stages weekly.

    Seeing where collectors stall shows which fix matters (onboarding, mounts, QA), and automation lets one part-time person manage thousands of collectors.

    Risks and how to handle them

    • Personal data spread across tools

      Keep payout details out of the CRM and limit access.

    Evidence

    • HiringStrategic Operations Intern (YC job board, Berkeley or remote US, $20-50 an hour, about 20 hours a week): manages contractors "across countries and time zones", equipment shipments and inventory, onboarding, monitoring contractor activity and following up on issues, and tracking data in spreadsheets and internal tools.Source
    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source
    • Your websiteSupport page: help for the Asimov Capture app (recordings, uploads, review status, payouts, account deletion) through a single support email inbox; no FAQ or help articles.Source

Big bets 4

Larger plays that can change the trajectory.

  • 19

    Asimov already holds what a provenance record needs: each collector accepts a dated Contributor Agreement (updated June 20, 2026) under which accepted data can be licensed, the Privacy Policy records consent status, and QA logs accept and reject decisions. Claru advertises datasheets with its deliveries. Lab buyers increasingly have to show where training data came from and that people agreed, and the cleaning homes (idea 8) and partner workplaces add consent layers of their own.

    High impactHigh effort First result in 1 quarter Founder

    First action

    Ask two customers what their legal team asks about data origin.

    How to measure it

    Buyer legal review time

    Now
    –
    Target
    -50%
    By
    2 quarters

    Steps

    1. 1

      Interview buyers

      What their legal and ethics reviews require.

      Founder (go-to-market) 2 weeks

    2. 2

      Design the schema

      Fields drawn from agreements, consent, QA and capture metadata.

      Founder (engineering) 1 week

    3. 3

      Link the records

      Each clip to its agreement version and consent record.

      Founder (engineering) 1 month

    4. 4

      Generate automatically

      Datasheet built with every export.

      Founder (engineering) 1 month

    5. 5

      Legal summary

      Counsel-reviewed PDF template.

      Outside counsel 2 weeks

    6. 6

      Launch

      On the spec sheet and in every delivery.

      Founder (go-to-market) 1 week

    7. 7

      Audit

      Sample datasheets against source records each quarter.

      QA lead quarterly

    First result
    1 quarter
    Cost
    About two months of engineering plus legal review
    Owner
    Founder (engineering)

    Smallest test

    Hand-build one datasheet for the next delivery and ask for feedback.

    Why it works

    Generate a machine-readable datasheet per delivery: agreement versions, consent counts, countries and environment types, device mix, capture sensors, QA acceptance rate, privacy steps applied, and any partner or cleaning-client consent. Include a summary PDF for legal teams. Make it a standard part of every delivery and a headline on the spec sheet.

    When data vendors look alike, the one whose data clears legal review fastest wins the contract, and provenance is hard for newcomers to fake later.

    Risks and how to handle them

    • Gaps in older records

      Start with new collections and flag legacy data separately.

    Builds on

    Evidence

    • Your websitePrivacy Policy (updated June 20, 2026): covers "approved adult collectors" using the Asimov Capture app; collects first-person video, microphone audio only when an assignment needs narration or the collector opts in, and ARKit pose and LiDAR/depth where supported; accepted data may be "sold, licensed, sublicensed, shared"; it does not mention face blurring and does not address cleaning clients or partner businesses.Source
    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
    • Your websiteCleaning page (/cleaning): "Bay Area Only"; "Your first cleaning is completely free with no commitment"; "Starting at $10/hr" after the first free 3-hour cleaning; badges "Privacy reviewed" and "Avoid faces & personal info". It does not mention the recording kits or how footage from a cleaning is used.Source
    • Competitor researchClaru publishes an "Asimov (YC W26) alternatives" comparison page. It lists Claru's delivery formats (WebDataset, HDF5, RLDS, with manifests, checksums and datasheets) and pilot datasets in under a week; for Asimov, delivery formats are not specified.Source
  • 20

    Asimov already structures its supply: every collector application records country, city, phone model, hours a week, recording locations and task types. Buyers, meanwhile, have no way to say what they want on the site. Matching the two in software is what turns a services business into a platform, and the YC page describes a goal of collecting from "thousands of different people, homes, and workplaces".

    High impactHigh effort First result in 2 quarters Founder

    First action

    Build a read-only supply estimate from current collector data.

    How to measure it

    Portal-originated orders

    Now
    –
    Target
    30% of new orders
    By
    3 quarters

    Steps

    1. 1

      Model supply

      Active collectors by country, device, task and weekly hours.

      Founder (engineering) 2 weeks

    2. 2

      Estimate turnaround

      From historic acceptance rates and hours.

      Founder (engineering) 2 weeks

    3. 3

      Build the spec form

      Tasks, environments, devices, hours.

      Founder (engineering) 1 month

    4. 4

      Price ranges

      Rules from pilot pricing.

      Founder (go-to-market) 1 week

    5. 5

      Create assignments

      Orders become app assignments.

      Founder (engineering) 1 month

    6. 6

      Beta with customers

      Three existing buyers first.

      Founder (go-to-market) 1 month

    7. 7

      Open it

      Link from the request form and spec sheet.

      Founder (go-to-market) 1 week

    First result
    2 quarters
    Cost
    One to two quarters of engineering
    Owner
    Founder (engineering)

    Smallest test

    Send buyers a hand-made supply estimate within a day of each request.

    Why it works

    Let a buyer sign in, choose tasks, environments, countries, device needs (for example, depth-capable phones) and hours, and see an instant estimate of available supply, turnaround and price range. Confirmed orders become assignments in the collector app automatically. Start with the pilot packages (idea 14) and expand to custom specs.

    Buyers get answers in minutes instead of weeks, and Asimov can serve many more teams without adding sales staff.

    Risks and how to handle them

    • Estimates prove wrong

      Show ranges and confirm by hand during the beta.

    Builds on

    Evidence

    • Your websiteCollector page (/collectors): "Record real-world tasks with your phone"; pay "$5-15/hour", "Exact rates are shown before you record", and only "approved recording time is eligible for payment"; the form asks for country, city, exact phone model, hours a week (1-5 up to 40+), mount, Wi-Fi, recording locations and task types, plus an optional referral code; headbands are not shipped "during early testing"; there is no FAQ.Source
    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Y CombinatorYC company profile and launch post: Winter 2026 batch, team of 3, San Francisco; "The largest robotics labs in the world are already our customers"; delivers "thousands of hours a day to leading labs"; the launch post asks robotics companies, businesses and contributors to get in touch through a founders' email address that does not appear on tryasimov.ai.Source
  • 21

    Asimov's search description already advertises "teleoperation services" and "remote operation", but no page or offer exists. The homepage's case is that lab teleoperation is scarce: "thousands of hours at best". Asimov has what that market lacks: a global, paid, remote workforce already trained to follow detailed task instructions, recruited "anywhere in the world".

    High impactHigh effort First result in 1 quarter Founder

    First action

    Ask current customers whether they would buy remote operator hours, and for which robot.

    How to measure it

    Paid operator hours a month

    Now
    –
    Target
    1,000
    By
    3 quarters

    Steps

    1. 1

      Validate demand

      Five customer conversations on robots, tasks and price.

      Founder (go-to-market) 2 weeks

    2. 2

      Pick operators

      Top collectors by acceptance and reliability.

      Strategic Operations Intern 1 week

    3. 3

      Train on one robot

      With the pilot customer's interface and safety rules.

      Founder (engineering) 1 month

    4. 4

      Set shifts and QA

      Coverage across time zones and demonstration review.

      QA lead 2 weeks

    5. 5

      Run the pilot

      One task family, fixed hours and price.

      Founder (go-to-market) 1 month

    6. 6

      Publish the page

      A /teleoperation page matching the description.

      Founder (go-to-market) 1 week

    7. 7

      Scale

      Second robot or second customer.

      Founder (go-to-market) 1 quarter

    First result
    1 quarter
    Cost
    Operator pay and training time, funded by the pilot customer
    Owner
    Founder (go to market)

    Smallest test

    Offer ten trained operators to one customer for a two-week paid trial.

    Why it works

    Pilot a remote-operation service: the most reliable collectors, trained on a customer's robot interface, run demonstrations or supervise deployed robots in shifts across time zones. Price per operator hour, and pair it with egocentric data of the same tasks so labs get human video and robot demonstrations together. Start with one customer's robot and one task family.

    It sells a second product to the same buyers, uses the supply Asimov already manages, and gives labs matched human and robot data that is hard to get elsewhere.

    Risks and how to handle them

    • Muddies the message against teleoperation

      Position it as "human data plus remote operation", not a replacement for capture.

    • Robot safety

      Follow each customer's safety protocol and start in labs, not public spaces.

    Builds on

    Evidence

    • Your websiteThe homepage description used by search results and link previews says "End-to-end data and teleoperation services for embodied AI. From egocentric capture to remote operation.", while the page argues that teleoperation gives "thousands of hours at best, usually in controlled labs" and sells human activity capture; no page describes a teleoperation or remote-operation service.Source
    • HiringData Collection Specialist (YC job board, remote, "anywhere in the world"): "$5-15/hr base pay" rising "up to $30/hr after your first 5 hours collected", plus volume bonuses and unlimited hours; the post says no audio is collected and that faces and personal information are automatically blurred.Source
  • 22

    Asimov's cleaning company, "Bay Area Only", cleans for over 100 founders, investors and students each weekend, with a free first three-hour cleaning and prices from $10 an hour. It is the one source where Asimov controls the workers, the kits and the environments. Every other home in the dataset depends on individual collectors. A second city adds new kinds of homes, a new revenue line and a template to repeat.

    Medium impactHigh effort First result in 1 quarter Founder

    First action

    Compare cost per accepted hour: cleaning vs crowd collectors in the Bay Area.

    How to measure it

    Cleaning-city cost per accepted hour

    Now
    –
    Target
    Below crowd collection
    By
    3 quarters

    Steps

    1. 1

      Run the numbers

      Bay Area cleaning revenue, cost and accepted hours.

      Founder (go-to-market) 1 week

    2. 2

      Pick the city

      Housing mix, collector applicants, cleaning prices.

      Strategic Operations Intern 2 weeks

    3. 3

      Recruit cleaners

      Five to ten local cleaners with kits.

      Strategic Operations Intern 1 month

    4. 4

      Launch the offer

      Free first cleaning, with the disclosure from idea 8.

      Founder (go-to-market) 2 weeks

    5. 5

      Seed demand

      University and startup communities.

      Strategic Operations Intern ongoing

    6. 6

      Compare costs

      Monthly against crowd collection.

      Founder (go-to-market) monthly

    7. 7

      Decide

      Scale, adjust or stop at 90 days.

      Founder (go-to-market) 90 days

    First result
    1 quarter
    Cost
    Cleaner wages and kits for one quarter, partly offset by cleaning revenue
    Owner
    Founder (go to market)

    Smallest test

    Offer free cleanings in one neighbourhood of the new city for a month.

    Why it works

    Choose a city with a different housing mix from the Bay Area and a pool of collectors already applying from it. Recruit cleaners locally, launch with the same free-first-cleaning offer aimed at student and startup communities, and track cost per accepted hour against crowd collection. Only expand if cleaning revenue plus data value beats the crowd cost.

    The cleaning model pays for itself and produces consented, consistent home data that buyers value for diversity. Proving it twice makes it a repeatable supply engine.

    Risks and how to handle them

    • Running a service business distracts the founders

      Set a cost-per-hour stop rule before launch.

    Builds on

    Evidence

    • Your websiteHomepage, businesses and cleaning: "We pay your workers' salaries while they do their normal jobs wearing lightweight recording equipment", with dataset revenue share "For larger partnerships"; "We also own a growing cleaning company, offering free and reduced-price cleanings. Our cleaners wear our kits"; it cleans apartments for "over 100 founders, investors, and students" in the Bay Area each weekend.Source
    • Your websiteCleaning page (/cleaning): "Bay Area Only"; "Your first cleaning is completely free with no commitment"; "Starting at $10/hr" after the first free 3-hour cleaning; badges "Privacy reviewed" and "Avoid faces & personal info". It does not mention the recording kits or how footage from a cleaning is used.Source

Wildcards 2

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

  • 23

    Asimov shows no sample data anywhere: the homepage has no clip or download, and /explore is blank without scripts. Robotics researchers pick datasets they can try, and today's students and postdocs become tomorrow's buyers at the labs Asimov sells to. Asimov's open operations role is based in Berkeley, a university town, which makes a first academic release easy to seed in person.

    Medium impactMedium effort First result in 1 quarter Founder

    First action

    Pick 3-5 hours of the cleanest, fully consented clips.

    How to measure it

    Requests citing the research sample

    Now
    –
    Target
    10
    By
    3 quarters

    Steps

    1. 1

      Select clips

      Varied tasks, all with full consent for research release.

      QA lead 1 week

    2. 2

      Write the licence

      Research-only, no redistribution.

      Outside counsel 2 weeks

    3. 3

      Write the note

      Capture, annotation, QA and known limits.

      Founder (engineering) 1 week

    4. 4

      Host it

      Download with a short sign-up for contact.

      Founder (engineering) 3 days

    5. 5

      Announce

      University robotics groups and dataset hubs.

      Founder (go-to-market) 1 week

    6. 6

      Follow up

      Offer full data to labs that download.

      Founder (go-to-market) ongoing

    First result
    1 quarter
    Cost
    A few hours of data plus legal review
    Owner
    Founder (engineering)

    Smallest test

    Share a one-hour sample privately with three university labs.

    Why it works

    Publish a few hours of consented, annotated household-task data under a research-only licence, with a short technical note on capture and annotation, a citation, and a clear path to commercial data. Announce it to university robotics groups and on the usual dataset hubs. Measure downloads, citations and inbound requests from labs.

    A research sample spreads by citation and word of mouth among exactly the engineers who later choose vendors, at the cost of a few hours of data.

    Risks and how to handle them

    • Undercutting paid sales

      Keep it small and research-only.

    Builds on

    Evidence

    • Your websiteHomepage (checked twice on November 10, 2026): headline "Human intelligence for robotics", "Backed by Y Combinator", "Collecting across 3 continents". The buyer buttons "Request Data", "Contact Us" and "Partner with Us" and the footer "Contact" are not links in the page as served, and the page has no email address, form or calendar link. Robotics teams are promised "clean, annotated datasets ready for training" ("3D body pose, depth maps, semantic labels, activity segmentation"), but no sample, dataset size, file format or price is shown, and the 3D viewer reads "Loading 3D viewer...".Source
    • Your websiteThe "Explore Data" page (/explore), linked from the footer, shows only the word "Asimov" to a reader that does not run scripts: no datasets, hours, environments or way to request data.Source
    • HiringStrategic Operations Intern (YC job board, Berkeley or remote US, $20-50 an hour, about 20 hours a week): manages contractors "across countries and time zones", equipment shipments and inventory, onboarding, monitoring contractor activity and following up on issues, and tracking data in spreadsheets and internal tools.Source
  • 24

    Pay depends on QA: estimates are "for planning only" and a payable amount exists only after QA accepts a session. Payouts also need Stripe onboarding and compliance checks. The support page's main topics are recordings, uploads, review status and payouts, all through one email inbox. When collectors can't see where their money stands, they write to support or quit.

    Medium impactMedium effort First result in 1 month Founder

    First action

    Count last month's support emails about review status and payouts.

    How to measure it

    Collectors active 30 days after first upload

    Now
    –
    Target
    +20%
    By
    2 quarters

    Steps

    1. 1

      Measure the load

      Support emails by topic.

      Strategic Operations Intern 2 days

    2. 2

      Design the screen

      Session status, minutes, pending and paid.

      Founder (engineering) 3 days

    3. 3

      Expose QA status

      API from the review system to the app.

      Founder (engineering) 1 week

    4. 4

      Show payout blockers

      Missing onboarding or checks, with a fix link.

      Founder (engineering) 3 days

    5. 5

      Add notifications

      Push on review and payout.

      Founder (engineering) 2 days

    6. 6

      Measure

      Support volume and 30-day retention.

      Strategic Operations Intern monthly

    First result
    1 month
    Cost
    About three weeks of engineering
    Owner
    Founder (engineering)

    Smallest test

    Email a weekly status summary to 50 collectors and compare their retention.

    Why it works

    Add an earnings screen: each session's status (in review, accepted, reshoot, rejected with reason), accepted minutes, rate applied, pending and paid amounts, and the next payout date, plus any onboarding step blocking a payout. Send a push when a session is reviewed.

    Seeing pay accrue keeps gig workers active, and self-serve status cuts the support load for a three-person team.

    Risks and how to handle them

    • Rejections become more visible and discourage people

      Pair each rejection with a fix tip (idea 10).

    Builds on

    Evidence

    • Your websiteContributor Agreement (updated June 20, 2026): "A payable amount is created only after Asimov QA accepts the uploaded data"; sessions can be rejected or sent for reshoot for "wrong task, insufficient length, bad camera angle, hands not visible, privacy issues", or duplicate or scripted activity; participation is limited to an "allowed launch country/region", with no list of countries; no referral program is described.Source
    • Your websiteSupport page: help for the Asimov Capture app (recordings, uploads, review status, payouts, account deletion) through a single support email inbox; no FAQ or help articles.Source
    • HiringData Collection Specialist (YC job board, remote, "anywhere in the world"): "$5-15/hr base pay" rising "up to $30/hr after your first 5 hours collected", plus volume bonuses and unlimited hours; the post says no audio is collected and that faces and personal information are automatically blurred.Source
More example playbooks and reviews

More growth playbooks

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

  • Interfere growth strategyProduction monitoring that turns errors, traces, sessions and releases into prioritized problems, with affected users, likely root cause…22 growth ideas
  • Workflows.io growth strategyA go-to-market agency for B2B tech companies that combines AI workflows with human expertise across four service lines: LinkedIn Content…24 growth ideas
  • SupplyHouse.com growth strategyAn online distributor of plumbing, heating, HVAC and electrical supplies, with over 180,000 products, four US fulfillment centers, an app…23 growth ideas
  • Almedia growth strategyA Berlin adtech company that runs Freecash, a rewards app where users earn money for playing games, trying apps, taking surveys and…21 growth ideas
  • Clasp growth strategyLoan-Linked Hiring: healthcare employers commit to repaying clinical students' loans over a multi-year term in exchange for tenure (Clasp…21 growth ideas
  • Nubank growth strategyA Brazilian digital bank: credit card, digital account, Pix, savings boxes, loans, insurance, investments and the NuCoin benefits program…20 growth ideas
  • Zellbury growth strategyA Pakistani value-fashion brand selling ready-to-wear, unstitched fabric, western wear and accessories for women, men and kids through its…22 growth ideas
  • Firecrawl growth strategyA web data API for AI agents (search, scrape, crawl, map and browser interaction that return LLM-ready markdown or JSON), open source…24 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. Asimov did not commission, review or endorse it, and Programmable is not affiliated with Asimov. 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 Asimov runs its business.

Something wrong, or do you represent Asimov? 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 309 ideas

    Give buyers a working request path, a spec sheet and a readable Explore page; fix homepage speed, the search description and share previews; one privacy promise, cleaning-client disclosure and the visible pay ramp.

  2. Days 31 to 609 ideas

    Scale and keep collectors (QA onboarding, referrals, demand board, pipeline CRM); open the partner and pilot offers; publish the comparison page, a customer story and the monthly delivery ledger.

  3. Days 61 to 906 ideas

    Commit to the big bets: provenance datasheets, a self-serve request portal, the teleoperation line and a second cleaning city; try the research sample and payout transparency.

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 data from real homes and workplaces instead of controlled labs. The most cited problem: teleoperation gives "thousands of hours at best".

  • An end-to-end pipeline: hardware, collection, QA and post-processing
  • Narrow datasets from controlled settings
  • In their words: “egocentric data”

Market

Competitive landscape

2 competitors compared: Claru and In-house teleoperation at robotics labs.

  • Gap to exploit: A buyer-facing spec sheet and viewable sample for household egocentric data
  • Claru: Broad focus across games, driving and cinematic data rather than household tasks
  • In-house teleoperation at robotics labs: In Asimov's words, "thousands of hours at best, usually in controlled labs"

Hiring signals

Asimov's two open roles are both on the supply side.

  • Data Collection Specialist: Global remote collectors; up to $30/hr after 5 hours; privacy claims that differ from the policy. Ideas 7, 9, 21, 24.
  • Strategic Operations Intern: Contractor operations across time zones, tracked in spreadsheets. Ideas 10, 11, 12, 18.

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