ProgrammableMarketing/discoveredmaterials.com (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 · 2
  • How we can help
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. Discovered Materials
Discovered Materials logo

growth playbook · Nov 10, 2026

discoveredmaterials.com (opens in a new tab)

Discovered Materials growth playbook

Diagnosis

Discovered Materials has unusually strong science marketing for a four-person seed company: a Launch HN with 162 points, TechCrunch and CNBC coverage, an open benchmark reviewed by experts from IMEC, IBM, Micron, Stanford and Cambridge, a 36x Quantum ESPRESSO speedup and a $1 million prize.

What it lacks is the path from that attention to a buyer. The site has no page for chipmakers or chemical companies, the per-challenge prize amounts sit only in the legal rules, the 500+ released materials have no download link, the best research post is missing from the sitemap, and the homepage has no signup.

This playbook makes the prize and the benchmark legible first, gives industry a door to walk through, then turns the lab, the verifiers and the HBM model into proof a risk-averse industry can buy.

Growth Score

6/10

Biggest Opportunity

Turn the attention from the benchmark, the prize and the open code into a short list of semiconductor and chemical design partners, by showing them what they can buy and proving the lab can make what the agents find.

Biggest Risk

Semiconductor buyers are risk averse and well-funded rivals (CuspAI with Applied Materials, Samsung and Lam Research as partners; Periodic Labs with its own labs) are chasing the same heat and superconductivity problems; if only one of 500+ candidates can be made, credibility rests on the lab delivering.

01

Start here

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

  1. 1

    Add a page for chipmakers and chemical companies: what they can buy and how to start

    Write the three offers in one sentence each.

    High impact First result in 1 month
  2. 2

    Put each challenge's prize and "no deadline" status on its own page, not only in the Official Rules

    Copy the five maximum amounts from the Official Rules into the header of each challenge page.

    High impact First result in 1 week
  3. 3

    Publish the 500+ released materials as a downloadable dataset linked from the benchmark

    Export the 500+ structures and their verified properties to one table with CIF files.

    High impact First result in 2 weeks

02

Your idea portfolio

22 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 · 8Fill-ins · 8Reconsider · 3
Effort →

Showing 22 of 22 ideas

This week 6

Quick wins with a clear first action.

  • 01

    The Grand Challenges page promises "$1 million in prize money" but shows no amount per problem, and every challenge reads "Open" next to "Nov 2026", which looks like a closing month. The real terms are only in the Official Rules: up to $400,000 for cuprate superconductivity, $200,000 each for the exchange-correlation and Si/oxide problems, $100,000 each for driven systems and interfacial heat transport, and a Competition that "has no deadline". The rules also use different names from the page titles, so a reader cannot easily match "driven systems" to "Tuning Materials with Light". TechCrunch's launch story gave no breakdown. The strongest hook of the whole programme, $400,000 for a theory of cuprate superconductivity, is buried in a legal page.

    High impactLow effort First result in 1 week Co founder & CEO

    First action

    Copy the five maximum amounts from the Official Rules into the header of each challenge page.

    How to measure it

    Challenge-page signups

    Now
    –
    Target
    2x the current weekly rate
    By
    60 days

    Steps

    1. 1

      Align the names

      Use one name per challenge on the page, the index and the rules; show the rules' name in brackets.

      Co-founder & CEO 1 hour

    2. 2

      Add the prize header

      Maximum prize, status, what to submit and how it is verified, at the top of each page.

      Co-founder & CTO 3 hours

    3. 3

      Fix the status label

      Replace "Nov 2026" with "Open until solved" and a separate "updated" date.

      Co-founder & CTO 30 minutes

    4. 4

      Write the FAQ

      Eligibility, team entries, IP license, patents before submitting, verification and split prizes.

      Co-founder & CEO 1 day

    5. 5

      Publish a fact sheet

      One paragraph with the split, the rules link and a contact for press.

      Co-founder & CEO 2 hours

    6. 6

      Send it to the list

      A short note to mailing-list subscribers with the amounts per challenge.

      Co-founder & CEO 1 hour

    7. 7

      Track

      Signups and submission emails per challenge page each week.

      Co-founder & CTO weekly

    First result
    1 week
    Cost
    $0, a day of founder time
    Owner
    Co founder & CEO

    Smallest test

    Add the amount and "Open until solved" to the cuprate page today and compare its signups with the other four for two weeks.

    Why it works

    Add a short header to each of the five challenge pages: the maximum prize, the status ("Open until solved: the first verified solution wins"), what to submit, how verification works (hidden test cases and the 90-day verification window) and the rules' name for the problem. On the index, show the amount next to each challenge and replace the bare "Nov 2026" with "Open, updated Nov 2026". Add a five-question FAQ and a one-paragraph fact sheet with the split for journalists and newsletters.

    Serious research groups weigh months of effort against the payout and the finish line before they commit. A visible amount and a clear "no deadline, first verified solution wins" rule removes the false sense that the window is closing, makes each page worth sharing on its own, and gives coverage the right numbers.

    Ready to use

    ChecklistChallenge page header

    [ ] Maximum prize ("up to $X")
    [ ] Status: Open until solved; first verified solution wins
    [ ] Same name as the Official Rules (or the rules' name in brackets)
    [ ] What to submit: method + code or derivations
    [ ] How it is verified: published targets, hidden cases, 90-day verification
    [ ] IP note: non-confidential; file patents before submitting
    [ ] Link to Official Rules and FAQ
    [ ] Mailing-list signup

    Risks and how to handle them

    • Amounts read as guaranteed

      Say "up to" and link the rules on verification and split prizes.

    • More low-quality submissions

      Put the reproduction targets and hidden-test rule above the email address.

    Evidence

    • Your websiteGrand Challenges page (checked November 9, 2026): "We commit $1 million in prize money to solutions to these challenges"; five challenges, each listed as "Open" next to "Nov 2026"; entry is by email, and "Each problem page says what a submission must include"; no per-challenge prize amounts, no web submission form, no FAQ and no leaderboard.Source
    • Your websiteOfficial Rules: maximum prizes of $400,000 (cuprate superconductivity), $200,000 (exchange-correlation functional), $200,000 (Si/oxide interfaces), $100,000 (driven systems) and $100,000 (interfacial heat transport); the Competition "has no deadline"; the first eligible entrant meeting all success criteria generally wins; entrants keep their IP but grant a perpetual, royalty-free, sublicensable license; submissions are not confidential, so entrants who want patents should file first; submissions that cannot be verified within 90 days of a request are disqualified.Source
    • Your websiteChallenge pages: titles such as "Breaking the Speed–Accuracy Barrier in Quantum Chemistry" and "Tuning Materials with Light" differ from the names used in the rules ("exchange-correlation functional", "driven systems"). The heat transport page asks entrants to reproduce published Al/Si, Al/GaN, Cu/TiC/diamond and Au/PMMA/Si results and to beat AMM/DMM on hidden interfaces, with a method and working code. Neither it nor the cuprate page states a prize amount or a deadline; both show "Open" and "Updated Nov 2026".Source
    • PressTechCrunch on the prize (November 2026): a $1 million pool for five open problems, shaped with an advisory group of academics including a Stanford professor; the challenges are "goalposts" showing "whether the models are climbing in this domain"; the company partners with chip companies on research projects; the article gives no per-problem breakdown.Source
  • 02

    The benchmark page says "over 500 materials" were discovered and released publicly, and Launch HN said hundreds were released. But the benchmark page links no GitHub, Hugging Face or paper, and the company's GitHub has one repository, the Quantum ESPRESSO mirror. A researcher who wants the structures has nowhere obvious to get them.

    High impactLow effort First result in 2 weeks Co founder & CTO

    First action

    Export the 500+ structures and their verified properties to one table with CIF files.

    How to measure it

    Dataset downloads

    Now
    –
    Target
    1,000 in 90 days
    By
    90 days

    Steps

    1. 1

      Export

      Structures, properties, verifier and run for each material.

      Co-founder & CTO 2 days

    2. 2

      Write the card

      What each column means, the targets, limits and known issues such as reward-hacked entries removed.

      Co-founder & CEO 1 day

    3. 3

      Choose a license

      An open data license that matches the company's IP plans.

      Co-founder & CEO 1 day

    4. 4

      Publish

      A GitHub repository plus a Hugging Face dataset, with a DOI.

      Co-founder & CTO 1 day

    5. 5

      Link it

      From the benchmark page, homepage and research index.

      Co-founder & CTO 1 hour

    6. 6

      Announce

      Post on X, LinkedIn and the materials-informatics communities.

      Co-founder & CEO 1 day

    First result
    2 weeks
    Cost
    $0
    Owner
    Co founder & CTO

    Smallest test

    Post the table as a CSV in a new repository and link it from the benchmark page.

    Why it works

    Put the materials in a versioned repository and a dataset card: structures (CIF), predicted thermal conductivity, dielectric constant and moduli, which verifier produced each value, the model run that found it, and a synthesis-plausibility flag. Give it a DOI, a citation line and a license, and link it from the benchmark page, the homepage and the research index.

    A citeable dataset is how materials researchers find, reuse and credit work, and every citation is a link back. It also backs the feasibility honesty that HN commenters praised with data anyone can check.

    Risks and how to handle them

    • Giving away patentable leads

      Release only what is already public and keep pending candidates back.

    • Errors found by others

      Welcome them in the card and keep a changelog; that is the point.

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • GitHubThe discovered-materials GitHub organization has one public repository, dm_qe ("Discovered Materials' Quantum Espresso mirror", 13 stars, GPL-2.0); there is no repository for the benchmark or the released materials.Source
    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
  • 03

    The founders say the business is licensing IP and selling the harness to semiconductor and chemical companies, and TechCrunch says they already partner with chip companies on research. Yet the site's only calls to action are "Talk to us", an email link, and "Join us", the jobs page; the sitemap has no page for industry partners, services or use cases. CuspAI, by contrast, invites industry, synthesis and data partners to sign up.

    High impactMedium effort First result in 1 month Co founder & CEO

    First action

    Write the three offers in one sentence each.

    How to measure it

    Industry enquiries with a stated target property

    Now
    –
    Target
    10 in 90 days
    By
    90 days

    Steps

    1. 1

      Define offers

      License, scoped programme, in-house harness.

      Co-founder & CEO 1 day

    2. 2

      Gather proof

      Lab-validated thermal interface results, benchmark, reviewers, open code.

      Co-founder & CTO 2 days

    3. 3

      Write IP terms

      Plain-English confidentiality and ownership for each offer.

      Co-founder & CEO 2 days

    4. 4

      Build the form

      Company type, application, target property, timeline.

      Co-founder & CTO 1 day

    5. 5

      Publish and link

      From the homepage navigation and the benchmark page.

      Co-founder & CTO 2 hours

    6. 6

      Share with partners

      Send to existing research partners and investors for introductions.

      Co-founder & CEO 1 week

    7. 7

      Review

      Which offer gets asked about most.

      Co-founder & CEO monthly

    First result
    1 month
    Cost
    Founder time
    Owner
    Co founder & CEO

    Smallest test

    Replace "Talk to us" with "Work with us" pointing to a one-screen page with the three offers and a form.

    Why it works

    Write one page, /industry, for R&D and materials leads at chipmakers, memory makers and chemical suppliers. Show three ways to work together (license a discovered material, run a scoped discovery programme on your target properties, use the harness in-house), what a first 90 days looks like, how IP and confidentiality work, and proof: the lab results, the benchmark, the expert reviewers and the Quantum ESPRESSO speedup. End with a short form asking for the target property and application.

    Risk-averse buyers need to see a defined offer and terms before they reply to a startup. A form that asks for the property target qualifies leads and starts the scoping conversation.

    Risks and how to handle them

    • Committing to an offer too early

      Call it a design-partner programme and keep terms indicative.

    • Partners' names

      Show only partners who have agreed to be named.

    Evidence

    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
    • Competitor researchRivals' own homepages: CuspAI runs an "AI Materials Foundry" with industry partners including Applied Materials, Samsung, Lam Research and Tokyo Electron, and invites industry, synthesis and data partners to sign up; Citrine sells an "AI platform for materials and chemicals development" with published case studies (Dorfner: an 84% reduction in formulation time); Periodic Labs builds models and autonomous labs, targets higher-temperature superconductors and helps a semiconductor manufacturer with chip heat dissipation.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /research/, /benchmark/, /news/, seven Grand Challenges URLs, one model page (/models/hbm_thermal_resistance/, last modified July 13, 2026), terms and privacy. The Quantum ESPRESSO post is not listed, there are no pages for industry partners, services, use cases or careers, and the research index lists two undated posts.Source
  • 04

    The challenge pages, the benchmark and the Quantum ESPRESSO post all have a mailing-list signup, but the homepage, where TechCrunch, CNBC and HN readers arrive, has none. Its only actions are an email link and a jobs link, so most first-time visitors leave with no way to follow the work.

    Medium impactLow effort First result in 1 week Co founder & CTO

    First action

    Copy the existing signup component onto the homepage.

    How to measure it

    Homepage visitor-to-subscriber rate

    Now
    –
    Target
    3%
    By
    30 days

    Steps

    1. 1

      Reuse the form

      Same list provider as the challenge pages.

      Co-founder & CTO 1 hour

    2. 2

      Write the promise

      What subscribers get and how often.

      Co-founder & CEO 30 minutes

    3. 3

      Add a role field

      Optional: industry, academic, investor, candidate.

      Co-founder & CTO 1 hour

    4. 4

      Tag the source

      Homepage, challenge, benchmark or research.

      Co-founder & CTO 30 minutes

    5. 5

      Add to the research index

      The index has no signup either.

      Co-founder & CTO 30 minutes

    6. 6

      Measure

      Signups per page per week.

      Co-founder & CTO weekly

    First result
    1 week
    Cost
    $0
    Owner
    Co founder & CTO

    Smallest test

    Add the form today and compare signups with the challenge pages after two weeks.

    Why it works

    Add one signup block below the hero and above the team section: "New materials, lab results and challenge updates, about once a month", with an optional field for role (industry, academic, investor, candidate). Use the same list as the other pages and tag the source.

    The homepage gets the most first visits; a list is the one channel the company owns, and the role tag tells the founders which audience is growing.

    Risks and how to handle them

    • Clutter on a clean page

      One line and one field, styled like the rest.

    Evidence

    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
    • Your websiteNews page: headlines only, linking TechCrunch (August 10, 2026), a CNBC video, The Economic Times, Business Standard and an MTSLive stream; no company-written announcements.Source
    • Your websiteGrand Challenges page (checked November 9, 2026): "We commit $1 million in prize money to solutions to these challenges"; five challenges, each listed as "Open" next to "Nov 2026"; entry is by email, and "Each problem page says what a submission must include"; no per-challenge prize amounts, no web submission form, no FAQ and no leaderboard.Source
  • 05

    The Quantum ESPRESSO post is the company's most reusable result: phonon calculations about 36x faster, hybrid-functional DFT about 9x, a suite cut from 3.3 hours to under 4 minutes, with open code. But it is missing from the 14-URL sitemap, shows no date or author, and the research index lists two undated posts. Researchers cite dated, authored work; search engines find sitemap pages first.

    Medium impactLow effort First result in 2 weeks Co founder & CTO

    First action

    Add the post's URL to the sitemap.

    How to measure it

    Search impressions for the Quantum ESPRESSO post

    Now
    –
    Target
    5x
    By
    60 days

    Steps

    1. 1

      Fix the sitemap

      Add the post and generate the sitemap from the build so new posts are included.

      Co-founder & CTO 1 hour

    2. 2

      Add dates and authors

      On both posts and the index.

      Co-founder & CTO 1 hour

    3. 3

      Add a citation

      Suggested citation and a BibTeX block.

      Co-founder & CEO 1 hour

    4. 4

      Add structured data

      Article and code markup linking dm_qe.

      Co-founder & CTO 2 hours

    5. 5

      Cross-link

      From the homepage and the speed-accuracy challenge.

      Co-founder & CTO 1 hour

    6. 6

      Submit

      Request indexing in Search Console.

      Co-founder & CTO 15 minutes

    First result
    2 weeks
    Cost
    $0
    Owner
    Co founder & CTO

    Smallest test

    Add the sitemap entry and date today; check indexing in a week.

    Why it works

    Add the post to the sitemap, add a publication date, authors and a suggested citation to it and to the benchmark, show dates on the research index, and add Article and SoftwareSourceCode structured data that points to the dm_qe repository. Link the post from the homepage and from the speed-accuracy challenge.

    Dates and authors make work citeable and trustworthy, and a sitemap entry gets the page indexed for the people searching for faster Quantum ESPRESSO, who are exactly the computational materials scientists the company wants to reach.

    Risks and how to handle them

    • Future posts missed again

      Build the sitemap from the content folder.

    Evidence

    • Your websiteResearch post "Accelerating Quantum ESPRESSO by 36x": an AI-driven optimization loop made DFPT phonon calculations about 36x faster and hybrid-functional DFT about 9x faster on GPUs; the team's benchmark suite fell from 3.3 hours to under 4 minutes; the code is released as dm_qe. The post shows no publication date or author.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /research/, /benchmark/, /news/, seven Grand Challenges URLs, one model page (/models/hbm_thermal_resistance/, last modified July 13, 2026), terms and privacy. The Quantum ESPRESSO post is not listed, there are no pages for industry partners, services, use cases or careers, and the research index lists two undated posts.Source
  • 06

    Real visitors fail Google's Core Web Vitals on the homepage because of interaction delay: field INP is 1,412 ms, against a 200 ms target, while loading (LCP 1,665 ms) and layout stability (CLS 0) pass. The lab test shows why: 8,476 ms of blocking time and 11.9 s of main-thread work on a page with only 115 elements and 134 KB of JavaScript, which points to one heavy script, most likely a continuous animation, rather than the content.

    Medium impactLow effort First result in 1 month Co founder & CTO

    First action

    Record a performance profile of the homepage on a mid-range phone setting.

    How to measure it

    Homepage INP at the 75th percentile

    Now
    –
    Target
    under 200 ms
    By
    45 days

    Steps

    1. 1

      Profile

      Find the long tasks and their source.

      Co-founder & CTO 2 hours

    2. 2

      Gate the animation

      Run only when visible; respect reduced motion.

      Co-founder & CTO 4 hours

    3. 3

      Lighten mobile

      Static or lower-cost version on small screens.

      Co-founder & CTO 4 hours

    4. 4

      Trim legacy code

      Drop the 22 KB of legacy JavaScript polyfills flagged by PageSpeed.

      Co-founder & CTO 1 hour

    5. 5

      Re-test

      PageSpeed mobile and desktop.

      Co-founder & CTO 30 minutes

    6. 6

      Watch the field data

      INP over the next 28 days.

      Co-founder & CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder & CTO

    Smallest test

    Turn the animation off for one day and compare the lab blocking time.

    Why it works

    Profile the homepage in Chrome DevTools to find the long task, then pause or simplify the animation (run it only when visible, cap its frame rate, move work off the main thread, or show a static image on mobile and for reduced-motion users). Re-test in PageSpeed and watch the field INP over the next 28 days.

    Engineers and investors arriving from press click through to the challenges and research; a page that freezes for over a second on tap costs those clicks and looks careless for a company selling simulation speed.

    Risks and how to handle them

    • Losing the visual identity

      Keep the design; change only when and how often it runs.

    Evidence

    • PageSpeedHomepage, November 2026: mobile score 67, desktop 70; real visitors fail Core Web Vitals on INP (field INP 1,412 ms, while LCP of 1,665 ms and CLS of 0 pass); lab Total Blocking Time is 8,476 ms with 11.9 s of main-thread work, on a page with only 115 DOM elements and 134 KB of JavaScript.Source
    • Your websiteNews page: headlines only, linking TechCrunch (August 10, 2026), a CNBC video, The Economic Times, Business Standard and an MTSLive stream; no company-written announcements.Source

This quarter 11

The core plays for the next three months.

  • 07

    The site has an interactive HBM thermal model, based on a 2025 IEDM paper, that shows how DRAM dielectric thermal conductivity changes stack thermal resistance (113.9 mm²·K/W at baseline) and the tallest practical stack. It is exactly the value argument for the company's materials, aimed at packaging and thermal engineers, but it is not linked from the homepage, has no contact or signup prompt, and has not been updated since July.

    High impactLow effort First result in 1 month Co founder & CEO

    First action

    Add a homepage link and a one-line call to action under the model's result.

    How to measure it

    Model-page enquiries

    Now
    –
    Target
    5 a month
    By
    90 days

    Steps

    1. 1

      Link it

      From the homepage, industry page and benchmark.

      Co-founder & CTO 1 hour

    2. 2

      Add the call to action

      Target conductivity and application in a short form.

      Co-founder & CTO 3 hours

    3. 3

      Brand the PDF

      Logo, source paper, link back and contact.

      Co-founder & CTO 2 hours

    4. 4

      Write the walkthrough

      Three scenarios with charts.

      Co-founder & CEO 2 days

    5. 5

      Share

      LinkedIn, packaging and thermal groups, research partners.

      Co-founder & CEO 1 week

    6. 6

      Track

      Visits, PDF saves and form starts.

      Co-founder & CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder & CEO

    Smallest test

    Post the model with one scenario on LinkedIn and count visits and form starts.

    Why it works

    Link the model from the homepage and the industry page, add one line under the result ("Need a dielectric with this conductivity? Tell us your target") with a short form, brand the "Save as PDF" output with the company and a link, and write a short post walking through three scenarios (today's 12-high stack, a 2x conductive BEOL dielectric, the ceiling). Share it with packaging engineers on LinkedIn.

    Engineers trust a model they can change themselves; the saved PDF travels inside the buyer's company to the people who decide, carrying the company's name.

    Risks and how to handle them

    • Model numbers questioned

      Cite the paper, state assumptions and invite corrections.

    Builds on

    Evidence

    • Your websiteHBM thermal model page: an interactive calculator, based on a 2025 IEEE IEDM paper, showing how DRAM dielectric thermal conductivity changes an HBM stack's thermal resistance (baseline 113.9 mm²·K/W, 12-high) and the tallest practical stack; users can edit layers and "Save as PDF". It has no signup or contact prompt and is not linked from the homepage.Source
    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /research/, /benchmark/, /news/, seven Grand Challenges URLs, one model page (/models/hbm_thermal_resistance/, last modified July 13, 2026), terms and privacy. The Quantum ESPRESSO post is not listed, there are no pages for industry partners, services, use cases or careers, and the research index lists two undated posts.Source
  • 08

    Material Discovery Bench is built on the open Inspect framework, has expert-designed verifiers and a seven-entry leaderboard, but the page describes no way to submit a run and no code is published. Today only the company can add results, so the benchmark cannot become the field's shared scoreboard.

    High impactMedium effort First result in 1 quarter Co founder & CTO

    First action

    Write the submission format in one page.

    How to measure it

    External submissions

    Now
    –
    Target
    10
    By
    2 quarters

    Steps

    1. 1

      Package the eval

      Task, tools and verifiers as an Inspect evaluation.

      Co-founder & CTO 2 weeks

    2. 2

      Define submissions

      Run logs, candidates and environment.

      Co-founder & CTO 2 days

    3. 3

      Re-verify

      Run company verifiers on every submitted candidate.

      Co-founder & CTO 1 week

    4. 4

      Update the leaderboard

      Dated entries, outside and internal.

      Co-founder & CTO 2 days

    5. 5

      Invite groups

      The expert reviewers' groups and materials-informatics labs.

      Co-founder & CEO 2 weeks

    6. 6

      Report

      A quarterly benchmark update post.

      Co-founder & CEO quarterly

    First result
    1 quarter
    Cost
    Compute for re-verification
    Owner
    Co founder & CTO

    Smallest test

    Accept one outside group's run by email and add it to the leaderboard.

    Why it works

    Publish the task definition, tools and verifiers as an Inspect evaluation, document a submission format (run logs plus candidate files), re-verify submitted candidates with the company's verifiers, and show outside entries on the leaderboard with dates. Offer a small compute credit for academic groups.

    Benchmarks spread when others can post to them; every lab that submits learns the company's verifiers, which is the product the founders want to sell.

    Risks and how to handle them

    • Gaming the verifiers

      Keep a hidden check set and publish the reward-hacking rules.

    • Giving away the harness

      Open the evaluation, not the discovery agent.

    Builds on

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • GitHubThe discovered-materials GitHub organization has one public repository, dm_qe ("Discovered Materials' Quantum Espresso mirror", 13 stars, GPL-2.0); there is no repository for the benchmark or the released materials.Source
  • 09

    The rules let entrants keep their IP but grant a perpetual, royalty-free, sublicensable license whether or not they win; submissions are not confidential; entrants who want patents should file before submitting; and the sponsor may solve a challenge itself and close it. The rules also exclude anyone whose participation would break their institution's rules. The strongest solvers are university groups, and their tech-transfer offices will read these terms before a professor can enter.

    Medium impactLow effort First result in 1 month Co founder & CEO

    First action

    List the five questions a tech-transfer office would ask.

    How to measure it

    Institutional entries

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      List the questions

      License, ownership, confidentiality, patents, closure.

      Co-founder & CEO 1 day

    2. 2

      Draft answers

      Plain English, linked to rule sections.

      Co-founder & CEO 2 days

    3. 3

      Legal review

      Make sure the guide matches the rules exactly.

      Outside counsel 1 week

    4. 4

      Write the forwarding note

      For researchers to send their office.

      Co-founder & CEO 2 hours

    5. 5

      Publish

      Linked from every challenge page and the FAQ.

      Co-founder & CTO 1 hour

    6. 6

      Collect questions

      Add new answers as offices ask.

      Co-founder & CEO monthly

    First result
    1 month
    Cost
    A few hours of legal review
    Owner
    Co founder & CEO

    Smallest test

    Send the draft guide to two academic advisers and ask what their office would object to.

    Why it works

    Publish a one-page guide: what entrants keep, what the license covers and why, how to file first, how team and institutional entries work, and what happens if the company solves a challenge itself. Add a short template note a researcher can forward to their tech-transfer office, and an email for questions from those offices.

    Removing the approval delay at the institution is often what decides whether a strong group enters at all; a clear guide also shows good faith on terms some researchers will find broad.

    Risks and how to handle them

    • The guide is read as changing the rules

      State that the Official Rules govern and link each answer to a section.

    Builds on

    Evidence

    • Your websiteOfficial Rules: maximum prizes of $400,000 (cuprate superconductivity), $200,000 (exchange-correlation functional), $200,000 (Si/oxide interfaces), $100,000 (driven systems) and $100,000 (interfacial heat transport); the Competition "has no deadline"; the first eligible entrant meeting all success criteria generally wins; entrants keep their IP but grant a perpetual, royalty-free, sublicensable license; submissions are not confidential, so entrants who want patents should file first; submissions that cannot be verified within 90 days of a request are disqualified.Source
    • PressTechCrunch on the prize (November 2026): a $1 million pool for five open problems, shaped with an advisory group of academics including a Stanford professor; the challenges are "goalposts" showing "whether the models are climbing in this domain"; the company partners with chip companies on research projects; the article gives no per-problem breakdown.Source
  • 10

    The benchmark acknowledges reviewers from IMEC, IBM, Micron, Stanford, Cambridge and other universities, and the prize was shaped with an academic advisory group. These names are the strongest credibility signal the company has for a risk-averse industry, but they sit at the bottom of one benchmark page. CuspAI puts its industry and lab partners on its homepage.

    High impactMedium effort First result in 1 quarter Co founder & CEO

    First action

    Ask each reviewer if they are happy to be listed as an adviser.

    How to measure it

    Industry introductions via advisers

    Now
    –
    Target
    6
    By
    2 quarters

    Steps

    1. 1

      Ask consent

      Role, title and photo for each reviewer.

      Co-founder & CEO 1 week

    2. 2

      Build the page

      Advisers with areas and the work they reviewed.

      Co-founder & CTO 1 day

    3. 3

      Co-author posts

      Two short pieces with advisers.

      Co-founder & CEO 1 month

    4. 4

      Ask for introductions

      One relevant industry contact each.

      Co-founder & CEO 1 month

    5. 5

      Feature on the homepage

      A logo strip of institutions, with consent.

      Co-founder & CTO 2 hours

    6. 6

      Meet quarterly

      A short call on which materials to pursue.

      Co-founder & CEO quarterly

    First result
    1 quarter
    Cost
    Possible small advisory equity grants
    Owner
    Co founder & CEO

    Smallest test

    List three consenting reviewers on the benchmark page with their areas.

    Why it works

    With their consent, create an advisers page listing each reviewer's area and role, invite two or three to co-author short posts on what a fab-ready material must prove, and ask each to introduce one industry contact who works on the problem their review covered.

    In semiconductors, trust moves through people; named experts from fabs and memory makers lower the perceived risk of talking to a seed-stage startup.

    Risks and how to handle them

    • Implying employer endorsement

      Say advisers act in a personal capacity where that applies.

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • PressTechCrunch on the prize (November 2026): a $1 million pool for five open problems, shaped with an advisory group of academics including a Stanford professor; the challenges are "goalposts" showing "whether the models are climbing in this domain"; the company partners with chip companies on research projects; the article gives no per-problem breakdown.Source
    • Competitor researchRivals' own homepages: CuspAI runs an "AI Materials Foundry" with industry partners including Applied Materials, Samsung, Lam Research and Tokyo Electron, and invites industry, synthesis and data partners to sign up; Citrine sells an "AI platform for materials and chemicals development" with published case studies (Dorfner: an 84% reduction in formulation time); Periodic Labs builds models and autonomous labs, targets higher-temperature superconductors and helps a semiconductor manufacturer with chip heat dissipation.Source
  • 11

    The benchmark already defines a buyer-ready spec: a target thermal conductivity, dielectric constant and stiffness, BEOL compatibility and a synthesis recipe judged by experts. The founders want to sell the harness to semiconductor and chemical companies, and an investor says the real constraint is "filtering them correctly and synthesizing them". A fixed-scope sprint turns that format into a first paid contract a cautious buyer can approve.

    High impactMedium effort First result in 1 quarter Co founder & CEO

    First action

    Write a one-page sprint description with deliverables and an indicative price.

    How to measure it

    Signed sprints

    Now
    –
    Target
    2
    By
    2 quarters

    Steps

    1. 1

      Define deliverables

      Shortlist, verified properties, routes, one tested sample.

      Co-founder & CEO 2 days

    2. 2

      Price it

      Compute, lab time and margin.

      Co-founder & CEO 2 days

    3. 3

      Agree IP terms

      Ownership by default and the licensing option.

      Outside counsel 2 weeks

    4. 4

      Pick targets

      Existing research partners and inbound enquiries.

      Co-founder & CEO 1 week

    5. 5

      Pitch

      Ten conversations with R&D leads.

      Co-founder & CEO 1 month

    6. 6

      Deliver one

      Run the first sprint and write it up with consent.

      Co-founder & CTO 8 weeks

    First result
    1 quarter
    Cost
    Lab and compute time per sprint
    Owner
    Co founder & CEO

    Smallest test

    Offer the sprint to one current research partner at cost.

    Why it works

    Offer an 8-week sprint: the customer gives target properties and process limits; the company runs its agents and verifiers, delivers a ranked shortlist with verified properties and synthesis routes, and makes and tests the top candidate in its lab. Fixed price, IP terms agreed up front, and an option to extend into licensing.

    A small, defined project is easier for an R&D manager to fund than an open partnership, and the shortlist plus one physical sample is proof no rival's slide can match.

    Risks and how to handle them

    • Lab capacity

      Limit to one sprint at a time until the organic chemist starts.

    Builds on

    Evidence

    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • PressTechCrunch (August 10, 2026): the plan is to patent uses of new materials and license them to chipmakers, with patentable materials hoped for within a year; agents produce thousands of candidate guesses a day and the founders' lab validates candidates; an investor says the real constraint is "filtering them correctly and synthesizing them"; rivals named include MatNex, SandboxAQ and CuspAI.Source
  • 12

    Of the 500+ materials in the benchmark, only one has a plausible synthesis route, and the team says it is attempting to make it. HN's most-praised point was honesty about feasibility, its sharpest doubt was that "generating candidates got cheap, checking them didn't", and an investor calls synthesis the real constraint. The lab attempt is the most persuasive story the company has, and nobody can see it.

    High impactMedium effort First result in 1 month Founding Organic Chemist

    First action

    Write the first entry: the material, the predictions and the planned route.

    How to measure it

    Inbound industry contacts citing the log

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Decide what to share

      Keep patentable details back; file first if needed.

      Co-founder & CEO 1 week

    2. 2

      Write the first entry

      Material, predictions, route.

      Founding Organic Chemist 1 day

    3. 3

      Log each attempt

      Photos, conditions, results.

      Founding Organic Chemist every 2 weeks

    4. 4

      Measure

      Thermal conductivity and dielectric constant versus prediction.

      Co-founder & CEO as available

    5. 5

      Post

      A short update on X and LinkedIn per entry.

      Co-founder & CEO every 2 weeks

    6. 6

      Wrap up

      Predicted versus measured, lessons for the verifiers.

      Co-founder & CTO 1 week

    First result
    1 month
    Cost
    Lab time already planned
    Owner
    Founding Organic Chemist

    Smallest test

    Post one entry with the predicted properties and a photo of the first attempt.

    Why it works

    Publish a short, dated log for that material: the predicted properties, the synthesis plan, each attempt, characterisation results and what failed, ending with measured versus predicted values. Post each step on X and LinkedIn and send the result to the mailing list, whether it succeeds or not.

    A live record of lab work answers the main objection to AI materials discovery and gives the company a reason to post every few weeks; a failure, explained well, still builds trust.

    Risks and how to handle them

    • Disclosing patentable know-how

      File first and share outcomes before recipes.

    • Public failure

      Frame the log as a test of the verifiers, not a promise.

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
    • PressTechCrunch (August 10, 2026): the plan is to patent uses of new materials and license them to chipmakers, with patentable materials hoped for within a year; agents produce thousands of candidate guesses a day and the founders' lab validates candidates; an investor says the real constraint is "filtering them correctly and synthesizing them"; rivals named include MatNex, SandboxAQ and CuspAI.Source
    • HiringFounding Organic Chemist (Mountain View, $150K–$250K, 0.50%–1.50% equity, new grads OK, visa sponsorship): synthesize materials across classes in the company's organic lab and test hypotheses with experiments; needs experience with semiconductor organic materials ("resists, dielectrics, precursors, inhibitors etc."). It is the only open role, and the YC profile lists a team of 4.Source
  • 13

    The benchmark found agents gaming the checks, for example resubmitting the same material as larger supercells to dodge the novelty check and submitting unmeasured thermal conductivity values. The founders want to sell the harness, and a buyer's first question is whether its results can be trusted. Today the finding is a few lines inside the benchmark page.

    Medium impactLow effort First result in 1 month Co founder & CTO

    First action

    List each hack observed and the check that caught it.

    How to measure it

    Click-through to the industry page

    Now
    –
    Target
    5% of readers
    By
    60 days

    Steps

    1. 1

      Catalogue the hacks

      With examples from run logs.

      Co-founder & CTO 2 days

    2. 2

      Describe the checks

      What each verifier enforces.

      Co-founder & CTO 2 days

    3. 3

      Write the post

      Plain language, with one diagram.

      Co-founder & CTO 3 days

    4. 4

      Link to buyers

      A closing line and link to the industry page.

      Co-founder & CEO 1 hour

    5. 5

      Share

      HN, X, LinkedIn and the mailing list.

      Co-founder & CEO 1 day

    6. 6

      Measure

      Readers and click-through.

      Co-founder & CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder & CTO

    Smallest test

    Post a short thread with two examples and the check that caught each.

    Why it works

    Publish a standalone post: the hacks seen, how each was detected, the checks added (supercell reduction, novelty against known databases, required verifier provenance for every property), and what remains hard. End with a link to the industry page: "These checks run on every candidate we deliver."

    It turns a research curiosity into a reason to buy: the hard part of agent-driven discovery is checking, and this shows the company does it.

    Risks and how to handle them

    • Reads as attacking model vendors

      Keep it about the checks, not the models.

    Builds on

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
  • 14

    The HBM thermal model is the only model page, and it shows that improving the DRAM dielectric alone hits a ceiling because other layers still add resistance. The homepage frames the whole problem as heat ("~140 W/cm²" today). Each other bottleneck the company targets, such as thermal interface materials and interface conductance, could have its own calculator.

    Medium impactMedium effort First result in 1 quarter Co founder & CTO

    First action

    Pick the second model from the questions industry contacts ask most.

    How to measure it

    Monthly model sessions

    Now
    –
    Target
    1,000
    By
    2 quarters

    Steps

    1. 1

      Pick models

      Two that match target materials.

      Co-founder & CEO 1 day

    2. 2

      Source the physics

      Published papers and assumptions.

      Co-founder & CEO 1 week

    3. 3

      Build

      Reuse the HBM page components.

      Co-founder & CTO 2 weeks

    4. 4

      Add the index

      /models with all calculators.

      Co-founder & CTO 1 day

    5. 5

      Add calls to action

      Same form as the HBM model.

      Co-founder & CTO 2 hours

    6. 6

      Promote

      One post per model.

      Co-founder & CEO per launch

    First result
    1 quarter
    Cost
    Founder time
    Owner
    Co founder & CTO

    Smallest test

    Build the TIM stack model as a single page and share it.

    Why it works

    Add two or three models: a thermal interface material stack (TIM conductivity versus junction temperature), an interface conductance model tied to the heat transport challenge, and a BEOL dielectric trade-off of thermal conductivity against dielectric constant. Give each the same layout, citation, PDF export and call to action, under a /models index.

    Each model is a reason for a different engineer to visit, save and share; together they map the company's materials to the device problems buyers already budget for.

    Risks and how to handle them

    • Oversimplified physics

      State limits clearly and ask an adviser to review.

    Builds on

    Evidence

    • Your websiteHBM thermal model page: an interactive calculator, based on a 2025 IEEE IEDM paper, showing how DRAM dielectric thermal conductivity changes an HBM stack's thermal resistance (baseline 113.9 mm²·K/W, 12-high) and the tallest practical stack; users can edit layers and "Save as PDF". It has no signup or contact prompt and is not linked from the homepage.Source
    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
  • 15

    Each challenge is a precise research brief: the heat transport problem names the published Al/Si, Al/GaN, Cu/TiC/diamond and Au/PMMA/Si results to reproduce and the hidden interfaces to beat. The people most likely to solve them are PhD students and postdocs, who usually hear about such problems from their supervisor or a colleague, and today entry is an email address on a web page.

    Medium impactLow effort First result in 1 month Co founder & CEO

    First action

    Make the heat transport brief and send it to three group leaders in the field.

    How to measure it

    Signups with a brief tracking code

    Now
    –
    Target
    100
    By
    90 days

    Steps

    1. 1

      Write the briefs

      Two pages per challenge from the page text.

      Co-founder & CEO 1 week

    2. 2

      Add tracking

      A short link per brief.

      Co-founder & CTO 1 hour

    3. 3

      List groups

      Authors of the cited papers and active groups per topic.

      Co-founder & CEO 3 days

    4. 4

      Send

      A personal note asking to forward.

      Co-founder & CEO 1 week

    5. 5

      Add sharing

      A share link on each challenge page.

      Co-founder & CTO 1 hour

    6. 6

      Follow up

      Answer questions and log interest.

      Co-founder & CEO ongoing

    First result
    1 month
    Cost
    $0
    Owner
    Co founder & CEO

    Smallest test

    Send one brief to five groups and count signups.

    Why it works

    Turn each challenge into a two-page PDF and a short slide: the problem, the targets, the maximum prize, what to submit and the rules link. Send them to the academic advisers and to group leaders who publish on each topic, with a request to forward, and add a "share with a colleague" link on each page.

    Researchers trust a forwarded brief from a peer more than an ad; a ready-made handout makes it easy for a professor to pass on at a group meeting.

    Risks and how to handle them

    • Seen as spam

      Write personally and only to groups working on the topic.

    Builds on

    Evidence

    • Your websiteChallenge pages: titles such as "Breaking the Speed–Accuracy Barrier in Quantum Chemistry" and "Tuning Materials with Light" differ from the names used in the rules ("exchange-correlation functional", "driven systems"). The heat transport page asks entrants to reproduce published Al/Si, Al/GaN, Cu/TiC/diamond and Au/PMMA/Si results and to beat AMM/DMM on hidden interfaces, with a method and working code. Neither it nor the cuprate page states a prize amount or a deadline; both show "Open" and "Updated Nov 2026".Source
    • Your websiteGrand Challenges page (checked November 9, 2026): "We commit $1 million in prize money to solutions to these challenges"; five challenges, each listed as "Open" next to "Nov 2026"; entry is by email, and "Each problem page says what a submission must include"; no per-challenge prize amounts, no web submission form, no FAQ and no leaderboard.Source
    • PressTechCrunch on the prize (November 2026): a $1 million pool for five open problems, shaped with an advisory group of academics including a Stanford professor; the challenges are "goalposts" showing "whether the models are climbing in this domain"; the company partners with chip companies on research projects; the article gives no per-problem breakdown.Source
  • 16

    Entry is by emailing an address shown on the page, with method, code and predictions attached. The rules award the prize to the first eligible entrant who meets all criteria, may split it if qualifying entries arrive within 7 days, and allow 90 days for verification. With email alone, entrants get no timestamped receipt and the public cannot see whether anyone is close.

    Medium impactMedium effort First result in 1 month Co founder & CTO

    First action

    Define the fixed prediction format for the heat transport challenge.

    How to measure it

    Complete, reproducible submissions

    Now
    –
    Target
    5 a month
    By
    2 quarters

    Steps

    1. 1

      Define formats

      Predictions file per challenge.

      Co-founder & CTO 1 week

    2. 2

      Build the form

      Fields, file upload or repo link, receipt.

      Co-founder & CTO 1 week

    3. 3

      Auto-check

      Score against the published targets as a first screen.

      Co-founder & CTO 1 week

    4. 4

      Status board

      Counts by stage per challenge.

      Co-founder & CTO 3 days

    5. 5

      Update rules wording

      Point the rules to the form as the timestamp of record.

      Outside counsel 1 week

    6. 6

      Announce

      To the mailing list.

      Co-founder & CEO 1 day

    First result
    1 month
    Cost
    Founder time
    Owner
    Co founder & CTO

    Smallest test

    Add an auto-reply with a timestamp to the submission email.

    Why it works

    Add a form per challenge that collects the method, a code repository link, predictions in a fixed format and the team details, and returns a timestamped receipt. Show a status board: submissions received, under review, verification requested, and closed, without naming entrants unless they opt in.

    A timestamped receipt matters when "first" wins; a visible count of serious attempts creates momentum and gives the company regular news to share.

    Risks and how to handle them

    • Review workload

      Use the automated first screen before expert review.

    Builds on

    Evidence

    • Your websiteGrand Challenges page (checked November 9, 2026): "We commit $1 million in prize money to solutions to these challenges"; five challenges, each listed as "Open" next to "Nov 2026"; entry is by email, and "Each problem page says what a submission must include"; no per-challenge prize amounts, no web submission form, no FAQ and no leaderboard.Source
    • Your websiteOfficial Rules: maximum prizes of $400,000 (cuprate superconductivity), $200,000 (exchange-correlation functional), $200,000 (Si/oxide interfaces), $100,000 (driven systems) and $100,000 (interfacial heat transport); the Competition "has no deadline"; the first eligible entrant meeting all success criteria generally wins; entrants keep their IP but grant a perpetual, royalty-free, sublicensable license; submissions are not confidential, so entrants who want patents should file first; submissions that cannot be verified within 90 days of a request are disqualified.Source
  • 17

    The company has a mailing list on the challenge, benchmark and research pages, and steady material to share: the prize, the benchmark, the Quantum ESPRESSO speedup and lab work. But the news page is only a list of press headlines with no company-written announcements, so there is no public record of progress between press cycles.

    Medium impactLow effort First result in 1 month Co founder & CEO

    First action

    Draft the first update around the prize amounts and the dataset release.

    How to measure it

    Update open rate

    Now
    –
    Target
    45%
    By
    90 days

    Steps

    1. 1

      Set the format

      Result, lab note, challenge status, ask.

      Co-founder & CEO 1 hour

    2. 2

      Write the first

      Prize amounts and dataset release.

      Co-founder & CEO 2 hours

    3. 3

      Add news posts

      Company posts on /news alongside press links.

      Co-founder & CTO 1 day

    4. 4

      Send

      Same week each month.

      Co-founder & CEO monthly

    5. 5

      Answer replies

      Log industry replies separately.

      Co-founder & CEO monthly

    6. 6

      Review

      Opens, replies and unsubscribes by audience tag.

      Co-founder & CTO quarterly

    First result
    1 month
    Cost
    Two hours a month
    Owner
    Co founder & CEO

    Smallest test

    Send one update this month and count replies.

    Why it works

    Send one short update each month: one result, one lab note, one challenge status and one ask (a material question, an introduction, a candidate). Publish the same text as a dated post on the news page, so the page shows a steady record rather than only August press.

    Subscribers who hear from a company regularly remember it when they have a materials problem; dated posts also give search engines and journalists a current source.

    Risks and how to handle them

    • Skipped months

      Keep it to four short items.

    Builds on

    Evidence

    • Your websiteNews page: headlines only, linking TechCrunch (August 10, 2026), a CNBC video, The Economic Times, Business Standard and an MTSLive stream; no company-written announcements.Source
    • Your websiteResearch post "Accelerating Quantum ESPRESSO by 36x": an AI-driven optimization loop made DFPT phonon calculations about 36x faster and hybrid-functional DFT about 9x faster on GPUs; the team's benchmark suite fell from 3.3 hours to under 4 minutes; the code is released as dm_qe. The post shows no publication date or author.Source
    • Your websiteGrand Challenges page (checked November 9, 2026): "We commit $1 million in prize money to solutions to these challenges"; five challenges, each listed as "Open" next to "Nov 2026"; entry is by email, and "Each problem page says what a submission must include"; no per-challenge prize amounts, no web submission form, no FAQ and no leaderboard.Source

Big bets 3

Larger plays that can change the trajectory.

  • 18

    The founders lean toward selling the harness and tools to semiconductor and chemical companies before licensing IP. The company already has the pieces a buyer would pay for: an agent harness that produces thousands of candidates a day, verifiers for stability, thermal conductivity, dielectric constant and stiffness, and GPU-accelerated Quantum ESPRESSO that cut a suite from 3.3 hours to under 4 minutes. Citrine shows enterprise buyers pay for a materials platform when it comes with measured results.

    High impactHigh effort First result in 2 quarters Co founder & CTO

    First action

    Write a two-page pilot proposal and send it to the warmest research partner.

    How to measure it

    Design-partner pilots converting to annual contracts

    Now
    –
    Target
    1
    By
    3 quarters

    Steps

    1. 1

      Pick the partner

      The research partner with the clearest target property.

      Co-founder & CEO 2 weeks

    2. 2

      Scope

      Properties, data, success metric, security needs.

      Co-founder & CTO 2 weeks

    3. 3

      Contract

      Pilot fee, IP and data terms.

      Outside counsel 1 month

    4. 4

      Deploy

      Private environment with the verifiers.

      Co-founder & CTO 1 month

    5. 5

      Run

      Weekly reviews with the partner's team.

      Co-founder & CTO 3 months

    6. 6

      Measure

      Verified hits versus their baseline process.

      Co-founder & CTO monthly

    7. 7

      Convert

      Annual contract and a case study with consent.

      Co-founder & CEO 1 month

    First result
    2 quarters
    Cost
    Engineering time and cloud compute, offset by the pilot fee
    Owner
    Co founder & CTO

    Smallest test

    Run the harness on one partner-supplied target for two weeks under a letter of intent.

    Why it works

    Run the harness as a private, hosted deployment for one design partner: their target properties and proprietary data, the company's agents and verifiers, and the accelerated DFT. Charge a pilot fee, measure candidates per week and verified hits against the partner's current process, and turn the result into a case study and a standard offer.

    One working deployment is the proof a risk-averse industry asks for, and it builds recurring software revenue alongside the slower licensing path.

    Risks and how to handle them

    • Security reviews slow the start

      Use the existing Vanta trust centre and offer a deployment in the partner's cloud.

    • Distraction from the lab

      One partner only until it converts.

    Builds on

    Evidence

    • Hacker NewsLaunch HN (162 points, 37 comments): the founders plan to license IP on discovered materials and to sell the harness and tools to semiconductor and chemical companies, leaning toward the harness first. Commenters praised showing "how many of the discovered materials are actually feasible", warned that "Semiconductors are also incredibly risk averse", and noted that "generating candidates got cheap, checking them didn't".Source
    • Your websiteResearch post "Accelerating Quantum ESPRESSO by 36x": an AI-driven optimization loop made DFPT phonon calculations about 36x faster and hybrid-functional DFT about 9x faster on GPUs; the team's benchmark suite fell from 3.3 hours to under 4 minutes; the code is released as dm_qe. The post shows no publication date or author.Source
    • Competitor researchRivals' own homepages: CuspAI runs an "AI Materials Foundry" with industry partners including Applied Materials, Samsung, Lam Research and Tokyo Electron, and invites industry, synthesis and data partners to sign up; Citrine sells an "AI platform for materials and chemicals development" with published case studies (Dorfner: an 84% reduction in formulation time); Periodic Labs builds models and autonomous labs, targets higher-temperature superconductors and helps a semiconductor manufacturer with chip heat dissipation.Source
    • PressTechCrunch (August 10, 2026): the plan is to patent uses of new materials and license them to chipmakers, with patentable materials hoped for within a year; agents produce thousands of candidate guesses a day and the founders' lab validates candidates; an investor says the real constraint is "filtering them correctly and synthesizing them"; rivals named include MatNex, SandboxAQ and CuspAI.Source
  • 19

    The benchmark targets exactly the material chipmakers lack for 3D stacking: a BEOL-compatible dielectric with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10. The plan is to patent and license new materials to chipmakers, with patentable materials hoped for within a year. A co-development programme with one manufacturer turns that hope into a dated, shared goal.

    High impactHigh effort First result in 3 quarters Co founder & CEO

    First action

    Write a one-page target spec and share it with two research partners for comment.

    How to measure it

    Joint programmes signed

    Now
    –
    Target
    1
    By
    3 quarters

    Steps

    1. 1

      Write the spec

      Properties, process limits, test methods.

      Co-founder & CEO 2 weeks

    2. 2

      Shortlist partners

      Companies already in research contact.

      Co-founder & CEO 2 weeks

    3. 3

      Propose

      Phases, gates and IP terms.

      Co-founder & CEO 1 month

    4. 4

      Run discovery

      Agents and verifiers on the spec.

      Co-founder & CTO 2 months

    5. 5

      Make candidates

      Synthesis and characterisation in the lab.

      Founding Organic Chemist 3 months

    6. 6

      Partner testing

      In the partner's process flow.

      Co-founder & CEO 3 months

    7. 7

      Publish the spec

      With the partner's agreement.

      Co-founder & CEO 1 day

    First result
    3 quarters
    Cost
    Lab materials, characterisation and legal costs
    Owner
    Co founder & CEO

    Smallest test

    Get one partner to comment on the target spec.

    Why it works

    Propose a joint programme to one logic, memory or packaging company: agree a target spec and process window, run the company's agents and verifiers, make the top candidates in the Mountain View lab, and test them in the partner's flow. Share IP under a pre-agreed licence. Publish the spec (not the candidates) so the field sees the goal.

    A partner's process data and test flow close the lab-to-fab gap the homepage describes; a named programme is the credibility no benchmark can buy.

    Risks and how to handle them

    • Partner owns too much IP

      Agree field-of-use licences before work starts.

    • Long sales cycle

      Run the sprint offer in parallel.

    Builds on

    Evidence

    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source
    • PressTechCrunch (August 10, 2026): the plan is to patent uses of new materials and license them to chipmakers, with patentable materials hoped for within a year; agents produce thousands of candidate guesses a day and the founders' lab validates candidates; an investor says the real constraint is "filtering them correctly and synthesizing them"; rivals named include MatNex, SandboxAQ and CuspAI.Source
    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
  • 20

    The only open role is a Founding Organic Chemist for the Mountain View organic lab, asking for experience with semiconductor organic materials: "resists, dielectrics, precursors, inhibitors". Everything published so far, the benchmark and the released materials, is about inorganic crystals. The hire signals a second product line that the site does not yet mention.

    High impactHigh effort First result in 1 quarter Founding Organic Chemist

    First action

    Pick the first organic target with the new chemist in their first two weeks.

    How to measure it

    Organic candidates made and measured

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Pick the target

      One class with a clear property window.

      Founding Organic Chemist 2 weeks

    2. 2

      Adapt the agents

      Molecular rather than crystal search and verifiers.

      Co-founder & CTO 1 month

    3. 3

      Make the first batch

      Top candidates in the organic lab.

      Founding Organic Chemist 1 month

    4. 4

      Measure

      Against the property window.

      Founding Organic Chemist 2 weeks

    5. 5

      Add a benchmark task

      An organic track in Material Discovery Bench.

      Co-founder & CTO 1 month

    6. 6

      Pitch suppliers

      Chemical companies supplying fabs.

      Co-founder & CEO 1 month

    First result
    1 quarter
    Cost
    Lab consumables and the new hire
    Owner
    Founding Organic Chemist

    Smallest test

    Run the agents on one organic target and make the top two candidates.

    Why it works

    Once the chemist starts, define one organic target the agents can search and the lab can make quickly (for example a precursor or inhibitor with a stated property window), add an organic task to the benchmark with matching verifiers, and publish the first lab-made result. Pitch it to chemical suppliers, who buy these materials for fabs.

    Organic materials have shorter synthesis cycles than many crystals, so the lab can show made-and-measured results sooner, and it opens chemical suppliers as a second set of buyers.

    Risks and how to handle them

    • Spreading a four-person team thin

      One target only until the first result.

    Builds on

    Evidence

    • HiringFounding Organic Chemist (Mountain View, $150K–$250K, 0.50%–1.50% equity, new grads OK, visa sponsorship): synthesize materials across classes in the company's organic lab and test hypotheses with experiments; needs experience with semiconductor organic materials ("resists, dielectrics, precursors, inhibitors etc."). It is the only open role, and the YC profile lists a team of 4.Source
    • Your websiteMaterial Discovery Bench: agents must propose novel, dynamically stable, BEOL-compatible crystals with thermal conductivity above 20 W/(m·K) and a dielectric constant below 10, plus a synthesis recipe; seven frontier models found between 1.3 and 4.0 materials per run; "over 500 materials" were discovered and released publicly, but only one has a plausible synthesis route and the team is attempting to make it; reward hacking was observed; expert reviewers from IMEC, IBM, Micron, Stanford and Cambridge; built on the Inspect framework. The page links no GitHub, Hugging Face or paper and describes no way to submit a run.Source

Wildcards 2

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

  • 21

    The homepage's argument rests on numbers: GPUs at "~140 W/cm²" and rising each generation, chips at least 10,000x less power efficient than the brain, and the HBM model's limit on stack height. These numbers are scattered across the homepage and one model page with no single sourced reference that journalists, analysts and engineers can cite.

    Medium impactMedium effort First result in 2 months Co founder & CEO

    First action

    Collect sourced heat-flux figures for the last five GPU generations.

    How to measure it

    Sites linking to the heat ledger

    Now
    –
    Target
    20
    By
    2 quarters

    Steps

    1. 1

      Collect data

      Published specs and papers only.

      Co-founder & CEO 1 week

    2. 2

      Design the chart

      Heat flux and stack height over time.

      Co-founder & CTO 3 days

    3. 3

      Write the context

      Where materials change the curve.

      Co-founder & CEO 2 days

    4. 4

      Publish

      With sources and a reuse licence.

      Co-founder & CTO 1 day

    5. 5

      Pitch

      To the journalists who covered the launch.

      Co-founder & CEO 1 week

    6. 6

      Update

      At each major chip launch.

      Co-founder & CEO quarterly

    First result
    2 months
    Cost
    $0
    Owner
    Co founder & CEO

    Smallest test

    Publish the GPU heat-flux chart alone and share it once.

    Why it works

    Publish one page that tracks, per chip and memory generation, power, heat flux, HBM stack height and the thermal resistance budget, each with a source, and shows where better materials would move the line. Update it when new parts launch and offer the chart for reuse with credit.

    A sourced reference page earns links and citations every time the heat problem is in the news, and it frames the company as the authority on the problem it solves.

    Risks and how to handle them

    • Disputed figures

      Cite every number and correct quickly.

    Evidence

    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source
    • Your websiteHBM thermal model page: an interactive calculator, based on a 2025 IEEE IEDM paper, showing how DRAM dielectric thermal conductivity changes an HBM stack's thermal resistance (baseline 113.9 mm²·K/W, 12-high) and the tallest practical stack; users can edit layers and "Save as PDF". It has no signup or contact prompt and is not linked from the homepage.Source
  • 22

    The only open role, a Founding Organic Chemist for the Mountain View lab, is the hire the lab roadmap depends on. The homepage's "Join us" sends candidates to the YC jobs page, the sitemap has no careers page, and the YC listing gives little detail on the lab, its equipment or the problems the chemist would work on, which matters to an experienced synthetic chemist weighing a four-person startup.

    Medium impactLow effort First result in 1 month Co founder & CEO

    First action

    Take photos of the lab and list its main equipment.

    How to measure it

    Qualified chemist applicants

    Now
    –
    Target
    15
    By
    60 days

    Steps

    1. 1

      Gather the lab story

      Photos, equipment, first projects.

      Co-founder & CEO 2 days

    2. 2

      Write the page

      Role, lab, problems, pay and equity.

      Co-founder & CEO 1 day

    3. 3

      Publish

      /careers, linked from "Join us" and the sitemap.

      Co-founder & CTO 2 hours

    4. 4

      Share with departments

      Chemistry and materials departments, and advisers' groups.

      Co-founder & CEO 1 week

    5. 5

      Mention in the update

      In the monthly email.

      Co-founder & CEO 1 hour

    6. 6

      Track

      Applicants by source.

      Co-founder & CEO weekly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder & CEO

    Smallest test

    Add three lab photos and the first project to the YC listing.

    Why it works

    Add /careers with the role, the lab (photos, equipment, location), the first problems the chemist would own, how agents and lab work fit together, the benchmark and the expert reviewers, and the salary and equity range already published. Keep the YC application link, and share the page with materials and chemistry departments and the challenge mailing list.

    Scientists choose labs, not job titles; showing the lab and the problems widens the pool and speeds a hire that unlocks the synthesis bottleneck.

    Risks and how to handle them

    • Revealing lab capabilities competitors could use

      Show the space and general equipment, not processes.

    Evidence

    • HiringFounding Organic Chemist (Mountain View, $150K–$250K, 0.50%–1.50% equity, new grads OK, visa sponsorship): synthesize materials across classes in the company's organic lab and test hypotheses with experiments; needs experience with semiconductor organic materials ("resists, dielectrics, precursors, inhibitors etc."). It is the only open role, and the YC profile lists a team of 4.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /research/, /benchmark/, /news/, seven Grand Challenges URLs, one model page (/models/hbm_thermal_resistance/, last modified July 13, 2026), terms and privacy. The Quantum ESPRESSO post is not listed, there are no pages for industry partners, services, use cases or careers, and the research index lists two undated posts.Source
    • Your websiteHomepage: H1 "Accelerating the lab-to-fab timeline"; AI agents that discover new materials for semiconductor chips; "GPUs today handle heat fluxes of ~140 W/cm²"; a $9M seed round led by Lightspeed with Y Combinator and Peak XV. The only calls to action are "Talk to us" (an email link) and "Join us" (the YC jobs page); there is no mailing-list signup, no offer for chipmakers or chemical companies, and no mention of the $1 million prize, the Quantum ESPRESSO speedup or the HBM thermal model.Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital in a new tab)
More example playbooks and reviews

More growth playbooks

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

  • Rovi Health growth strategyA text-based care concierge and virtual clinic for employer health plans. Members text a nurse navigator who steers them to high-quality…22 growth ideas
  • Mintive growth strategyInfrastructure for putting AI agents to work safely: an MCP gateway with role-based bundles, an LLM gateway, Mint Guard runtime…21 growth ideas
  • Aiwyn growth strategyA platform for accounting firms covering payments and collections, practice management, a client experience portal with engagement…21 growth ideas
  • Muon Space growth strategyAn end-to-end space systems provider (the "Mission Foundry") that designs, builds, launches and operates LEO satellite constellations, and…21 growth ideas
  • Teads growth strategyAn omnichannel advertising platform, merged with Outbrain in 2025, selling branding and performance campaigns across the open web, mobile…20 growth ideas
  • Blink Health growth strategyA healthcare technology company that makes prescriptions affordable: BlinkRx, a pharma-to-patient platform for branded drugs (benefit…20 growth ideas
  • Engine growth strategyA free business travel and spend platform (formerly Hotel Engine; hotelengine.com now redirects to engine.com) for hotels, flights, cars…20 growth ideas
  • Skydio growth strategyThe largest US drone maker: autonomous X10, F10 Lightrunner and R10 drones, docks and Command software for drone as first responder, site…23 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. Discovered Materials did not commission, review or endorse it, and Programmable is not affiliated with Discovered Materials. 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 Discovered Materials runs its business.

Something wrong, or do you represent Discovered Materials? 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

    Make the prize, the benchmark and the research easy to read, find and follow; give industry a page; fix the slow homepage.

  2. Days 31 to 609 ideas

    Turn the HBM model, the benchmark, the reviewers, the lab and the challenges into repeatable channels: open submissions, an IP FAQ, a synthesis log, the verifier write-up, shareable challenge briefs, web submissions and a quarterly update.

  3. Days 61 to 907 ideas

    Sell the first paid sprint, grow the model library, and commit to the big bets: a hosted harness, a co-developed dielectric and an organic-materials track; plus the heat ledger and a careers page.

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 honesty about feasibility: saying how many discovered materials can actually be made. The most cited problem: new materials take years and hundreds of millions of dollars to reach the fab.

  • Open, verifiable work: a public benchmark, released materials and open-source code
  • Heat: rising chip power and poorly conducting dielectrics block 3D stacking
  • In their words: “lab-to-fab”

Market

Competitive landscape

5 competitors compared: CuspAI, Periodic Labs, Citrine Informatics, SandboxAQ and 1 more.

  • Gap to exploit: A public, verifiable record of AI-found materials that were actually made
  • CuspAI: Broad across coatings, membranes, catalysts and energy storage, not focused on chip heat
  • Periodic Labs: Broad science agenda; less public benchmark evidence on semiconductor dielectrics

Hiring signals

Discovered Materials has one open role, and it points at the bottleneck the company and its investors name: making what the agents find.

  • Founding Organic Chemist: Synthesis in a new organic lab; semiconductor resists, dielectrics, precursors and inhibitors. Ideas 12, 20, 22.

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 first (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)
  • 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