ProgrammableMarketing/b12-labs.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 · 8
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 1
  • How we can help
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. b12 labs
b12 labs logo

growth playbook · Nov 10, 2026

b12-labs.com (opens in a new tab)

b12 labs growth playbook

Diagnosis

b12 labs is a three-person Y Combinator (Summer 2025) company with serious science behind it: 12 papers including the ChemCrow agent, a Selenium benchmark recovering expert strategies in 80% of cases, and, by its own account, paid pilots with large pharma and a top-3 pharma repeat customer. The website hides most of that.

Every commercial path ends in an email link, the Foundry asks humans for a molecule by email while its API asks agents for a structured brief, none of the pharma traction appears on the site, and the free academic tier is mentioned only in a blog post. The first-plate numbers vary from page to page (99% yield, >80% conversion, 91%, >95%), and the products go by two sets of names.

This playbook starts with a structured Foundry brief form, then on-site proof, the academic offer, one set of numbers and one set of names. Next come credit pricing, a public benchmark, instrument pages and a simulation-first Cobalt trial. The big bets are self-serve Foundry ordering, a pre-booked Bay Area lab and a partner program for platforms that embed Selenium.

Growth Score

6/10

Biggest Opportunity

Turn real pharma results and the Foundry into a self-serve path: a structured brief, published proof and prices, so a chemist can go from "try Selenium" to a quoted Foundry project without waiting on an email thread.

Biggest Risk

With a team of three and email-only sales, demand from pharma, academia and agent users can outrun follow-up, while better-funded rivals such as Chemify and Iktos sell the same promise of AI planning plus robotic execution.

01

Start here

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

  1. 1

    Give the Foundry a "send us your molecule" form that asks people for the same brief the Foundry API already asks agents for

    Copy the Foundry API's Molecule, Reaction and Custom fields into a simple web form on /foundry/.

    High impact First result in 2 weeks
  2. 2

    Put the pharma track record from the YC profile on b12-labs.com, which today shows no customer proof

    Agree the exact wording of the pharma claims that can go on the homepage.

    High impact First result in 1 month
  3. 3

    Say "free for academic and nonprofit researchers" on the homepage, not only inside a July blog post

    Add the free-academic line under the Selenium button on the homepage.

    High impact First result in 2 weeks

02

Your idea portfolio

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

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

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

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 4Big bets · 8Fill-ins · 6Reconsider · 5
Effort →

Showing 23 of 23 ideas

This week 8

Quick wins with a clear first action.

  • 01

    The Foundry page invites chemists to "Send us your molecule" or "Discuss your experiments", but both buttons open an email, and so does every "Request a Demo" and "Contact Us" on the homepage. Meanwhile the Foundry API docs define exactly what a good brief looks like: Molecule (target, amount, purity), Reaction (scale per well, well limit, measurement goal) or Custom, and each estimate comes back with price, delivery dates and assumptions. An agent gets a structured intake and a status trail. A human gets a blank email and has to guess what to include, and b12 gets briefs that need another round of questions before anyone can quote. Rivals such as Enamine use a "Request a quote" form.

    High impactLow effort First result in 2 weeks Co Founder & CEO

    First action

    Copy the Foundry API's Molecule, Reaction and Custom fields into a simple web form on /foundry/.

    How to measure it

    Quotable Foundry briefs (no follow-up questions needed)

    Now
    –
    Target
    2x the current monthly email briefs
    By
    60 days

    Steps

    1. 1

      Map the fields

      Take the Molecule, Reaction and Custom fields from the Foundry API schema and mark which are required for a first estimate.

      Co-Founder & CTO half a day

    2. 2

      Build the form

      A brief-type picker with conditional fields, structure upload or SMILES, deliverable choice and chemistry flags, on /foundry/.

      Founding Product Engineer 2 days

    3. 3

      Connect the pipeline

      Write each submission as a Foundry request with the same statuses as the API, so web and agent briefs share one queue.

      Co-Founder & CTO 1 day

    4. 4

      Write the confirmation

      Say what happens next, when an estimate is expected, and how confidential structures are handled.

      Co-Founder & CEO 2 hours

    5. 5

      Swap the buttons

      Point "Send us your molecule", "Discuss your experiments" and "Request a Demo" at the form; keep email as a fallback link.

      Founding Product Engineer 2 hours

    6. 6

      Review every brief

      A chemist checks each brief within one working day and notes which fields were missing.

      Research Scientist ongoing

    7. 7

      Measure

      Compare briefs, quotes and acceptances with the last quarter of email enquiries.

      Co-Founder & CEO monthly

    First result
    2 weeks
    Cost
    About three days of engineering; no new tools
    Owner
    Co Founder & CEO

    Smallest test

    Add a one-page form with the Molecule fields only and send half the Foundry button clicks to it for two weeks.

    Why it works

    Put a short web form on /foundry/ built on the API's three brief types. A picker (molecule, reaction, custom) shows only the fields that brief needs: a SMILES or structure upload and the amount and purity for molecules; scale per well, well count and measurement goal for reactions; free text for custom work. Add optional fields for deadline, chemistry flags (air-sensitive, photochemical, high-pressure, which the Zurich lab already handles) and whether the customer wants compounds, data or both. Each submission becomes a Foundry request with the same statuses the API uses, and the submitter gets a confirmation saying what happens next and when to expect an estimate. Keep the email link as a fallback and add a note for confidential structures (NDA first). Use the same form for "Request a Demo", with a field that routes Foundry, Selenium, Palladium and Cobalt requests.

    A structured brief removes the back-and-forth that slows every first quote, and it tells buyers what b12 needs and can do. Because the form writes into the pipeline the API already uses, it needs almost no new back-office work, and every enquiry becomes measurable: briefs submitted, briefs quoted, quotes accepted.

    Ready to use

    ChecklistFoundry brief form fields

    [ ] Brief type: Molecule / Reaction / Custom
    [ ] Molecule: structure (SMILES or file) or code name under NDA, amount, purity
    [ ] Reaction: transformation, scale per well, number of wells, measurement goal
    [ ] Custom: mixing, dilution, sample preparation, input matrix
    [ ] Deliverable: compounds, data (including failed reactions), or both
    [ ] Chemistry flags: air-sensitive, photochemical, high-pressure
    [ ] Target date and budget range
    [ ] Work email and organisation
    [ ] Confirmation: what happens next and when the estimate arrives

    Risks and how to handle them

    • Pharma teams won't paste structures into a web form

      Allow a code name or "structure under NDA" option and start with the scope fields only.

    • More low-quality enquiries

      Make amount, purity or scale required, and add a short budget-range field.

    Evidence

    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
    • Your websiteFoundry API docs: a pilot that "requests quotes" and "does not accept orders, take payment or authorize lab execution"; one intake with three brief types: Molecule (target, requested amount and purity), Reaction (scale per well, well limit, measurement goal) and Custom (mixing, dilution, sample preparation); a published estimate "includes price, currency, delivery dates and assumptions"; request statuses received, in_review, on_hold and estimate_available; credentials are issued privately per customer.Source
    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Competitor researchEnamine sells custom synthesis alongside "400 Thousand compounds in stock" and over 650 chemists, with a "Request a quote" form and a separate custom request page.Source
  • 02

    b12's YC profile and job post say it runs paid pilots with large pharma, supports 5 of the top 10 pharma companies, has a top-3 pharma repeat customer, and serves flavor and fragrance, energy, biotech and university customers. b12-labs.com shows none of it: no logos, no testimonials and no case studies, only the founders' former employers. A pharma buyer has to find the YC page to learn that peers already pay for the product.

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

    First action

    Agree the exact wording of the pharma claims that can go on the homepage.

    How to measure it

    Homepage-to-enquiry conversion

    Now
    –
    Target
    +30%
    By
    90 days

    Steps

    1. 1

      Agree claims

      Which numbers and industries can be published without names.

      Co-Founder & CEO 1 day

    2. 2

      Ask for logos

      Ask academic and biotech users, who are often freer to be named, for logo permission.

      Co-Founder & CEO 2 weeks

    3. 3

      Write two stories

      One Palladium first-plate project and one Foundry delivery, anonymised.

      Research Scientist 1 week

    4. 4

      Customer review

      Each customer approves its anonymised story.

      Co-Founder & CEO 1 week

    5. 5

      Publish

      Homepage proof band plus a /customers page.

      Founding Product Engineer 1 day

    6. 6

      Reuse

      Put the same proof in demo decks and Foundry confirmations.

      Co-Founder & CEO ongoing

    First result
    1 month
    Cost
    Staff time only
    Owner
    Co Founder & CEO

    Smallest test

    Add the one-line pharma claim under the hero this week.

    Why it works

    Add a "Who uses b12" band under the hero with the claims b12 can publish: "Used by 5 of the top 10 pharma companies", "Repeat customer: a top-3 global pharma", and the industries served. Then write two anonymised project stories (one Palladium first-plate result, one Foundry delivery) with the problem, what b12 did, and the result. Add a /customers page with the stories and logos as permissions come through.

    Scientific buyers in large companies look for peer adoption before they bring in procurement or security. Anonymised stories with real numbers are common in pharma, where many customers can't be named.

    Risks and how to handle them

    • Customers object to even anonymous mentions

      Use only aggregate claims ("5 of the top 10") until a customer approves a story.

    Evidence

    • Y Combinator profileYC Summer 2025, San Francisco, team of 3, one-liner "Making hard molecules for Pharma and Biotech". The profile says b12 runs paid pilots with large pharma (one reached full conversion in a single attempt), works with three large pharma companies including a top-3 global pharma that came back as a repeat customer, also serves flavor and fragrance, energy, biotech and university customers, and that an AI drug discovery platform has built b12's synthesis agent into its own pipeline. The Foundry opened in summer 2025 and is producing molecules for clients.Source
    • HiringFounding Product Engineer (Full Stack), San Francisco, $120K-$160K plus 0.50%-2.00% equity, open on November 10, 2026. The role owns the reliability, performance and usability of the customer-facing platform and the links between AI agents, scientific applications, databases and long-running workers, and talks directly with users. The post says b12 supports 5 of the top 10 pharma companies and wants the product to "feel fast, dependable, and coherent".Source
    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.Source
  • 03

    The Selenium launch post says academic and nonprofit researchers can use Selenium free with no application or demo request. The homepage, the support page and the sitemap say nothing about it, and there is no academic page. The batch post mentions free access again, but a professor or PhD student landing on the homepage sees only "Get Started" and "Request a Demo", with no hint that it costs nothing.

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

    First action

    Add the free-academic line under the Selenium button on the homepage.

    How to measure it

    Academic sign-ups

    Now
    –
    Target
    3x the current monthly rate
    By
    90 days

    Steps

    1. 1

      Homepage line

      Free-for-academics note under Try Selenium.

      Founding Product Engineer 1 hour

    2. 2

      Write /academic

      Scope, eligibility, data policy, how to cite.

      Co-Founder & CTO 1 day

    3. 3

      Tag sign-ups

      Flag academic email domains in the app so the tier is measurable.

      Founding Product Engineer 1 day

    4. 4

      Announce

      LinkedIn post and notes to co-authors' groups.

      Co-Founder & CTO 1 day

    5. 5

      Add to sitemap

      Make /academic indexable and link it from the footer.

      Founding Product Engineer 1 hour

    6. 6

      Review

      Sign-ups, searches run and conversions to commercial use.

      Co-Founder & CEO monthly

    First result
    2 weeks
    Cost
    $0 plus compute for free searches
    Owner
    Co Founder & CTO

    Smallest test

    Add the homepage line and compare academic sign-ups for two weeks before and after.

    Why it works

    Add a line under "Try Selenium" ("Free for academic and nonprofit researchers") and build a short /academic page: what's free, how to sign up with a university email, the data policy (structures not used for training), how to cite b12's papers, and a link to commercial access for spin-outs. Share it with the chemistry groups and preprint channels the founders already reach through their publications.

    Academic chemists become the next generation of pharma chemists and often publish routes they planned. A free tier only grows if people know it exists.

    Risks and how to handle them

    • Compute cost of free usage

      Set a monthly search allowance per academic account.

    Evidence

    • Your websiteSelenium launch post (July 22, 2026): on 100 molecules with published expert syntheses, Selenium recovered the expert's strategy in 80% of cases, against "at most 20%" for traditional retrosynthesis systems and 56% for general coding agents. Academic and nonprofit researchers can use Selenium free with no application or demo request; commercial teams are asked to email. Submitted structures are not used for training. Early users are at top pharmas, biotechs, CROs and universities. The post promises a coming end-to-end walkthrough of one hard target.Source
    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.Source
    • Your websiteBatch retrosynthesis post (September 24, 2026): across 100 targets Selenium found a shared intermediate in 85 routes, and after one plain-language instruction routed all 100 through it, working through an assistant connected to the Selenium MCP; batch routes are easier to automate and parallelise and need fewer starting materials.Source
  • 04

    The homepage shows Palladium at "99% yield on first plate" and ">80% conversion rate". The SLAS page says ">80% conversion on the first plate", the SLAS recap says "up to 99% yield on the first plate", and the Palladium post reports 91% from one internal plate and above 95% from partners. Each may be true for a different reaction, but a process chemist reading two pages sees four numbers without context, and in a field built on reproducibility that looks like rounding up.

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

    First action

    List every first-plate figure on the site and the experiment behind each.

    How to measure it

    Claim consistency

    Now
    –
    Target
    100% of pages
    By
    2 weeks

    Steps

    1. 1

      Inventory claims

      Homepage, /slas, SLAS recap, Palladium post, decks.

      Founding Product Engineer 2 hours

    2. 2

      Trace each number

      Reaction, metric, plates, who ran it.

      Research Scientist 1 day

    3. 3

      Write the note

      A short results table with method notes.

      Co-Founder & CTO 1 day

    4. 4

      Choose headline figures

      A typical and a best case, with yield and conversion labelled separately.

      Co-Founder & CEO 1 hour

    5. 5

      Update pages

      Same words and a link to the note everywhere.

      Founding Product Engineer half a day

    6. 6

      Keep it current

      Add each new plate result.

      Research Scientist monthly

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

    Smallest test

    Fix the homepage and /slas figures to match today.

    Why it works

    Write one short "First-plate results" note: each reaction (anonymised), the metric (yield or conversion), the number of plates, and who ran it (internal, partner or Foundry). Replace the homepage and SLAS figures with a typical result and a best case, both linked to the note. Use the same wording on every page and in demo decks.

    Chemists trust numbers they can trace. One sourced figure is more persuasive than several bigger ones without context, and it also protects the company when a pharma reviewer checks the claims.

    Risks and how to handle them

    • Lower headline numbers

      Lead with the reduction in optimisation rounds, which is the real benefit.

    Evidence

    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Your websiteSLAS 2026 recap (February 3, 2026): visitors wanted to start small with one painful workflow, trust was earned by reviewing the protocol before it ran, "No coding required" landed immediately, and several asked for demos; Plate Designer reached "up to 99% yield on the first plate". The /slas page says ">80% conversion on the first plate", still uses the names Automation Copilot, Plate Designer and Synthesis Expert, and its "Request a Virtual Demo" buttons are email links.Source
    • Your websitePalladium post (May 18, 2026): an internal Palladium-designed first plate gave 91% yield with no optimisation rounds; early partners reported yields above 95% on first-plate designs; the aim is to cut the usual 3 to 10 rounds of Bayesian optimisation or DoE to about one; users can review the references behind every well.Source
  • 05

    The homepage and app use Selenium, Palladium and Cobalt. The /slas page and the YC launch post still say Automation Copilot, Plate Designer and Synthesis Expert, and llms.txt uses both. People who met b12 at SLAS, or read the launch, may search for names the homepage no longer uses, and AI assistants may describe six products instead of three.

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

    First action

    Add "(formerly ...)" lines to the three product sections on the homepage.

    How to measure it

    Branded search impressions

    Now
    –
    Target
    +25%
    By
    90 days

    Steps

    1. 1

      Decide the scheme

      Element name plus job subtitle.

      Co-Founder & CEO 1 hour

    2. 2

      Update /slas

      New names and subtitles.

      Founding Product Engineer 1 hour

    3. 3

      Annotate old posts

      A note at the top of the launch post.

      Founding Product Engineer 1 hour

    4. 4

      Fix llms.txt

      One name per product.

      Co-Founder & CTO 1 hour

    5. 5

      Align decks and app

      Same pairs everywhere.

      Co-Founder & CEO 1 day

    6. 6

      Watch search

      Search Console queries by product.

      Founding Product Engineer monthly

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

    Smallest test

    Fix /slas today.

    Why it works

    Pick one scheme: the element name first, with the job as a subtitle ("Palladium, plate designer"). Update /slas, the launch post and llms.txt, and add "formerly Plate Designer" on each product section for a year. Use the same pairs in decks, LinkedIn and the app.

    Consistent names concentrate search, links and word of mouth on one term per product, and the subtitle explains what the element names mean.

    Risks and how to handle them

    • Element names are hard to search on their own

      Always pair them with "b12" and the subtitle.

    Evidence

    • Your websiteSLAS 2026 recap (February 3, 2026): visitors wanted to start small with one painful workflow, trust was earned by reviewing the protocol before it ran, "No coding required" landed immediately, and several asked for demos; Plate Designer reached "up to 99% yield on the first plate". The /slas page says ">80% conversion on the first plate", still uses the names Automation Copilot, Plate Designer and Synthesis Expert, and its "Request a Virtual Demo" buttons are email links.Source
    • Your websiteYC launch post (August 15, 2025): b12 says it works across Chemspeed, Unchained, Opentrons and manual setups; lab robots "sit underutilized"; the post names the tools Automation Copilot, Plate Designer and Synthesis Expert.Source
    • Your websitellms.txt: Selenium is for medicinal chemists and synthesis planners, Palladium for process chemists and HTE teams, Cobalt for automation scientists and lab engineers; Selenium gives retrosynthesis plans "in minutes, not days"; commercial, partnership and product enquiries go to a founders' email address.Source
  • 06

    Selenium can be used through MCP, a REST API with an OpenAPI schema, or a downloadable agent skill, and the batch post shows 100 targets routed through an assistant connected to the Selenium MCP. Yet the homepage footer links only the Foundry API and Cobalt MCP docs. The Selenium docs exist only in the sitemap and llms.txt, so the developers and computational chemists most likely to wire Selenium into a pipeline can't find them from the homepage.

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

    First action

    Add the footer link.

    How to measure it

    Selenium docs visits

    Now
    –
    Target
    3x
    By
    60 days

    Steps

    1. 1

      Footer link

      Alongside the Foundry API and Cobalt MCP links.

      Founding Product Engineer 30 minutes

    2. 2

      Selenium section link

      "Use it from your agent or pipeline".

      Founding Product Engineer 30 minutes

    3. 3

      Build /docs/

      A hub with the three docs pages.

      Founding Product Engineer half a day

    4. 4

      Link from support

      Point API questions to the hub.

      Co-Founder & CTO 30 minutes

    5. 5

      Add a quick start

      First route search in five minutes.

      Co-Founder & CTO 1 day

    6. 6

      Track

      Docs visits to first paid search.

      Founding Product Engineer monthly

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

    Smallest test

    Add the footer link today.

    Why it works

    Add "Selenium API & MCP docs" to the footer and a "Use Selenium from your agent or pipeline" link in the Selenium section. Put a short "Developers" hub at /docs/ listing all three docs pages with a one-line purpose each, and point the support page to it.

    Agent and API users are a self-serve channel that needs no demo. They should be able to find the docs in one click.

    Risks and how to handle them

    • More unsupported API users

      Keep usage caps and the quote-before-search step.

    Evidence

    • Your websiteSelenium Route Search docs: access by MCP, a REST API with an OpenAPI schema, or a downloadable agent skill; work is billed in credits, with a credits balance check and a quote before any paid search; no credit prices are published. The homepage footer links the Foundry API and Cobalt MCP docs but not this page.Source
    • Your websiteBatch retrosynthesis post (September 24, 2026): across 100 targets Selenium found a shared intermediate in 85 routes, and after one plain-language instruction routed all 100 through it, working through an assistant connected to the Selenium MCP; batch routes are easier to automate and parallelise and need fewer starting materials.Source
    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.Source
  • 07

    One of only 18 URLs in b12's sitemap is /hex-background/, an internal "Hex Background Lab" with sliders and an "Export SVG" button. The sitemap asks search engines to index it, while pages that matter, such as an academic page or a customers page, don't exist yet.

    Low impactLow effort First result in 2 weeks Founding Product Engineer

    First action

    Exclude /hex-background/ in the sitemap configuration.

    How to measure it

    Content share of the sitemap

    Now
    –
    Target
    100%
    By
    2 weeks

    Steps

    1. 1

      Exclude

      Filter the page out of sitemap generation.

      Founding Product Engineer 15 minutes

    2. 2

      Noindex

      Add a robots noindex tag to the page.

      Founding Product Engineer 15 minutes

    3. 3

      Request removal

      Search Console removal if it's indexed.

      Founding Product Engineer 15 minutes

    4. 4

      Add new pages

      Academic, customers and docs hub as they ship.

      Founding Product Engineer ongoing

    5. 5

      Check quarterly

      Review the sitemap against the nav.

      Co-Founder & CTO quarterly

    6. 6

      Check llms.txt

      Keep it in step with the sitemap.

      Co-Founder & CTO quarterly

    First result
    2 weeks
    Cost
    $0
    Owner
    Founding Product Engineer

    Smallest test

    Remove it from the sitemap today.

    Why it works

    Remove /hex-background/ from the sitemap and add a noindex tag (or move it off the public site). Use the sitemap for real pages and add new ones (academic, customers, instruments, docs hub) as they launch.

    On an 18-page site, every indexed URL shapes how search engines and AI assistants summarise the company.

    Risks and how to handle them

    • The team still needs the tool

      Keep it at an unlisted path with noindex.

    Evidence

    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.Source
  • 23

    Real Chrome visitors already pass Core Web Vitals on b12-labs.com (p75 LCP 1.0s, INP 53 ms, CLS 0), so this is not urgent. But in Google's phone lab test the largest element, the Foundry video, appears at 5.2s, the homepage downloads 10.1 MB, and the mobile score is 74 against 98 on desktop. Scientists on hotel or conference Wi-Fi, for example after SLAS, see the slow version.

    Low impactLow effort First result in 1 week Founding Product Engineer

    First action

    Export a poster frame from the Foundry video and add it to the video tag.

    How to measure it

    Mobile lab LCP

    Now
    –
    Target
    under 2.5s
    By
    2 weeks

    Steps

    1. 1

      Poster frame

      Compressed WebP or AVIF.

      Founding Product Engineer 1 hour

    2. 2

      Defer the video

      Load after interaction or on idle.

      Founding Product Engineer 2 hours

    3. 3

      Mobile encoding

      A smaller file for narrow screens.

      Founding Product Engineer 2 hours

    4. 4

      Image delivery

      Apply the PageSpeed image fix.

      Founding Product Engineer 2 hours

    5. 5

      Retest

      PageSpeed mobile and desktop.

      Founding Product Engineer 30 minutes

    6. 6

      Watch field data

      Confirm real-visitor metrics stay green.

      Founding Product Engineer monthly

    First result
    1 week
    Cost
    $0
    Owner
    Founding Product Engineer

    Smallest test

    Add the poster attribute today and retest.

    Why it works

    Add a compressed poster image to the hero video, load the video only after the page is interactive (or on tap on mobile), and serve a smaller mobile encoding. Apply the "Improve image delivery" fix PageSpeed suggests (about 0.5s).

    A poster image becomes the largest element and loads in a fraction of the time, while the video still plays for visitors on good connections.

    Risks and how to handle them

    • Less visual impact on mobile

      Use a strong still frame from the lab.

    Evidence

    • PageSpeedGoogle PageSpeed (November 10, 2026): mobile score 74, desktop 98; in the phone lab test the largest element, the Foundry video, appears at 5.2s and the homepage downloads 10.1 MB; real Chrome visitors pass, with p75 LCP 1.0s (44% faster than before), INP 53 ms and CLS 0; the top fix is "Improve image delivery" (about 0.5s).Source

This quarter 11

The core plays for the next three months.

  • 08

    Every demo button on the homepage, the /slas page and the SLAS recap is an email link, so a prospect writes an email and waits. b12's own llms.txt already splits its buyers into three roles: medicinal chemists and synthesis planners (Selenium), process chemists and HTE teams (Palladium), and automation scientists and lab engineers (Cobalt). SLAS visitors said they want to start small with one painful workflow.

    High impactLow effort First result in 2 weeks Co Founder & CEO

    First action

    Set up a booking link with three meeting types.

    How to measure it

    Demos held

    Now
    –
    Target
    +50%
    By
    60 days

    Steps

    1. 1

      Three meeting types

      Route planning, plate design, automation.

      Co-Founder & CEO 2 hours

    2. 2

      Build /demo

      One question, then the calendar.

      Founding Product Engineer 1 day

    3. 3

      Write agendas

      What to bring for each type.

      Co-Founder & CTO 2 hours

    4. 4

      Swap buttons

      Homepage, /slas and blog CTAs.

      Founding Product Engineer 1 hour

    5. 5

      Prepare examples

      Run the prospect's molecule or method before the call.

      Research Scientist per demo

    6. 6

      Track

      Booked, held, pilots started.

      Co-Founder & CEO monthly

    First result
    2 weeks
    Cost
    A scheduling tool, under $20 a month
    Owner
    Co Founder & CEO

    Smallest test

    Replace the homepage "Request a Demo" link with a booking link for one week.

    Why it works

    A /demo page asks one question ("What do you want to fix first?") with three answers that match the roles, then shows a calendar with 30-minute slots. Each answer gets a tailored agenda: bring a target molecule, a reaction to optimise, or an existing Chemspeed method. Confirmations ask the prospect to prepare that example, so the first call works on their own chemistry.

    Booking a slot directly removes a round of emails, and a demo on the prospect's own molecule or method fits how SLAS visitors said they want to start.

    Risks and how to handle them

    • Founders' calendars fill up

      Limit slots per week and offer the free academic path to non-commercial users.

    Builds on

    Evidence

    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Your websitellms.txt: Selenium is for medicinal chemists and synthesis planners, Palladium for process chemists and HTE teams, Cobalt for automation scientists and lab engineers; Selenium gives retrosynthesis plans "in minutes, not days"; commercial, partnership and product enquiries go to a founders' email address.Source
    • Your websiteSLAS 2026 recap (February 3, 2026): visitors wanted to start small with one painful workflow, trust was earned by reviewing the protocol before it ran, "No coding required" landed immediately, and several asked for demos; Plate Designer reached "up to 99% yield on the first plate". The /slas page says ">80% conversion on the first plate", still uses the names Automation Copilot, Plate Designer and Synthesis Expert, and its "Request a Virtual Demo" buttons are email links.Source
  • 09

    Selenium is billed in credits, with a balance check and a quote before any paid search, and the support page says assistant-based use draws on "existing credits". But no credit price appears anywhere on the site, and commercial teams are told to email. A biotech chemist who wants to try ten molecules has no way to buy without a sales conversation.

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

    First action

    Work out the average credits per standard search and set a starter-pack price.

    How to measure it

    Starter packs sold

    Now
    –
    Target
    20 in the first quarter
    By
    1 quarter

    Steps

    1. 1

      Measure usage

      Credits per search by tier.

      Co-Founder & CTO 2 days

    2. 2

      Set prices

      Starter pack and enterprise floor.

      Co-Founder & CEO 1 day

    3. 3

      Add checkout

      Card payment for starter packs in the app.

      Founding Product Engineer 1 week

    4. 4

      Write /pricing

      Three tiers and a cost per search example.

      Co-Founder & CEO 1 day

    5. 5

      Link it

      Homepage nav, support page and Selenium docs.

      Founding Product Engineer 1 hour

    6. 6

      Review

      Pack sales, repeat purchases, upgrades.

      Co-Founder & CEO monthly

    First result
    1 month
    Cost
    Payment processing fees and about a week of engineering
    Owner
    Co Founder & CEO

    Smallest test

    Offer a fixed-price starter pack by email link to the next ten commercial enquiries.

    Why it works

    Publish a /pricing page: free for academics; a commercial starter pack (a fixed number of credits for a set price, card payment) for small teams; and annual enterprise plans with SSO, data terms and Foundry discounts on request. Show what a typical search costs in credits by tier. Keep the quote step in the product so users still see the cost before each search.

    Biotechs and CROs often have small budgets they can spend without procurement. A published starter pack turns free trials into revenue without a founder on every deal.

    Risks and how to handle them

    • Public prices anchor enterprise deals low

      Publish the starter pack only and say enterprise is quoted.

    Builds on

    Evidence

    • Your websiteSelenium Route Search docs: access by MCP, a REST API with an OpenAPI schema, or a downloadable agent skill; work is billed in credits, with a credits balance check and a quote before any paid search; no credit prices are published. The homepage footer links the Foundry API and Cobalt MCP docs but not this page.Source
    • Your websiteSupport page: help for Selenium, Palladium, Cobalt and accounts is by email only; Selenium used inside AI assistants draws on the organisation's "existing credits"; the page lists no response times, pricing or FAQs.Source
    • Your websiteSelenium launch post (July 22, 2026): on 100 molecules with published expert syntheses, Selenium recovered the expert's strategy in 80% of cases, against "at most 20%" for traditional retrosynthesis systems and 56% for general coding agents. Academic and nonprofit researchers can use Selenium free with no application or demo request; commercial teams are asked to email. Submitted structures are not used for training. Early users are at top pharmas, biotechs, CROs and universities. The post promises a coming end-to-end walkthrough of one hard target.Source
  • 10

    Selenium's launch post reports its strongest result yet: on 100 molecules with published expert syntheses, it recovered the expert's strategy in 80% of cases against at most 20% for traditional tools and 56% for general coding agents. The result sits inside one blog post. The team has published benchmark-heavy work before (Tango*, a 2026 Matter paper on strategy-aware planning), and rivals such as Iktos and Chemify lead with capability claims rather than open comparisons.

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

    First action

    Check which of the 100 reference syntheses can be listed publicly.

    How to measure it

    Referring domains to /benchmark

    Now
    –
    Target
    25
    By
    2 quarters

    Steps

    1. 1

      Clear the data

      Confirm the molecules and references can be published.

      Co-Founder & CTO 2 days

    2. 2

      Write the method

      Strategy recovery and feasibility scoring.

      Research Scientist 3 days

    3. 3

      Build the page

      Table by version, dated.

      Founding Product Engineer 2 days

    4. 4

      Open submissions

      Rules for outside systems.

      Co-Founder & CTO 1 day

    5. 5

      Announce

      LinkedIn and co-authors' networks.

      Co-Founder & CEO 1 day

    6. 6

      Re-run per release

      Update the table with each Selenium version.

      Research Scientist per release

    First result
    1 quarter
    Cost
    Staff time and compute
    Owner
    Co Founder & CTO

    Smallest test

    Publish the method and Selenium's numbers as a standing page.

    Why it works

    Publish /benchmark: the molecule list (or the public references), the scoring method, Selenium's results by version, and baselines. Re-run with each release and date it. Invite academic groups to submit their own systems. Link it from the homepage Selenium section in place of the "70% faster" line, or next to it.

    Computational chemists and pharma evaluation teams compare tools on benchmarks. An open, re-run benchmark gets cited and linked, and it makes b12 the reference point.

    Risks and how to handle them

    • Rivals overfit to the set

      Keep a held-out set and report both.

    Builds on

    Evidence

    • Your websiteSelenium launch post (July 22, 2026): on 100 molecules with published expert syntheses, Selenium recovered the expert's strategy in 80% of cases, against "at most 20%" for traditional retrosynthesis systems and 56% for general coding agents. Academic and nonprofit researchers can use Selenium free with no application or demo request; commercial teams are asked to email. Submitted structures are not used for training. Early users are at top pharmas, biotechs, CROs and universities. The post promises a coming end-to-end walkthrough of one hard target.Source
    • Your websitePublications page: 12 selected papers in 8 venues, including a 2026 Matter paper on strategy-aware synthesis planning, the 2024 Nature Machine Intelligence paper introducing the ChemCrow chemistry agent, and Tango* in Digital Discovery (2025), plus several 2025 preprints.Source
    • Competitor researchIktos pairs Spaya AI retrosynthesis with its own robotic chemistry and biology labs and cites over 60 pharma, biotech and academic collaborations since 2016.Source
  • 11

    The homepage says Cobalt writes protocols for "Chemspeed, Opentrons, or other automation platforms", and the launch post adds Unchained and manual setups. The Cobalt MCP docs describe Chemspeed AutoSuite and emulated ArkSuite only. SLAS visitors wanted to escape bespoke coding on the instruments they already own. An automation engineer searches by instrument, and b12 has no page for any of them.

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

    First action

    Write the Chemspeed page from the Cobalt MCP docs.

    How to measure it

    Instrument-page demos

    Now
    –
    Target
    10 a month
    By
    2 quarters

    Steps

    1. 1

      Confirm support

      Status per instrument and software version.

      Co-Founder & CTO 1 day

    2. 2

      Chemspeed page

      Simulation, replay, diagnosis, export.

      Co-Founder & CTO 3 days

    3. 3

      Record a demo

      Natural language to an AutoSuite method.

      Research Scientist 2 days

    4. 4

      Other instruments

      Opentrons and Unchained with status and waitlist.

      Co-Founder & CTO 1 week

    5. 5

      Align the homepage

      Match Cobalt's instrument list to the pages.

      Founding Product Engineer 1 hour

    6. 6

      Track

      Search queries and demos by instrument.

      Founding Product Engineer monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Co Founder & CTO

    Smallest test

    Publish the Chemspeed page alone.

    Why it works

    Write one page per instrument: Chemspeed first (simulate and replay AutoSuite methods, diagnose stopped steps, export for review), then Opentrons and Unchained with an honest status (supported, beta or on request) and a waitlist. Each page shows a before-and-after method, a short video and a "bring your method" demo link.

    Instrument names are what automation buyers type into search. Being clear about support builds the trust SLAS visitors said they need.

    Risks and how to handle them

    • Promising instruments not yet supported

      Mark the status clearly and use waitlists to decide what to build next.

    Evidence

    • Your websiteCobalt MCP docs: check and simulate Chemspeed AutoSuite methods without touching hardware, replay recorded runs against emulated ArkSuite, find the first condition that stopped a step and rank where a run spends its time; setup is documented for four agent clients; sign-in uses a b12 account with automation access; usage is capped per organisation and user; edited apps are exported for review rather than pushed to the instrument.Source
    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
    • Your websiteYC launch post (August 15, 2025): b12 says it works across Chemspeed, Unchained, Opentrons and manual setups; lab robots "sit underutilized"; the post names the tools Automation Copilot, Plate Designer and Synthesis Expert.Source
    • Your websiteSLAS 2026 recap (February 3, 2026): visitors wanted to start small with one painful workflow, trust was earned by reviewing the protocol before it ran, "No coding required" landed immediately, and several asked for demos; Plate Designer reached "up to 99% yield on the first plate". The /slas page says ">80% conversion on the first plate", still uses the names Automation Copilot, Plate Designer and Synthesis Expert, and its "Request a Virtual Demo" buttons are email links.Source
  • 12

    b12 is built by people from Chemspeed, its Foundry runs Chemspeed workflows from solid dispensing to LC-MS in Zurich, and Cobalt simulates and diagnoses Chemspeed AutoSuite methods. The launch post says lab robots "sit underutilized", and SLAS visitors complained about bespoke coding on Chemspeed stations. Every Chemspeed owner is a Cobalt prospect.

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

    First action

    Draft a one-page partner proposal with a joint webinar date.

    How to measure it

    Partner-sourced pilots

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Proposal

      Benefit to their customers and to them.

      Co-Founder & CEO 2 days

    2. 2

      First meeting

      Through the founders' existing contacts.

      Co-Founder & CEO 2 weeks

    3. 3

      Joint webinar

      Live method from plain language.

      Co-Founder & CTO 1 month

    4. 4

      Referral path

      A route for customers with programming bottlenecks.

      Co-Founder & CEO 1 month

    5. 5

      Approve wording

      Agree public wording with the partner.

      Co-Founder & CEO 1 week

    6. 6

      Repeat

      Approach a second instrument vendor.

      Co-Founder & CEO next quarter

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Co Founder & CEO

    Smallest test

    Run one joint webinar.

    Why it works

    Propose a joint offer to Chemspeed: a webinar for its users ("get more from your Chemspeed with plain-language methods"), a listing in its partner or software pages, and a referral path for customers who struggle with method programming. Back it with a free method-replay check from idea 13. Check what Chemspeed is willing to say publicly before any co-branding.

    Instrument vendors want their robots used more, and higher use means renewals for them. A partner that makes the hardware easier to use is welcome, and it gives b12 access to a list of qualified buyers.

    Risks and how to handle them

    • The vendor sees Cobalt as competing with its own software

      Position Cobalt as exporting reviewable methods into the vendor's software, as the MCP already does.

    Builds on

    Evidence

    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
    • Your websiteCobalt MCP docs: check and simulate Chemspeed AutoSuite methods without touching hardware, replay recorded runs against emulated ArkSuite, find the first condition that stopped a step and rank where a run spends its time; setup is documented for four agent clients; sign-in uses a b12 account with automation access; usage is capped per organisation and user; edited apps are exported for review rather than pushed to the instrument.Source
    • Your websiteYC launch post (August 15, 2025): b12 says it works across Chemspeed, Unchained, Opentrons and manual setups; lab robots "sit underutilized"; the post names the tools Automation Copilot, Plate Designer and Synthesis Expert.Source
    • Your websiteHomepage (checked November 10, 2026): H1 "AI-Powered Chemical Synthesis"; Selenium ("2 vs 5 steps to target", "70% faster route design"), Palladium ("99% yield on first plate", ">80% conversion rate") and Cobalt (protocols "for Chemspeed, Opentrons, or other automation platforms"); "Request a Demo" and "Contact Us" are protected email links, with no form, booking or pricing link anywhere on the page; no customer logos or testimonials; "Built by researchers from Roche, Chemspeed, Max Planck Institute, and EPFL"; "Backed by Y Combinator"; SOC 2 Type 2.Source
  • 13

    Cobalt's MCP can already simulate AutoSuite methods without hardware, replay recorded runs, find the first condition that stopped a step and rank where a run spends its time. SLAS visitors said they want to start small and that trust is earned by reviewing before anything runs. Today the only way in is a demo email.

    High impactMedium effort First result in 6 weeks Co Founder & CTO

    First action

    Run the replay and diagnosis tools on three real methods and turn the output into a one-page report template.

    How to measure it

    Method-check to pilot rate

    Now
    –
    Target
    15%
    By
    2 quarters

    Steps

    1. 1

      Report template

      Stalls, time ranking, suggested edits.

      Co-Founder & CTO 3 days

    2. 2

      Upload page

      Secure upload with clear data terms.

      Founding Product Engineer 1 week

    3. 3

      Automate the run

      Simulation and diagnosis through the existing engine.

      Co-Founder & CTO 1 week

    4. 4

      Human review

      A chemist checks each report before it goes out.

      Research Scientist per report

    5. 5

      Promote

      Instrument pages, LinkedIn and partner webinars.

      Co-Founder & CEO ongoing

    6. 6

      Follow up

      Offer a Cobalt rebuild call within two days.

      Co-Founder & CEO per report

    First result
    6 weeks
    Cost
    About two weeks of engineering
    Owner
    Co Founder & CTO

    Smallest test

    Offer the check by email to ten SLAS contacts and do the reports by hand.

    Why it works

    A /method-check page where an automation scientist uploads an existing method (under the same data terms as the app) and gets back a report: where it would stall, where time goes, and two suggested edits exported for review. The report ends with an offer to rebuild the method from plain language with Cobalt.

    It starts with the customer's own problem, carries no hardware risk, and shows Cobalt's value before any sales call. Every report is a qualified lead with a known instrument and pain point.

    Risks and how to handle them

    • Methods contain confidential chemistry

      Offer anonymised reagent names and point to SOC 2 Type 2 and the Trust Center.

    Builds on

    Evidence

    • Your websiteCobalt MCP docs: check and simulate Chemspeed AutoSuite methods without touching hardware, replay recorded runs against emulated ArkSuite, find the first condition that stopped a step and rank where a run spends its time; setup is documented for four agent clients; sign-in uses a b12 account with automation access; usage is capped per organisation and user; edited apps are exported for review rather than pushed to the instrument.Source
    • Your websiteSLAS 2026 recap (February 3, 2026): visitors wanted to start small with one painful workflow, trust was earned by reviewing the protocol before it ran, "No coding required" landed immediately, and several asked for demos; Plate Designer reached "up to 99% yield on the first plate". The /slas page says ">80% conversion on the first plate", still uses the names Automation Copilot, Plate Designer and Synthesis Expert, and its "Request a Virtual Demo" buttons are email links.Source
  • 14

    Support is email only, with no stated response times. It covers three products, assistant-based use that draws on "existing credits", and "unexpected usage" questions. The Founding Product Engineer role is about making the platform "feel fast, dependable, and coherent" and talking directly with users. Pharma pilots renew when the champion can show usage and get quick help.

    Medium impactMedium effort First result in 1 quarter Founding Product Engineer

    First action

    Publish a response-time target on /support.

    How to measure it

    Pilot renewal

    Now
    –
    Target
    70%
    By
    2 quarters

    Steps

    1. 1

      Response targets

      Publish them by tier on /support.

      Co-Founder & CEO 1 hour

    2. 2

      Help button

      Pre-fills product, organisation and job reference.

      Founding Product Engineer 1 week

    3. 3

      Usage data

      Per-organisation counts and credits.

      Founding Product Engineer 1 week

    4. 4

      Monthly summary

      An email champions can forward.

      Founding Product Engineer 3 days

    5. 5

      Ticket review

      Top issues each week.

      Co-Founder & CTO weekly

    6. 6

      Measure

      Time to first reply and renewals.

      Co-Founder & CEO monthly

    First result
    1 quarter
    Cost
    About three weeks of engineering
    Owner
    Founding Product Engineer

    Smallest test

    Send a hand-made usage summary to the three largest accounts.

    Why it works

    Add an in-app help button that attaches the product, organisation and last job reference automatically (the details support already asks for). Publish response targets by tier on /support. Send each organisation a monthly summary: searches, plates and methods, credits used, and best results, which the champion can forward internally.

    Fast, context-rich help cuts friction during pilots, and a monthly summary gives the internal champion the numbers to argue for renewal.

    Risks and how to handle them

    • A three-person team can't meet the targets

      Set realistic targets, such as one working day.

    Evidence

    • Your websiteSupport page: help for Selenium, Palladium, Cobalt and accounts is by email only; Selenium used inside AI assistants draws on the organisation's "existing credits"; the page lists no response times, pricing or FAQs.Source
    • HiringFounding Product Engineer (Full Stack), San Francisco, $120K-$160K plus 0.50%-2.00% equity, open on November 10, 2026. The role owns the reliability, performance and usability of the customer-facing platform and the links between AI agents, scientific applications, databases and long-running workers, and talks directly with users. The post says b12 supports 5 of the top 10 pharma companies and wants the product to "feel fast, dependable, and coherent".Source
  • 15

    Academic and nonprofit researchers use Selenium free, and the team's own papers (ChemCrow in Nature Machine Intelligence, a 2026 Matter paper) show how much this field runs on citation. Nothing in the product or on the site asks free users to cite b12 or shows papers that used Selenium. Every published route that credits Selenium is a referral to the next group that reads the paper.

    Medium impactLow effort First result in 6 months Co Founder & CTO

    First action

    Write the citation text and methods sentence.

    How to measure it

    Citing papers listed

    Now
    –
    Target
    10
    By
    12 months

    Steps

    1. 1

      Write the citation

      Paper reference plus a methods sentence.

      Co-Founder & CTO 1 hour

    2. 2

      Add the button

      On each Selenium result.

      Founding Product Engineer 2 days

    3. 3

      Papers list

      "Papers using b12" on /publications.

      Founding Product Engineer 1 day

    4. 4

      Credit reward

      Extra credits for published use.

      Co-Founder & CEO 1 day

    5. 5

      Monitor

      Alerts for papers and preprints mentioning Selenium.

      Research Scientist monthly

    6. 6

      Amplify

      Share each new paper on LinkedIn.

      Co-Founder & CEO per paper

    First result
    6 months
    Cost
    Free credits
    Owner
    Co Founder & CTO

    Smallest test

    Add the citation line to the Selenium results page.

    Why it works

    Add a "Cite this route" button to each Selenium result with a ready citation and a one-line methods sentence. Add a "Papers using b12" section to /publications and invite authors to submit theirs. Give groups that publish with Selenium extra free credits.

    In academic chemistry, peers adopt the tools that appear in papers they read. A citation prompt turns the free tier into a channel that compounds as more papers come out.

    Risks and how to handle them

    • Slow publishing cycles

      Count preprints too.

    Builds on

    Evidence

    • Your websiteSelenium launch post (July 22, 2026): on 100 molecules with published expert syntheses, Selenium recovered the expert's strategy in 80% of cases, against "at most 20%" for traditional retrosynthesis systems and 56% for general coding agents. Academic and nonprofit researchers can use Selenium free with no application or demo request; commercial teams are asked to email. Submitted structures are not used for training. Early users are at top pharmas, biotechs, CROs and universities. The post promises a coming end-to-end walkthrough of one hard target.Source
    • Your websitePublications page: 12 selected papers in 8 venues, including a 2026 Matter paper on strategy-aware synthesis planning, the 2024 Nature Machine Intelligence paper introducing the ChemCrow chemistry agent, and Tango* in Digital Discovery (2025), plus several 2025 preprints.Source
  • 16

    The Selenium post promises a coming end-to-end walkthrough of one hard target. The blog has five posts in 15 months, the last on September 24. b12 is the only company in this set that plans, optimises, automates and makes the molecule, but no post yet shows all four steps on one molecule.

    Medium impactMedium effort First result in 1 month Research Scientist

    First action

    Choose the first public target and run it through Selenium.

    How to measure it

    Series-sourced enquiries

    Now
    –
    Target
    5 a month
    By
    2 quarters

    Steps

    1. 1

      Pick targets

      Six public hard molecules.

      Research Scientist 1 day

    2. 2

      Run the stack

      Route, plate, method, Foundry where useful.

      Research Scientist 2 weeks each

    3. 3

      Write

      Post with figures and the failures.

      Research Scientist 3 days each

    4. 4

      Short video

      Two-minute walkthrough.

      Co-Founder & CEO 1 day each

    5. 5

      Distribute

      LinkedIn, X and chemistry communities.

      Co-Founder & CEO 1 day each

    6. 6

      Measure

      Tracked links to the Foundry form.

      Founding Product Engineer monthly

    First result
    1 month
    Cost
    Staff time and Foundry runs
    Owner
    Research Scientist

    Smallest test

    Publish one planning-only walkthrough.

    Why it works

    Each month, take one public hard target (a natural product or a published drug) through Selenium, Palladium, Cobalt and, where it makes sense, the Foundry. Show the route, the plate, the method and the outcome, including failures. Publish it as a post, a short video and a LinkedIn thread, with a "bring your hardest molecule" link to the Foundry form.

    Chemists share worked examples, and a real end-to-end story shows the platform working together better than product sections can.

    Risks and how to handle them

    • Lab time pulled from paying work

      Use planning-only episodes in busy months.

    Builds on

    Evidence

    • Your websiteSelenium launch post (July 22, 2026): on 100 molecules with published expert syntheses, Selenium recovered the expert's strategy in 80% of cases, against "at most 20%" for traditional retrosynthesis systems and 56% for general coding agents. Academic and nonprofit researchers can use Selenium free with no application or demo request; commercial teams are asked to email. Submitted structures are not used for training. Early users are at top pharmas, biotechs, CROs and universities. The post promises a coming end-to-end walkthrough of one hard target.Source
    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.Source
    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
  • 17

    The Foundry promises chemistry data "with results from successful and failed reactions", and the YC profile says customers get full experimental data, including failed reactions. Failed reactions are scarce in published data because papers report what worked. A small, clean, open sample from b12's own internal runs would be unusual and useful to modellers, and would show the quality of Foundry data.

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

    First action

    Pick one internal screen whose data b12 fully owns.

    How to measure it

    Data briefs from dataset visitors

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Choose the screen

      Internal, non-confidential, well documented.

      Co-Founder & CTO 1 day

    2. 2

      Clean and annotate

      Conditions, outcomes, analytics.

      Research Scientist 2 weeks

    3. 3

      Data card

      Method, scope and limitations.

      Research Scientist 2 days

    4. 4

      Host and license

      A dataset page with a citation.

      Founding Product Engineer 2 days

    5. 5

      Announce

      Preprint note and community channels.

      Co-Founder & CTO 1 week

    6. 6

      Convert

      "Order a dataset like this" link to the Foundry form.

      Co-Founder & CEO 1 day

    First result
    1 quarter
    Cost
    Staff time; the experiments are already done
    Owner
    Co Founder & CTO

    Smallest test

    Release one plate of results with a data card.

    Why it works

    Release one dataset from internal (non-customer) screens: reactions, conditions, outcomes including zero-yield wells, and LC-MS summaries, with a data card. Host it on the site with a citation, announce it to the ML-for-chemistry community, and link it to "order a dataset like this" on the Foundry form.

    Open datasets get cited and downloaded by exactly the computational teams who buy reaction data, and they show the Foundry's rigour better than a sales page can.

    Risks and how to handle them

    • Leaking customer data

      Use only b12's own internal experiments and review each row.

    Evidence

    • Y Combinator profileYC Summer 2025, San Francisco, team of 3, one-liner "Making hard molecules for Pharma and Biotech". The profile says b12 runs paid pilots with large pharma (one reached full conversion in a single attempt), works with three large pharma companies including a top-3 global pharma that came back as a repeat customer, also serves flavor and fragrance, energy, biotech and university customers, and that an AI drug discovery platform has built b12's synthesis agent into its own pipeline. The Foundry opened in summer 2025 and is producing molecules for clients.Source
    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
  • 18

    Selenium's batch feature found a shared intermediate across 85 of 100 targets, then routed all 100 through it on request. Shared routes are easier to automate, run in parallel and need fewer starting materials, which is exactly what a Foundry run needs. Today batch planning and Foundry synthesis are sold separately, and medicinal chemists order analog series all the time.

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

    First action

    Price a sample 20-compound series from an internal batch run.

    How to measure it

    Analog-series projects sold

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Cost model

      Per-compound cost with shared intermediates.

      Co-Founder & CTO 1 week

    2. 2

      Define scope

      Series size, amounts, purity, data.

      Co-Founder & CEO 2 days

    3. 3

      Add brief type

      "Analog series" in the form and API.

      Founding Product Engineer 1 week

    4. 4

      Write the offer

      A section on /foundry/ with an example.

      Co-Founder & CEO 1 day

    5. 5

      Pilot

      Offer it to two existing pharma customers.

      Co-Founder & CEO 1 month

    6. 6

      Review margins

      Compare actual and quoted cost.

      Co-Founder & CTO per project

    First result
    1 quarter
    Cost
    Staff time; lab costs covered by quotes
    Owner
    Co Founder & CEO

    Smallest test

    Offer a batch-planned quote to one existing customer.

    Why it works

    Package a fixed-scope offer: send 10 to 50 analogs; b12 batch-plans them through shared intermediates, quotes the set with a per-compound price, and delivers compounds plus full data, failed reactions included. Add it as a fourth brief type ("Analog series") in the Foundry form and the API.

    Using one intermediate for many targets lowers b12's cost per compound, and the bundle shows why planning and making together beat buying them separately.

    Risks and how to handle them

    • Some analogs fail on the shared route

      Quote with a success threshold and data-only credit for failures.

    Builds on

    Evidence

    • Your websiteBatch retrosynthesis post (September 24, 2026): across 100 targets Selenium found a shared intermediate in 85 routes, and after one plain-language instruction routed all 100 through it, working through an assistant connected to the Selenium MCP; batch routes are easier to automate and parallelise and need fewer starting materials.Source
    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
    • Your websiteFoundry API docs: a pilot that "requests quotes" and "does not accept orders, take payment or authorize lab execution"; one intake with three brief types: Molecule (target, requested amount and purity), Reaction (scale per well, well limit, measurement goal) and Custom (mixing, dilution, sample preparation); a published estimate "includes price, currency, delivery dates and assumptions"; request statuses received, in_review, on_hold and estimate_available; credentials are issued privately per customer.Source

Big bets 3

Larger plays that can change the trajectory.

  • 19

    The Foundry API is a quoting pilot: it "does not accept orders, take payment or authorize lab execution". It already has the hard parts (structured briefs, scientific review, estimates with price and delivery dates, safe retries), and Selenium shows agents running 100-target batch plans. The Founding Product Engineer role covers authentication, task execution and the links between agents and long-running workers. An agent that can plan a route but not place the order hands the last step back to email.

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

    First action

    Design the approval and spending-limit model with one existing customer.

    How to measure it

    API-placed orders

    Now
    –
    Target
    10
    By
    3 quarters

    Steps

    1. 1

      Customer design session

      Approval rules, POs and limits.

      Co-Founder & CEO 2 weeks

    2. 2

      Accept endpoint

      Approval tied to a human or a spending limit.

      Founding Product Engineer 1 month

    3. 3

      Payments

      Purchase orders and card for small orders.

      Founding Product Engineer 1 month

    4. 4

      Execution statuses

      In execution, shipped, data delivered.

      Co-Founder & CTO 2 weeks

    5. 5

      Legal terms

      Order terms and liability.

      Co-Founder & CEO 1 month

    6. 6

      Pilot

      Three customers, low-value orders.

      Co-Founder & CEO 1 quarter

    7. 7

      Open up

      Wider release with docs and the agent skill.

      Co-Founder & CTO 1 quarter

    First result
    2 quarters
    Cost
    Engineering time plus legal review
    Owner
    Co Founder & CTO

    Smallest test

    Let one customer accept an estimate in the app with a PO number.

    Why it works

    Add an accept step to published estimates: an authorised human approves in the app (or a spending limit allows it automatically), a purchase order or card payment is attached, and the request moves to in_execution, then shipped, with data delivered. Start with existing customers and low-value orders, keeping scientific review mandatory.

    The first Foundry customers to buy through their own planning agents will reorder with little friction. It also sets b12 apart from synthesis services that still quote by email.

    Risks and how to handle them

    • Accidental or unauthorised orders

      Human approval by default; spending limits only when the customer opts in.

    • Regulated or hazardous requests

      Keep scientific and safety review before every acceptance.

    Builds on

    Evidence

    • Your websiteFoundry API docs: a pilot that "requests quotes" and "does not accept orders, take payment or authorize lab execution"; one intake with three brief types: Molecule (target, requested amount and purity), Reaction (scale per well, well limit, measurement goal) and Custom (mixing, dilution, sample preparation); a published estimate "includes price, currency, delivery dates and assumptions"; request statuses received, in_review, on_hold and estimate_available; credentials are issued privately per customer.Source
    • HiringFounding Product Engineer (Full Stack), San Francisco, $120K-$160K plus 0.50%-2.00% equity, open on November 10, 2026. The role owns the reliability, performance and usability of the customer-facing platform and the links between AI agents, scientific applications, databases and long-running workers, and talks directly with users. The post says b12 supports 5 of the top 10 pharma companies and wants the product to "feel fast, dependable, and coherent".Source
    • Your websiteBatch retrosynthesis post (September 24, 2026): across 100 targets Selenium found a shared intermediate in 85 routes, and after one plain-language instruction routed all 100 through it, working through an assistant connected to the Selenium MCP; batch routes are easier to automate and parallelise and need fewer starting materials.Source
  • 20

    The Foundry page says "Our own Bay Area lab is next", and b12 is based in San Francisco among the Bay Area biotechs that buy hard molecules. Today all Foundry work runs through the partner lab in Zurich. Chemify promotes its own Chemifarm facility as a selling point. A new lab is expensive to fill from scratch and cheap to fill if capacity is sold first.

    High impactHigh effort First result in 1 quarter Co Founder & CEO

    First action

    Add a Bay Area lab waitlist to the Foundry page.

    How to measure it

    Pre-sold capacity

    Now
    –
    Target
    50% of the first year's capacity
    By
    before opening

    Steps

    1. 1

      Waitlist

      On /foundry/ and in the brief form.

      Founding Product Engineer 1 day

    2. 2

      Founding offer

      Reserved capacity, priority and a fixed price.

      Co-Founder & CEO 1 week

    3. 3

      Local outreach

      Bay Area biotechs and existing customers.

      Co-Founder & CEO 1 quarter

    4. 4

      Size the lab

      Instruments based on commitments.

      Co-Founder & CTO 1 month

    5. 5

      Contracts

      Simple capacity agreements.

      Co-Founder & CEO 1 month

    6. 6

      Launch

      Opening with founding-customer stories.

      Co-Founder & CEO at opening

    First result
    1 quarter
    Cost
    Staff time before opening; the lab itself is a separate investment
    Owner
    Co Founder & CEO

    Smallest test

    Add the waitlist and count sign-ups for a month.

    Why it works

    Open a Bay Area waitlist on /foundry/ with a founding-customer offer: reserved monthly capacity, priority scheduling and a fixed price for the first year in return for a commitment. Use the commitments to size the lab and its instruments, and announce the opening with the founding customers' stories.

    Pre-commitments cut the lab's financial risk, show demand to investors, and turn the opening into a sales event rather than a cost.

    Risks and how to handle them

    • Opening delays break commitments

      Offer Zurich capacity at the same price until the Bay Area lab opens.

    Builds on

    Evidence

    • Your websiteFoundry page: "We make the molecules AI designs." Customers "Send us molecules or a chemistry workflow"; a partner HTE lab at the University of Zurich "runs Chemspeed workflows from solid dispensing to LC-MS analysis" and handles "air-sensitive, photochemical, and high-pressure chemistry"; "Our own Bay Area lab is next." The "Send us your molecule" and "Discuss your experiments" buttons open email; the page gives no prices, turnaround times or example projects.Source
    • Y Combinator profileYC Summer 2025, San Francisco, team of 3, one-liner "Making hard molecules for Pharma and Biotech". The profile says b12 runs paid pilots with large pharma (one reached full conversion in a single attempt), works with three large pharma companies including a top-3 global pharma that came back as a repeat customer, also serves flavor and fragrance, energy, biotech and university customers, and that an AI drug discovery platform has built b12's synthesis agent into its own pipeline. The Foundry opened in summer 2025 and is producing molecules for clients.Source
    • Competitor researchChemify calls itself "a hyperscaler for chemistry and materials science": its Genesis engine pairs AI design with robotic chemistry in its own Chemifarm facility and claims scale from 1 mg to over 40 g on one platform.Source
  • 21

    The YC profile says an AI drug discovery platform has already built b12's synthesis agent into its own pipeline, and Selenium is available by MCP, REST with an OpenAPI schema, and an agent skill. Generative design platforms need to know whether their molecules can be made. PostEra and Iktos run their own discovery pipelines; many others have no synthesis planning of their own. One embedding customer could be the start of a channel.

    High impactHigh effort First result in 2 quarters Co Founder & CEO

    First action

    Ask the existing embedding customer what would make them use it more, and whether they would co-announce.

    How to measure it

    Partner credit volume

    Now
    –
    Target
    25% of Selenium usage
    By
    12 months

    Steps

    1. 1

      Learn from partner one

      Usage, gaps and what they would pay for.

      Co-Founder & CEO 2 weeks

    2. 2

      Volume pricing

      Tiered credits for embedders.

      Co-Founder & CEO 2 weeks

    3. 3

      Partner docs

      Embed guide and a badge.

      Co-Founder & CTO 1 month

    4. 4

      Design-to-Foundry link

      Send partner designs straight to a Foundry brief.

      Founding Product Engineer 1 month

    5. 5

      Recruit

      Five more design platforms.

      Co-Founder & CEO 2 quarters

    6. 6

      Partners page

      Once partners agree to be named.

      Founding Product Engineer 1 day

    First result
    2 quarters
    Cost
    Staff time; volume discounts
    Owner
    Co Founder & CEO

    Smallest test

    Offer the existing embedder a volume price and a co-announcement.

    Why it works

    Formalise an embed program: volume credit pricing, a "synthesizability by b12" badge partners can show, joint documentation, and a direct route from a partner's design to a Foundry brief. Publish a partners page once the first partner agrees to be named.

    Each platform that embeds Selenium brings its own users and their molecules, and gives a natural route into Foundry orders.

    Risks and how to handle them

    • Partners build their own planning

      Compete on the published benchmark and the Foundry link.

    Builds on

    Evidence

    • Y Combinator profileYC Summer 2025, San Francisco, team of 3, one-liner "Making hard molecules for Pharma and Biotech". The profile says b12 runs paid pilots with large pharma (one reached full conversion in a single attempt), works with three large pharma companies including a top-3 global pharma that came back as a repeat customer, also serves flavor and fragrance, energy, biotech and university customers, and that an AI drug discovery platform has built b12's synthesis agent into its own pipeline. The Foundry opened in summer 2025 and is producing molecules for clients.Source
    • Your websiteSelenium Route Search docs: access by MCP, a REST API with an OpenAPI schema, or a downloadable agent skill; work is billed in credits, with a credits balance check and a quote before any paid search; no credit prices are published. The homepage footer links the Foundry API and Cobalt MCP docs but not this page.Source
    • Competitor researchPostEra describes itself as building an "AI-first biotech" around its Proton platform, with partners including Pfizer, Amgen and Merck KGaA.Source

Wildcards 1

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

  • 22

    The YC profile says b12 already serves flavor and fragrance and energy companies, not only pharma and biotech. Every page on b12-labs.com speaks to pharma R&D ("from years to months"), so a fragrance chemist sees nothing written for them. Flavor and fragrance chemistry is synthesis-heavy, often uses HTE, and is less crowded with AI vendors than drug discovery.

    Medium impactMedium effort First result in 1 quarter Co Founder & CEO

    First action

    Ask the existing flavor and fragrance customer what problem b12 solved for them.

    How to measure it

    Flavor and fragrance briefs

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Customer interview

      The problem, the result, the words they use.

      Co-Founder & CEO 1 week

    2. 2

      Write the page

      Products described in their terms.

      Co-Founder & CEO 3 days

    3. 3

      Anonymised story

      One result.

      Research Scientist 1 week

    4. 4

      Small campaign

      LinkedIn to fragrance and flavor chemists.

      Co-Founder & CEO 1 month

    5. 5

      Route to the Foundry

      Tag briefs from this page.

      Founding Product Engineer 1 hour

    6. 6

      Decide

      Expand, repeat for energy, or stop.

      Co-Founder & CEO after 1 quarter

    First result
    1 quarter
    Cost
    A small ad budget and staff time
    Owner
    Co Founder & CEO

    Smallest test

    Publish the page and send it to ten fragrance chemists.

    Why it works

    Write one /flavor-and-fragrance page using the same products in that field's terms: routes to new aroma molecules from cheap building blocks, plates that optimise selectivity, and Foundry samples at gram scale. Add one anonymised story, run a small LinkedIn campaign to fragrance chemists, and measure enquiries against the pharma pages. If it works, repeat for energy and materials.

    A market where b12 already has customers but no competitor targets it can be cheaper to win than another pharma account.

    Risks and how to handle them

    • Spreading a small team too thin

      One page and one campaign only, with a clear stop rule.

    Builds on

    Evidence

    • Y Combinator profileYC Summer 2025, San Francisco, team of 3, one-liner "Making hard molecules for Pharma and Biotech". The profile says b12 runs paid pilots with large pharma (one reached full conversion in a single attempt), works with three large pharma companies including a top-3 global pharma that came back as a repeat customer, also serves flavor and fragrance, energy, biotech and university customers, and that an AI drug discovery platform has built b12's synthesis agent into its own pipeline. The Foundry opened in summer 2025 and is producing molecules for clients.Source
    • Your websiteSitemap (18 URLs, checked November 10, 2026): the homepage, the blog and its 5 posts, Foundry, three docs pages (Cobalt MCP, Foundry API, Selenium Route Search), publications, SLAS, support, team, terms, privacy, and /hex-background/, a "Hex Background Lab" design tool with an "Export SVG" button. There are no pricing, customer, case-study, academic-access, use-case, instrument or careers pages.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.

  • Rainbrand growth strategyThe brand layer for agents: Rainbrand turns a brand's website, guides, logos, fonts and voice into shared, versioned brand context that…23 growth ideas
  • Bestow growth strategyA life and annuity technology platform for carriers: digital quote, illustration and application, agent tools, underwriting engine and…22 growth ideas
  • Orium growth strategyA composable commerce and digital experience consultancy and systems integrator (formerly Myplanet) that designs and builds enterprise…22 growth ideas
  • Bild AI growth strategyAI takeoffs for Division 8: Bild AI reads plans and specs, extracts door schedules and hardware sets, sizes hardware, flags…22 growth ideas
  • Form Energy growth strategyAn American manufacturing and energy technology company that makes iron-air batteries able to store energy for 100 hours, for utilities…22 growth ideas
  • Gruve growth strategyAn AI services and infrastructure company: managed AI-powered security operations, Cisco-aligned security, data center and networking…21 growth ideas
  • SplitMetrics growth strategyMobile marketing software and services: SplitMetrics Acquire for Apple Ads management with the Iris AI agent, App Radar for ASO, Optimize…21 growth ideas
  • Aunt Laurie's growth strategyAn online gift basket shop in Bluffton, South Carolina, selling handwoven, reusable baskets of Southern and Lowcountry foods and gifts…21 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. b12 labs did not commission, review or endorse it, and Programmable is not affiliated with b12 labs. 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 b12 labs runs its business.

Something wrong, or do you represent b12 labs? 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

    Foundry brief form, on-site pharma proof, the free academic offer, one set of first-plate numbers, one set of product names, the Selenium docs link, sitemap cleanup, the hero video and a demo booking page.

  2. Days 31 to 608 ideas

    Credit pricing, the public benchmark, instrument pages, Chemspeed co-marketing, the simulation-first Cobalt trial, support standards, academic citations and the hard-target series.

  3. Days 61 to 906 ideas

    Failed-reaction data, the batch-to-Foundry bundle, then the big bets: self-serve Foundry orders, the Bay Area lab, an embedding partner program and the first non-pharma vertical.

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 natural-language control of lab robots with no coding. The most cited problem: bespoke coding languages and brittle scripts on expensive instruments.

  • Being able to review a protocol or route before it runs
  • Many rounds of reaction optimisation
  • In their words: “first plate”

Market

Competitive landscape

4 competitors compared: Chemify, Iktos, Enamine and PostEra.

  • Gap to exploit: Free, self-serve AI retrosynthesis for academic chemists
  • Chemify: Programme-based access rather than self-serve tools
  • Iktos: No free tier shown

Hiring signals

b12 is hiring one role, and it says a lot: a Founding Product Engineer to make a platform that already serves 5 of the top 10 pharma companies "feel fast, dependable, and coherent", owning the links between AI agents…

  • Founding Product Engineer (Full Stack): Reliability and usability for a platform used by top-10 pharma; agent and worker interfaces. Ideas 1, 9, 14…

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