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

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 10
  • This quarter · 7
  • Big bets · 3
  • Wildcards · 2
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. Halluminate
Halluminate logo

growth playbook · Nov 10, 2026

halluminate.ai (opens in a new tab)

Halluminate growth playbook

Diagnosis

Halluminate went from $0 to a mid-eight-figure run rate in 10 months, is profitable, and raised a $30 million Series A in October 2026 with only around ten people. Its buyers are a handful of frontier labs, so growth does not depend on a wide marketing funnel. It depends on three things: research credibility with those labs, a much larger and better-run network of finance experts (its own hiring speaks of a "1000x scale up"), and a steady supply of real deal data.

Halluminate is hiring for all three: a research lead for "public-facing benchmark and leaderboard efforts", a People Platform Lead and a Senior Recruiter for the expert network, and a Data Platform Lead to license data rooms, models and decks. The website has not caught up. The two finance benchmarks live only in blog posts, with no leaderboard, dataset or contact path. The expert network is one page and one form for ten role groups.

The data partnerships page doesn't say what data is wanted or what partners get. This playbook starts by giving the benchmarks a permanent home, role-specific expert pages and a clear data-partner offer. It then builds the expert funnel, community and referrals, and moves on to the big bets: a finance task map, an annual report and the first adjacent vertical.

Growth Score

7/10

Biggest Opportunity

Make the research do the selling: permanent, cited leaderboards for the finance benchmarks keep Halluminate in front of the labs that buy from it, and the same pages recruit the experts and data partners it needs.

Biggest Risk

Revenue is concentrated in a few frontier labs, and much larger rivals (Mercor, Scale AI, Surge AI) already publish finance and accounting leaderboards and recruit at far greater scale, so expert supply or lab attention could move elsewhere.

01

Start here

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

  1. 1

    Give the finance benchmarks a permanent leaderboard page, like the one Web Bench still has

    Put the Diligence Bench results table on its own page today, with a run date and a contact link.

    High impact First result in 1 month
  2. 2

    One page per expert role group, with its pay, a sample task and the time commitment

    Ship the private equity page first, since it pays the most and matches the flagship benchmarks.

    High impact First result in 1 month
  3. 3

    Show the four-step application as a timeline with a sample assessment, and measure drop-off at each step

    Pull last month's applicants and count how many reached each step.

    High impact First result in 3 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 · 4Big bets · 11Fill-ins · 5Reconsider · 2
Effort →

Showing 22 of 22 ideas

This week 10

Quick wins with a clear first action.

  • 01

    Halluminate's homepage sends visitors to "Our latest benchmark", the 88-problem Westworld Finance Diligence Bench. That post has charts and a trajectory explorer, but no leaderboard page, no dataset or sample tasks, no repository and no way to get in touch. Three months after publication its last line is still "We will release an accompanying paper soon." DealTrace, the private-equity benchmark, has none of these either. Yet the 2025 browser benchmark, Web Bench, had all of them: a leaderboard at webbench.ai, a Hugging Face dataset, a GitHub repo (still the company's most-starred, with 100 stars) and a contact line. That leaderboard's entries all date from May 2025. Rivals put finance leaderboards on their homepages: Mercor shows current top scores for APEX-Accounting and τ³-Banking, and Surge AI lists DAYJOB: Finance. Halluminate's research post exists to "Lead our public-facing benchmark and leaderboard efforts for frontier models", but the core finance results are still a one-time blog post.

    High impactMedium effort First result in 1 month Member of Technical Staff

    First action

    Put the Diligence Bench results table on its own page today, with a run date and a contact link.

    How to measure it

    Leaderboard citations plus inbound evaluation requests

    Now
    –
    Target
    10 citations and 5 qualified requests
    By
    90 days

    Steps

    1. 1

      Decide what is public

      Which scores, sample tasks and trajectories can be shown without exposing customer work or deal data.

      CEO 2 days

    2. 2

      Build the hub

      /leaderboards with one page per finance benchmark: table by model and harness, run dates, methods, citation block.

      Member of Technical Staff, Platform Engineering 1 week

    3. 3

      Release a sample

      Ten to twenty anonymized tasks per benchmark on GitHub and Hugging Face, with a README and license.

      Member of Technical Staff, Research / Post-Training 1 week

    4. 4

      Add a contact path

      "Evaluate your model or agent" form routed to the forward-deployed engineers.

      Member of Technical Staff, Forward Deployed Engineer 1 day

    5. 5

      Link it everywhere

      Homepage, research page, every benchmark post and the GitHub org profile.

      Member of Technical Staff, Platform Engineering 1 day

    6. 6

      Set a rerun rule

      Rerun within two weeks of any major model release and date-stamp each change.

      Member of Technical Staff, Research / Post-Training ongoing

    7. 7

      Decide on webbench.ai

      Refresh it or mark it archived and point to the hub.

      Member of Technical Staff, Research / Post-Training 1 day

    8. 8

      Track citations

      Monthly count of papers, model cards and posts citing the benchmarks.

      Member of Technical Staff, Finance Researcher monthly

    First result
    1 month
    Cost
    Staff time plus evaluation compute for reruns
    Owner
    Member of Technical Staff, Research / Post Training

    Smallest test

    Publish the Diligence Bench table as a dated page with a contact link and count inbound requests for 30 days.

    Why it works

    Build one /leaderboards hub on halluminate.ai with a page each for Diligence Bench and DealTrace. Each page gets the current scores by model and harness, the date of each run, a methods summary, a small public sample of tasks (anonymized, as the posts already are), the explorer, a citation block, and a "Have a model or agent to evaluate?" contact path. Rerun scores when major models ship and date-stamp every update. Link the hub from the homepage, the research page and every benchmark post. Then either refresh webbench.ai or mark it archived and point it to the hub.

    Labs and agent companies cite benchmarks that are kept up to date in their release notes and papers, and every citation is a free reminder of Halluminate to the few buyers that matter. A live leaderboard also gives the research team a reason to talk to each lab at every model launch. The same page is a credibility signal for finance experts deciding whether to apply, and for firms weighing a data partnership.

    Ready to use

    ChecklistFinance leaderboard page launch list

    [ ] /leaderboards hub linked from homepage and /research
    [ ] Diligence Bench page: scores by model and harness, run dates
    [ ] DealTrace page: stage-by-stage scores
    [ ] Methods summary and citation block on each
    [ ] Public sample tasks with README and license
    [ ] "Evaluate your model or agent" contact form
    [ ] Rerun rule and "last updated" date
    [ ] webbench.ai refreshed or marked archived

    Risks and how to handle them

    • A public sample lets labs train on the test set

      Publish a small sample and keep the scored set private, as private-benchmark leaderboards do.

    • A stale leaderboard hurts credibility

      A written rerun rule and a visible "last updated" date.

    Evidence

    • Your websiteWestworld Finance Diligence Bench (August 11, 2026), the homepage's "Our latest benchmark": 88 due-diligence problems written and reviewed by finance deal professionals, run in desktop environments over hundreds of steps; the best configuration's mean score is 0.61. The post links an interactive trajectory explorer, but no leaderboard page, dataset, GitHub repository or contact path, and ends: "We will release an accompanying paper soon."Source
    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
    • Your websiteWeb Bench (May 30, 2025), released "in partnership with Skyvern": an official leaderboard at webbench.ai, a Hugging Face dataset, a GitHub repository and a direct contact for teams interested in environments. Every leaderboard entry is still dated 2025-05-30.Source
    • HiringMember of Technical Staff, Research / Post-Training: "Lead our public-facing benchmark and leaderboard efforts for frontier models"; "Train open-source frontier models on our in-house environments to validate quality and build research insights"; "Publish findings in blogs and papers to share with the broader research community".Source
    • Competitor researchMercor's homepage: "5M+ experts", a "$90/hr average" rate and "$4M+ daily payouts"; benchmark leaderboards with current top scores (APEX-Accounting, τ³-Banking, APEX-SWE and others); open roles shown with hourly pay and recently-hired counts (for example Retail Trader at about $400 an hour).Source
    • Competitor researchSurge AI's homepage lists benchmarks including "DAYJOB: Finance" and an Expert Workforce offer; Scale AI's says "Our leaderboards run private benchmarks for the most ambitious AI companies" and that 90% of leading generative AI model builders use its platform.Source
  • 02

    A public DealTrace repository exists on GitHub, but it has one commit, no description and no README, and the DealTrace post doesn't link to it. The post is dated May 15, 2026 on the research page and June 17, 2026 on the blog index. Lab researchers who check a benchmark's provenance will notice both.

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

    First action

    Write the DealTrace README and correct the date today.

    How to measure it

    DealTrace repository stars

    Now
    –
    Target
    50
    By
    90 days

    Steps

    1. 1

      Pick the date

      One publication date on /research and /blogs.

      Member of Technical Staff, Research / Post-Training 15 minutes

    2. 2

      Write the README

      Purpose, stages, finding, citation, what is and isn't released.

      Member of Technical Staff, Research / Post-Training half a day

    3. 3

      Add license and topics

      Repository description, website link and topics.

      Member of Technical Staff, Platform Engineering 1 hour

    4. 4

      Link both ways

      Post to repository and repository to post.

      Member of Technical Staff, Platform Engineering 1 hour

    5. 5

      Audit the rest

      Same check for all posts and the eight repositories.

      Member of Technical Staff, Research / Post-Training 1 day

    6. 6

      Pin the best

      Pin finance repositories on the GitHub org profile.

      Member of Technical Staff, Research / Post-Training 15 minutes

    First result
    1 week
    Cost
    $0
    Owner
    Member of Technical Staff, Research / Post Training

    Smallest test

    Add the README and the link, then watch stars and referral traffic for a month.

    Why it works

    Add a README (what DealTrace measures, the five stages, the headline finding, how to cite it), a license, a description and topics, and link the repository from the post. Pick one publication date and use it on both pages. Do the same check across all eight posts and repositories.

    Benchmarks are judged on rigor. A finished repository and consistent dates are cheap and remove a reason to doubt the work.

    Risks and how to handle them

    • Releasing more than intended

      The README states exactly what is public and what stays private.

    Evidence

    • Your websiteGitHub organization: 12 followers and 8 public repositories. WebBench has 100 stars and westworld 19; the DealTrace repository has 0 stars, one commit, no description and no README shown. The ballot study repository (November 2026) publishes all code and trajectories.Source
    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
  • 03

    The Expert Network has ten role groups, from private equity at $150-$250 an hour to insurance and FP&A at $100-$200, but they share one page and one form, and the sitemap has no page per role. The Senior Recruiter is hired to run LinkedIn outbound and "A/B test paid acquisition". Every ad and message lands on the same generic page. Mercor shows each open role with its own pay and how many people were recently hired.

    High impactMedium effort First result in 1 month Senior Recruiter

    First action

    Ship the private equity page first, since it pays the most and matches the flagship benchmarks.

    How to measure it

    Visit-to-application rate on role pages

    Now
    –
    Target
    +40% against /experts
    By
    60 days

    Steps

    1. 1

      Pick five roles

      By demand from current projects.

      Strategic Project Lead 1 day

    2. 2

      Write each page

      Rate, sample task, commitment, earnings, testimonial.

      Senior Recruiter 1 week

    3. 3

      Build the template

      One template with the role pre-filled on the form.

      Member of Technical Staff, Platform Engineering 3 days

    4. 4

      Tag the traffic

      Tracked links per channel and role.

      Senior Recruiter 1 day

    5. 5

      Point outbound and ads

      Each sequence and ad set lands on its role page.

      Senior Recruiter ongoing

    6. 6

      Compare

      Completion and approval rate against the generic page.

      People Platform Lead monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Senior Recruiter

    Smallest test

    One private equity page used for one LinkedIn sequence, compared with the generic page for two weeks.

    Why it works

    Create /experts/private-equity, /experts/investment-banking, /experts/consulting, /experts/accounting and /experts/insurance (more later). Each page has its rate range, an anonymized sample task from that field, the weekly commitment, earnings for that group, one testimonial from the same background, and the form with the role pre-selected. Use them as the landing page for each outbound sequence and ad set.

    A private equity associate converts better on a page about PE work at PE rates than on one about ten professions. Separate pages also let the recruiter compare channels and roles cleanly, and they can rank for searches like "private equity AI training contract".

    Risks and how to handle them

    • Pay ranges change by project

      Show ranges and a "last updated" date.

    Evidence

    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /blogs, /research, /data-partnerships, /experts, seven blog posts and /llms.txt. No leaderboard page, no page per expert role, no case study page and no careers or about page.Source
    • HiringSenior Recruiter (posted September 4, 2026), "to build our finance expert network": "Data Driven Outbound: LinkedIn Recruiter / Sales Navigator, alumni and professional networks"; "A/B test paid acquisition, partnerships with expert networks and boutique firms"; "Design incentives and referral programs that attract and keep high-quality experts".Source
    • Competitor researchMercor's homepage: "5M+ experts", a "$90/hr average" rate and "$4M+ daily payouts"; benchmark leaderboards with current top scores (APEX-Accounting, τ³-Banking, APEX-SWE and others); open roles shown with hourly pay and recently-hired counts (for example Retail Trader at about $400 an hour).Source
  • 04

    Before their first project, an expert goes through a resume review, a 60-minute assessment, a 30-minute call and a 3-day mock problem build, then about two weeks (15-20 hours) of training. Milestones pay $1,500-$4,000, but that is buried in the page. The People Platform Lead is hired to support a "1000x scale up", and at that scale each lost step costs hundreds of qualified people.

    High impactLow effort First result in 3 weeks People Platform Lead

    First action

    Pull last month's applicants and count how many reached each step.

    How to measure it

    Share of applicants reaching first paid task

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Measure the funnel

      Counts and days between each step.

      People Platform Lead 2 days

    2. 2

      Draw the timeline

      Steps, days and pay at each.

      People Platform Lead 1 day

    3. 3

      Write samples

      One sample assessment question and one mock-build brief.

      Strategic Project Lead 2 days

    4. 4

      Publish

      Top of /experts and every role page.

      Member of Technical Staff, Platform Engineering 1 day

    5. 5

      Add nudges

      Email after three idle days at any step.

      People Platform Lead 2 days

    6. 6

      Review weekly

      Biggest drop-off step and one fix.

      People Platform Lead weekly

    First result
    3 weeks
    Cost
    Staff time only
    Owner
    People Platform Lead

    Smallest test

    Add the timeline only and compare assessment start rates for two weeks.

    Why it works

    Put a timeline at the top of /experts: days from applying to first pay, and what is paid at each step (training milestones, hiring bonus). Add a short sample of the assessment and the mock build so applicants know what to expect. Instrument each step in the funnel and send a nudge to anyone stuck for more than three days.

    Busy finance professionals drop out of long processes they can't see the end of. Showing that training is paid and how long each step takes keeps good applicants moving.

    Risks and how to handle them

    • Samples make the assessment easy to game

      Publish a retired question, not a live one.

    Evidence

    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
    • HiringPeople Platform Lead (posted October 6, 2026): "Our goal is to build the best network of RL experts in the world"; "Productizing our people platform and processes to support 1000x scale up"; "Organize community events for our people network to drive signups, engagement, and retention". A second People Platform Lead post from August 14 is also open, with a different pay range.Source
  • 05

    The application form already has an optional referral link, and the FAQ admits a referral link can be invalid, but the page names no reward and no way to get a link. The Senior Recruiter is asked to "Design incentives and referral programs". Deal professionals know other deal professionals, so referral is likely the cheapest channel for this network.

    High impactLow effort First result in 1 month Senior Recruiter

    First action

    Check last month's applications for broken referral links and fix the validation.

    How to measure it

    Referred share of new approved experts

    Now
    –
    Target
    30%
    By
    90 days

    Steps

    1. 1

      Fix validation

      Links checked on the form, with a clear error.

      Member of Technical Staff, Platform Engineering 1 day

    2. 2

      Set the reward

      Paid after the referred expert completes training.

      Senior Recruiter 1 day

    3. 3

      Issue links

      Personal link in each expert's dashboard.

      Member of Technical Staff, Platform Engineering 2 days

    4. 4

      State it publicly

      Reward and rules on /experts and in the FAQ.

      Senior Recruiter 1 hour

    5. 5

      Time the ask

      Email after each expert's first paid milestone.

      People Platform Lead 2 days

    6. 6

      Track quality

      Approval and retention of referred experts against other channels.

      Senior Recruiter monthly

    First result
    1 month
    Cost
    Referral rewards, paid only on success
    Owner
    Senior Recruiter

    Smallest test

    Email current experts one month of a doubled reward and count the referrals.

    Why it works

    Give every expert who passes the mock build a personal link from their dashboard, a clear reward paid when the referred expert completes training, and a monthly leaderboard. Validate links on the form so they cannot fail silently. Ask for referrals right after a first paid milestone, when goodwill is highest.

    Referred experts come pre-vetted by peers with the same background, approve at higher rates and cost far less than paid ads.

    Risks and how to handle them

    • Low-quality mass referrals

      Pay only after training is completed and cap rewards per person.

    Evidence

    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
    • HiringSenior Recruiter (posted September 4, 2026), "to build our finance expert network": "Data Driven Outbound: LinkedIn Recruiter / Sales Navigator, alumni and professional networks"; "A/B test paid acquisition, partnerships with expert networks and boutique firms"; "Design incentives and referral programs that attract and keep high-quality experts".Source
  • 06

    "Every RL environment we ship is built on real-world financial data", and the Data Platform Lead is hired to "Source, license, and purchase" data rooms, filings, models and decks. Yet the Data Partnerships page is a headline and a contact form. It names no data types, says nothing about what a partner receives or how data is anonymized, and doesn't point to the company's trust center. A cautious finance firm has nothing to forward to its compliance team.

    High impactLow effort First result in 1 month Data Platform Lead

    First action

    List the five data types the next environments need most.

    How to measure it

    Qualified inquiries a month

    Now
    –
    Target
    10
    By
    90 days

    Steps

    1. 1

      List the wants

      Five data types by priority.

      Data Platform Lead 1 day

    2. 2

      Explain the process

      Anonymization and compliance, step by step.

      Data Platform Lead 2 days

    3. 3

      State the terms

      How partners are paid, in plain ranges or models.

      CEO 2 days

    4. 4

      Link trust

      Trust center and security summary.

      Data Platform Lead 1 hour

    5. 5

      Add qualifying fields

      Data type, volume and deal years on the form.

      Member of Technical Staff, Platform Engineering 1 day

    6. 6

      Make the PDF

      One page for partner legal teams.

      Data Platform Lead 2 days

    First result
    1 month
    Cost
    Staff time and legal review
    Owner
    Data Platform Lead

    Smallest test

    Add the data types and the trust link, then compare inquiry quality for a month.

    Why it works

    Rewrite /data-partnerships around the data actually wanted (closed-deal data rooms, financial models, decks, workpapers), how payment works (license fee or revenue share), the anonymization and compliance pipeline step by step, a link to trust.halluminate.ai, and a one-page PDF for the partner's legal team. Add the "anonymized data from private transactions" behind Diligence Bench as proof of what the data becomes.

    Inbound data partners need to know they are a fit and that their data is safe before they write in. A specific page brings in the right firms and shortens legal review.

    Risks and how to handle them

    • Publishing terms limits negotiation

      Describe the models, not fixed prices.

    Evidence

    • Your websiteData Partnerships page: "Turn your company's data and workflows into a new revenue opportunity." A contact form (name, work email, company, role, website, message) and "Share Your Interest"; no examples of the data wanted, nothing on what a partner receives or how data is anonymized, no terms and no logos. Halluminate runs a separate trust center at trust.halluminate.ai.Source
    • HiringData Platform Lead (posted August 17, 2026): "Every RL environment we ship is built on real-world financial data"; "Source, license, and purchase real-world financial data and artifacts — data rooms, filings, models, decks"; "Build the anonymization and compliance pipeline that turns sensitive industry data into training-safe assets"; own provider relationships "including negotiation and renewal".Source
    • Your websiteWestworld Finance Diligence Bench (August 11, 2026), the homepage's "Our latest benchmark": 88 due-diligence problems written and reviewed by finance deal professionals, run in desktop environments over hundreds of steps; the best configuration's mean score is 0.61. The post links an interactive trajectory explorer, but no leaderboard page, dataset, GitHub repository or contact path, and ends: "We will release an accompanying paper soon."Source
  • 07

    Halluminate has published eight posts, including seven research items, but the site has no newsletter signup and no RSS feed. Lab researchers, finance experts and possible data partners who like one post have no way to hear about the next, except LinkedIn and X.

    Medium impactLow effort First result in 2 weeks Member of Technical Staff

    First action

    Turn on an RSS feed for the blog.

    How to measure it

    Subscribers

    Now
    –
    Target
    1,500
    By
    90 days

    Steps

    1. 1

      Add RSS

      Feed for all posts.

      Member of Technical Staff, Platform Engineering 2 hours

    2. 2

      Add signup

      On every post and /research.

      Member of Technical Staff, Platform Engineering 1 day

    3. 3

      Tag readers

      Researcher, finance professional, data partner.

      Member of Technical Staff, Research / Post-Training 1 hour

    4. 4

      Write issue one

      The finance benchmarks and what they show.

      Member of Technical Staff, Finance Researcher 1 day

    5. 5

      Route tags

      Finance tags get the expert invite; partners get the data offer.

      People Platform Lead 1 day

    6. 6

      Send with launches

      One issue per benchmark or leaderboard update.

      Member of Technical Staff, Research / Post-Training monthly

    First result
    2 weeks
    Cost
    Under $50 a month for an email tool
    Owner
    Member of Technical Staff, Research / Post Training

    Smallest test

    Add the signup to the Diligence Bench post and count signups in two weeks.

    Why it works

    Add an RSS feed and a short email signup ("New benchmarks and findings, about once a month") to every post and the research page, with optional tags for researcher, finance professional or data partner. Send one short digest with each new benchmark or leaderboard update.

    Each research launch then reaches an audience Halluminate owns, and the tags show which readers could become experts or partners.

    Risks and how to handle them

    • No time to write

      Send only with launches; three paragraphs each.

    Evidence

    • Your websiteResearch page and blog index: seven research items (Web Bench, BrowserBench, Westworld v1, DealTrace, a sim-to-real study, Diligence Bench and a November 2026 ballot study) among eight posts; the research page lists no leaderboards, datasets or papers; no newsletter signup or RSS feed.Source
  • 08

    The Diligence Bench post ends "We will release an accompanying paper soon", three months later there is no paper, and the research page lists no papers at all. The research hire is asked to "Publish findings in blogs and papers". Lab researchers cite papers; they rarely cite blog posts.

    High impactMedium effort First result in 1 month Member of Technical Staff

    First action

    Set a submission date and an outline this week.

    How to measure it

    Citations

    Now
    –
    Target
    10
    By
    6 months

    Steps

    1. 1

      Outline

      Reuse the post's sections and add methods.

      Member of Technical Staff, Research / Post-Training 2 days

    2. 2

      Write methods

      Expert authoring and review process.

      Member of Technical Staff, Finance Researcher 1 week

    3. 3

      Check claims

      Reproduce the headline numbers.

      Member of Technical Staff, Research / Post-Training 1 week

    4. 4

      Review

      What can be shown about the data.

      Data Platform Lead 2 days

    5. 5

      Post

      arXiv and the research page.

      Member of Technical Staff, Research / Post-Training 1 day

    6. 6

      Announce

      Digest, LinkedIn and X; send to lab contacts.

      CEO 1 day

    First result
    1 month
    Cost
    Staff time only
    Owner
    Member of Technical Staff, Research / Post Training

    Smallest test

    Post a short technical report first and expand it later.

    Why it works

    Turn the post into a short preprint (task design, expert authoring and review, harness setup, results, failure analysis), post it on arXiv, then link it from the post, the leaderboard page and the research page. Put a citation block on every benchmark page.

    A citable paper is how a benchmark enters other labs' evaluation suites, and inclusion there keeps Halluminate in front of its buyers.

    Risks and how to handle them

    • Paper reveals too much about the private set

      Describe the method and publish only sample tasks.

    Builds on

    Evidence

    • Your websiteWestworld Finance Diligence Bench (August 11, 2026), the homepage's "Our latest benchmark": 88 due-diligence problems written and reviewed by finance deal professionals, run in desktop environments over hundreds of steps; the best configuration's mean score is 0.61. The post links an interactive trajectory explorer, but no leaderboard page, dataset, GitHub repository or contact path, and ends: "We will release an accompanying paper soon."Source
    • HiringMember of Technical Staff, Research / Post-Training: "Lead our public-facing benchmark and leaderboard efforts for frontier models"; "Train open-source frontier models on our in-house environments to validate quality and build research insights"; "Publish findings in blogs and papers to share with the broader research community".Source
    • Your websiteResearch page and blog index: seven research items (Web Bench, BrowserBench, Westworld v1, DealTrace, a sim-to-real study, Diligence Bench and a November 2026 ballot study) among eight posts; the research page lists no leaderboards, datasets or papers; no newsletter signup or RSS feed.Source
  • 09

    Real visitors already pass Core Web Vitals (LCP 1.2 s at the 75th percentile), so this is not urgent. In the lab test, though, the mobile score is 90, and the analytics script accounts for 178 KB of a 258 KB page. PageSpeed suggests longer cache lifetimes (about 0.6 s) and less unused JavaScript (about 0.45 s).

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

    First action

    Defer the analytics script.

    How to measure it

    Mobile performance score

    Now
    –
    Target
    95+
    By
    30 days

    Steps

    1. 1

      Defer analytics

      Load after first render.

      Member of Technical Staff, Platform Engineering 1 hour

    2. 2

      Trim features

      Only the analytics features in use.

      Member of Technical Staff, Platform Engineering 1 hour

    3. 3

      Cache assets

      Long cache lifetimes for static files.

      Member of Technical Staff, Platform Engineering 1 hour

    4. 4

      Test key pages

      /experts and the benchmark posts on mobile.

      Member of Technical Staff, Platform Engineering 1 hour

    5. 5

      Confirm tracking

      Funnel events still record.

      People Platform Lead 1 hour

    6. 6

      Recheck

      PageSpeed after a week.

      Member of Technical Staff, Platform Engineering 15 minutes

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

    Smallest test

    Defer the script on /experts only and compare.

    Why it works

    Load the analytics script after the page renders and turn off features you don't use (such as session recording on research pages), set long cache lifetimes for static assets, and check the expert and benchmark pages, which mobile visitors from LinkedIn reach first.

    Expert recruiting leans on LinkedIn outreach, which many people open on their phones, so a faster first view helps a little at no cost.

    Risks and how to handle them

    • Lost events

      Check the funnel events after the change.

    Evidence

    • Google PageSpeedHomepage (tested November 10, 2026): mobile score 90, desktop 99; real visitors pass Core Web Vitals (LCP 1.2 s at the 75th percentile). The analytics script is 178 KB of a 258 KB page; suggested fixes are cache lifetimes (about 0.6 s) and unused JavaScript (about 0.45 s).Source
    • HiringSenior Recruiter (posted September 4, 2026), "to build our finance expert network": "Data Driven Outbound: LinkedIn Recruiter / Sales Navigator, alumni and professional networks"; "A/B test paid acquisition, partnerships with expert networks and boutique firms"; "Design incentives and referral programs that attract and keep high-quality experts".Source
  • 10

    Halluminate plans to grow "from 10 to 25+ people in next 6 months", and the Chief of Staff is asked to help with "selling important hires". But Careers is an off-site link to Ashby, and the site has no careers or about page. Two People Platform Lead posts are open at once with different pay ranges, which confuses candidates.

    Medium impactLow effort First result in 1 month Chief of Staff

    First action

    Merge the two People Platform Lead posts today.

    How to measure it

    Qualified applicants per role

    Now
    –
    Target
    +30%
    By
    90 days

    Steps

    1. 1

      Merge the duplicate

      One post, one pay range.

      Senior Recruiter 1 hour

    2. 2

      Write the page

      Mission, story, traction, research, how the team works.

      Chief of Staff 2 days

    3. 3

      Embed jobs

      Pull open roles from Ashby.

      Member of Technical Staff, Platform Engineering half a day

    4. 4

      Link it

      Navigation, Series A post and footer.

      Member of Technical Staff, Platform Engineering 1 hour

    5. 5

      Add to sitemap

      So it can be found in search.

      Member of Technical Staff, Platform Engineering 15 minutes

    6. 6

      Review

      Applicant source and quality.

      Senior Recruiter monthly

    First result
    1 month
    Cost
    $0
    Owner
    Chief of Staff

    Smallest test

    Publish a simple page with the embed and the Series A summary.

    Why it works

    Add /careers with the mission, the founders' story, the Series A, the research and the team's backgrounds, how the team works on-site in San Francisco, and the open roles pulled from Ashby. Close or merge the duplicate post and keep one pay range.

    Senior engineers and researchers compare employers on the company's own site, and a page with the research and traction makes a nine-person company look like the leader it is.

    Risks and how to handle them

    • Page goes stale

      Jobs pulled live from Ashby.

    Evidence

    • HiringChief of Staff: "scale from 10 to 25+ people in next 6 months", covering "important business partnerships" and "selling important hires". Strategic Project Lead: "Recruit and train our contractor finance staff" and "support $MMs in revenue this year". All 10 open roles are in San Francisco.Source
    • HiringPeople Platform Lead (posted October 6, 2026): "Our goal is to build the best network of RL experts in the world"; "Productizing our people platform and processes to support 1000x scale up"; "Organize community events for our people network to drive signups, engagement, and retention". A second People Platform Lead post from August 14 is also open, with a different pay range.Source
    • Your websiteSitemap (14 URLs, checked November 9, 2026): the homepage, /blogs, /research, /data-partnerships, /experts, seven blog posts and /llms.txt. No leaderboard page, no page per expert role, no case study page and no careers or about page.Source

This quarter 7

The core plays for the next three months.

  • 11

    The people most likely to read a post on where agents fail at private-equity diligence are finance professionals curious about AI, which is exactly who the expert network wants. Yet the Diligence Bench and DealTrace posts have no call to action, and their only related path is the Careers link.

    High impactLow effort First result in 1 month Senior Recruiter

    First action

    Add the box to the Diligence Bench and DealTrace posts.

    How to measure it

    Expert applications sourced from research

    Now
    –
    Target
    100 a month
    By
    90 days

    Steps

    1. 1

      Write the box

      One line, the rate, the link.

      Senior Recruiter 1 hour

    2. 2

      Add to posts

      Finance posts and leaderboard pages.

      Member of Technical Staff, Platform Engineering half a day

    3. 3

      Tag the source

      Tracked links per post.

      Senior Recruiter 1 hour

    4. 4

      Cut findings

      Three LinkedIn-sized findings per benchmark.

      Member of Technical Staff, Finance Researcher 2 days

    5. 5

      Post to finance audiences

      Company and expert accounts.

      Senior Recruiter weekly

    6. 6

      Compare quality

      Approval rate against other channels.

      People Platform Lead monthly

    First result
    1 month
    Cost
    $0
    Owner
    Senior Recruiter

    Smallest test

    Add the box to one post and count applications for two weeks.

    Why it works

    End each finance post and leaderboard page with a short box: "These tasks were written by deal professionals. Build the next ones: $150-$250 an hour, remote." Link it to the matching role page. Pull one striking finding (for example, flagged discrepancies reach the recommendation only 7-29% of the time) into LinkedIn posts aimed at finance audiences, each with the same invitation.

    The research proves the work is serious and interesting, which is the main doubt a senior banker has about "AI training" gigs.

    Risks and how to handle them

    • Research looks like recruiting

      Keep it to one box at the end.

    Builds on

    Evidence

    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
    • Your websiteWestworld Finance Diligence Bench (August 11, 2026), the homepage's "Our latest benchmark": 88 due-diligence problems written and reviewed by finance deal professionals, run in desktop environments over hundreds of steps; the best configuration's mean score is 0.61. The post links an interactive trajectory explorer, but no leaderboard page, dataset, GitHub repository or contact path, and ends: "We will release an accompanying paper soon."Source
    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
  • 12

    The Finance Researcher post asks for a public "taxonomy of AI tasks/workflows in finance", later extended "to other industries", built from conversations with leaders at banks, PE firms and hedge funds. DealTrace already shows where agents break (forecasting). A public map would turn scattered benchmarks into one reference.

    High impactMedium effort First result in 1 quarter Member of Technical Staff

    First action

    Draft the top-level list of 20 finance workflows.

    How to measure it

    Lab and partner conversations citing the map

    Now
    –
    Target
    10
    By
    2 quarters

    Steps

    1. 1

      Draft the list

      20 workflows and their sub-tasks.

      Member of Technical Staff, Finance Researcher 1 week

    2. 2

      Validate

      Ten conversations with practitioners.

      Member of Technical Staff, Finance Researcher 3 weeks

    3. 3

      Score

      Tag each task with benchmark results or "untested".

      Member of Technical Staff, Research / Post-Training 2 weeks

    4. 4

      Build the page

      Interactive map with links out.

      Member of Technical Staff, Platform Engineering 2 weeks

    5. 5

      Launch

      Digest, LinkedIn and press.

      CEO 1 week

    6. 6

      Refresh

      With each benchmark release.

      Member of Technical Staff, Finance Researcher quarterly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Member of Technical Staff, Finance Researcher

    Smallest test

    Publish the list of workflows as a post and see who responds.

    Why it works

    Publish an interactive map of the main finance workflows (diligence, modeling, memo writing, reconciliation and so on), each tagged with today's agent performance from Halluminate's benchmarks, with a short explainer per task. Update it with every benchmark release. Each cell links to the benchmark, the matching expert role page and the data-partner offer.

    Labs use it to plan what to train next, which is the conversation Halluminate wants to have. Finance leaders use it to judge agents, and it gives the press a reason to quote Halluminate.

    Risks and how to handle them

    • Reveals a lab's training priorities

      Use only public benchmark results.

    Builds on

    Evidence

    • HiringMember of Technical Staff, Finance Researcher (posted August 14, 2026): "Define and publish taxonomy of AI tasks/workflows in finance, and expanding that playbook to other industries"; "Talk with real-world financial services leaders at major banks, private equity firms, hedge funds"; define "state of the art benchmarks, leaderboards, new research directions".Source
    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
  • 13

    The experts page shows logos from Harvard Business School, Stanford GSB and Wharton, and all three testimonials come from those programs. The Senior Recruiter is asked to use "alumni and professional networks" and partnerships with "boutique firms". MBA students and recent alumni from banking and PE fit the 20-hour, six-week commitment well.

    High impactMedium effort First result in 6 weeks Senior Recruiter

    First action

    Ask the three testimonial experts to introduce their school's finance club.

    How to measure it

    Approved experts via clubs and alumni groups

    Now
    –
    Target
    75
    By
    2 quarters

    Steps

    1. 1

      Pick ten groups

      Clubs and alumni groups that match current projects.

      Senior Recruiter 1 week

    2. 2

      Build the talk

      30 minutes from DealTrace and Diligence Bench.

      Member of Technical Staff, Finance Researcher 1 week

    3. 3

      Make club pages

      Tracked landing page per group.

      Member of Technical Staff, Platform Engineering 2 days

    4. 4

      Set club rewards

      Per approved expert, paid to the club.

      People Platform Lead 1 day

    5. 5

      Run sessions

      Two a month.

      Senior Recruiter ongoing

    6. 6

      Extend to boutiques

      Same format for advisory firms.

      Senior Recruiter next quarter

    First result
    6 weeks
    Cost
    Club rewards and event costs, a few thousand dollars a quarter
    Owner
    Senior Recruiter

    Smallest test

    One session with one club, then measure approved experts.

    Why it works

    Run a small partner program with finance and PE clubs at business schools and with alumni groups: an evening session on where agents fail at deal work (built from DealTrace), a club-specific landing page, and a referral reward for the club. Repeat the format with boutique advisory firms between deals.

    Club endorsements carry trust that ads lack, and each session reaches dozens of qualified people at once.

    Risks and how to handle them

    • Employer conflicts for working professionals

      State outside-work rules clearly in the FAQ.

    Builds on

    Evidence

    • HiringSenior Recruiter (posted September 4, 2026), "to build our finance expert network": "Data Driven Outbound: LinkedIn Recruiter / Sales Navigator, alumni and professional networks"; "A/B test paid acquisition, partnerships with expert networks and boutique firms"; "Design incentives and referral programs that attract and keep high-quality experts".Source
    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
  • 14

    Training already runs across five levels, with paid milestones, but the levels end once training does. The People Platform Lead is asked to run community events for "signups, engagement, and retention". Experts in this network are high earners with other options; status and peers keep them as much as pay does.

    High impactMedium effort First result in 2 months People Platform Lead

    First action

    Define the rank criteria from current quality data.

    How to measure it

    Active experts still working at week 12

    Now
    –
    Target
    +20%
    By
    2 quarters

    Steps

    1. 1

      Define ranks

      Criteria, pay band and access per rank.

      People Platform Lead 1 week

    2. 2

      Build the badge

      Rank in the dashboard and a shareable credential.

      Member of Technical Staff, Platform Engineering 2 weeks

    3. 3

      First session

      Research team on what expert tasks changed.

      Member of Technical Staff, Research / Post-Training monthly

    4. 4

      Dinners

      Small in-person events in two cities.

      People Platform Lead quarterly

    5. 5

      Announce promotions

      Monthly, with consent.

      People Platform Lead monthly

    6. 6

      Measure

      Retention by rank and cohort.

      People Platform Lead monthly

    First result
    2 months
    Cost
    Event costs and staff time
    Owner
    People Platform Lead

    Smallest test

    One virtual session for active experts; compare their activity with non-attendees.

    Why it works

    Extend the five levels into a lasting rank (Builder, Senior Builder, Reviewer, Lead and so on) tied to quality scores, with the pay band, project access and a LinkedIn-ready credential for each. Hold a monthly virtual session where the research team shows how experts' tasks changed model performance, plus small in-person dinners in New York and San Francisco.

    Visible progression and seeing their work matter keep strong experts active longer, and a shareable credential markets the network to their peers.

    Risks and how to handle them

    • Confidential customer work in sessions

      Show only public benchmark results.

    Evidence

    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
    • HiringPeople Platform Lead (posted October 6, 2026): "Our goal is to build the best network of RL experts in the world"; "Productizing our people platform and processes to support 1000x scale up"; "Organize community events for our people network to drive signups, engagement, and retention". A second People Platform Lead post from August 14 is also open, with a different pay range.Source
  • 15

    The platform engineer is asked to build "platform analytics for environment cost, hours, bottlenecks, and SME management", and the recruiter to A/B test paid acquisition. Without one view from first click to productive hours, the team can't tell which channel or role page actually produces delivered work.

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

    First action

    Agree the definition of a "productive hour".

    How to measure it

    Recruiting and onboarding cost per productive hour

    Now
    –
    Target
    -20%
    By
    2 quarters

    Steps

    1. 1

      Define the metric

      Productive hour and quality bar.

      Strategic Project Lead 2 days

    2. 2

      Join the data

      Source, steps, training, quality and hours.

      Member of Technical Staff, Platform Engineering 2 weeks

    3. 3

      Build the view

      By channel, role and cohort.

      Member of Technical Staff, Platform Engineering 1 week

    4. 4

      Review weekly

      Recruiting and project leads together.

      People Platform Lead weekly

    5. 5

      Shift budget

      Toward the best channels.

      Senior Recruiter monthly

    6. 6

      Flag bottlenecks

      Alerts when a step slows.

      Member of Technical Staff, Platform Engineering 1 week

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

    Smallest test

    A spreadsheet version for last quarter's cohort.

    Why it works

    Join recruiting source, application steps, training completion, quality scores and paid hours into one dashboard. The key number is cost per productive expert hour by channel, role group and cohort, so recruiting budget follows the channels that deliver.

    Revenue is delivered work, and delivered work depends on experts. Measuring the full chain shows where recruiting money and onboarding fixes pay back.

    Risks and how to handle them

    • Sensitive contractor data

      Role-based access and aggregate views.

    Builds on

    Evidence

    • HiringMember of Technical Staff, Platform Engineering: "Build software that 10–100x's the quality and throughput of RL environment creation"; verifiers "for decks, Excel modeling, quantitative trading, and beyond"; "Build platform analytics for environment cost, hours, bottlenecks, and SME management".Source
    • HiringSenior Recruiter (posted September 4, 2026), "to build our finance expert network": "Data Driven Outbound: LinkedIn Recruiter / Sales Navigator, alumni and professional networks"; "A/B test paid acquisition, partnerships with expert networks and boutique firms"; "Design incentives and referral programs that attract and keep high-quality experts".Source
  • 16

    Halluminate needs "data rooms, filings, models, decks". DealTrace already shows what one anonymized deal package becomes: a staged review that exposes where agents fail. Mid-market PE firms and boutique banks hold years of closed-deal material with no further use, and the data partnerships page promises "a new revenue opportunity".

    High impactMedium effort First result in 1 quarter Data Platform Lead

    First action

    Draft the before-and-after brief from one DealTrace package.

    How to measure it

    New data partners signed

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Write the brief

      Before-and-after, compliance steps, payment model.

      Data Platform Lead 1 week

    2. 2

      Build the list

      50 mid-market PE firms and boutiques.

      Data Platform Lead 1 week

    3. 3

      Ask for warm intros

      Willing network members, within employer rules.

      People Platform Lead 2 weeks

    4. 4

      Offer a pilot

      One closed deal, paid.

      Data Platform Lead ongoing

    5. 5

      Legal pack

      Standard license and anonymization terms.

      Chief of Staff 3 weeks

    6. 6

      Review

      Pipeline and cost per usable deal.

      Data Platform Lead monthly

    First result
    1 quarter
    Cost
    License fees and legal time
    Owner
    Data Platform Lead

    Smallest test

    Send the brief to ten firms and count meetings.

    Why it works

    Build a short outbound program for mid-market PE firms and advisory boutiques. A two-page brief shows a sanitized before-and-after: deal materials in, anonymized environment and benchmark out, with the compliance steps and the payment model. Offer a paid pilot on one closed deal. Use expert-network members who work at such firms (with their consent and within their employer's rules) for warm introductions.

    Firms say yes when they can see exactly what happens to their data and that it is safe; a concrete example beats a generic form.

    Risks and how to handle them

    • Confidentiality obligations on old deals

      Partner's counsel confirms rights before transfer; anonymize before use.

    Builds on

    Evidence

    • HiringData Platform Lead (posted August 17, 2026): "Every RL environment we ship is built on real-world financial data"; "Source, license, and purchase real-world financial data and artifacts — data rooms, filings, models, decks"; "Build the anonymization and compliance pipeline that turns sensitive industry data into training-safe assets"; own provider relationships "including negotiation and renewal".Source
    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
    • Your websiteData Partnerships page: "Turn your company's data and workflows into a new revenue opportunity." A contact form (name, work email, company, role, website, message) and "Share Your Interest"; no examples of the data wanted, nothing on what a partner receives or how data is anonymized, no terms and no logos. Halluminate runs a separate trust center at trust.halluminate.ai.Source
  • 17

    Halluminate's llms.txt mentions a case study, Yutori, which "post-trained its Navigator web agent using Westworld environments". But the sitemap has no case study page, and the homepage shows no customer logos or proof beyond the benchmark. The lab customers can't be named; this published one can.

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

    First action

    Confirm with Yutori what can be published.

    How to measure it

    Inbound inquiries from agent companies

    Now
    –
    Target
    5 a quarter
    By
    2 quarters

    Steps

    1. 1

      Get approval

      Scope and quote from the partner.

      Member of Technical Staff, Forward Deployed Engineer 1 week

    2. 2

      Write it

      Problem, environments, result, quote.

      Member of Technical Staff, Forward Deployed Engineer 3 days

    3. 3

      Publish

      /customers page in the sitemap.

      Member of Technical Staff, Platform Engineering 1 day

    4. 4

      Add proof to homepage

      One line of traction and the story.

      CEO 1 day

    5. 5

      Update llms.txt

      Link the page.

      Member of Technical Staff, Platform Engineering 15 minutes

    6. 6

      Ask for more

      Other public partners.

      Member of Technical Staff, Forward Deployed Engineer quarterly

    First result
    1 month
    Cost
    $0
    Owner
    Member of Technical Staff, Forward Deployed Engineer

    Smallest test

    Add the traction line to the homepage this week.

    Why it works

    Publish /customers/yutori with the problem, the environments used, what changed in the agent and a quote, and add it to the homepage under a short proof line (four of the five leading US closed-source labs, mid-eight-figure run rate, profitable). Ask other published partners, such as the Web Bench co-release partner, for permission to add a story.

    Agent companies and enterprise teams that may buy later look for proof from a peer, and a citable page also feeds the AI-assistant summaries that llms.txt is meant for.

    Risks and how to handle them

    • Partner declines

      Publish an anonymized version with numbers only.

    Evidence

    • Your websitellms.txt: three offerings (RL environments for enterprise and financial software, expert-authored tasks graded with verifiable rewards, proprietary and open-source benchmarks) and one case study, Yutori, which "post-trained its Navigator web agent using Westworld environments". The sitemap has no case study page.Source
    • Your websiteHomepage: "Halluminate is a data research lab building benchmarks and RL environments to push the frontier of AI in knowledge work", with financial services and consulting as the first focus. Navigation: Research, Expert Network, Careers (off-site, to Ashby) and Data Partnerships. Calls to action: "Our latest benchmark" and a "Contact Us" email link. No customer logos.Source
    • Your websiteSeries A post (October 1, 2026): $30 million led by Oak HC/FT, $38.5 million raised in total; "Built and scaled RL environments with four of the five leading closed-source U.S. AI labs"; grew "from $0 to a mid-eight-figure revenue run rate while remaining strongly profitable" in 10 months; "rapidly expanding to adjacent verticals", aiming at finance, consulting, insurance and operations.Source

Big bets 3

Larger plays that can change the trajectory.

  • 18

    The Series A post says Halluminate is "rapidly expanding to adjacent verticals" across consulting, insurance and operations, and the Finance Researcher is to extend the task map "to other industries". The expert page already recruits insurance, accounting and compliance professionals at $100-$200 an hour, so part of the supply is already being built.

    High impactHigh effort First result in 2 quarters CEO

    First action

    Score three candidate verticals on demand, data and expert supply.

    How to measure it

    First paid lab project in the new vertical

    Now
    –
    Target
    1 signed
    By
    2 quarters

    Steps

    1. 1

      Score options

      Demand, data, expert supply.

      Chief of Staff 2 weeks

    2. 2

      Map tasks

      Task map for the chosen vertical.

      Member of Technical Staff, Finance Researcher 1 month

    3. 3

      Recruit a cohort

      Role page and 30 experts.

      Senior Recruiter 1 month

    4. 4

      Secure data

      One data partner.

      Data Platform Lead 2 months

    5. 5

      Ship a benchmark

      Public, with a leaderboard.

      Member of Technical Staff, Research / Post-Training 2 months

    6. 6

      Take it to labs

      Scope the first environment project.

      Member of Technical Staff, Forward Deployed Engineer ongoing

    First result
    2 quarters
    Cost
    Expert pay, data licensing and research time
    Owner
    CEO

    Smallest test

    A 20-task pilot benchmark in the chosen vertical.

    Why it works

    Choose one vertical (insurance claims and underwriting, or consulting) using three tests: lab demand, available data and expert supply. Then launch it the way finance worked: one public benchmark to establish authority, a role page and expert cohort, and one data partner, before selling environments to labs.

    Repeating the finance pattern (research, experts, data) lowers the risk of a new vertical and lets the benchmark open lab conversations.

    Risks and how to handle them

    • Distracts from finance with a small team

      Ring-fence one project pod and review at 90 days.

    Builds on

    Evidence

    • Your websiteSeries A post (October 1, 2026): $30 million led by Oak HC/FT, $38.5 million raised in total; "Built and scaled RL environments with four of the five leading closed-source U.S. AI labs"; grew "from $0 to a mid-eight-figure revenue run rate while remaining strongly profitable" in 10 months; "rapidly expanding to adjacent verticals", aiming at finance, consulting, insurance and operations.Source
    • HiringMember of Technical Staff, Finance Researcher (posted August 14, 2026): "Define and publish taxonomy of AI tasks/workflows in finance, and expanding that playbook to other industries"; "Talk with real-world financial services leaders at major banks, private equity firms, hedge funds"; define "state of the art benchmarks, leaderboards, new research directions".Source
    • Your websiteExpert Network page, "Help Train Financial Superintelligence": one page and one application form for every role group; $150-$250 an hour for private equity and investment banking, $100-$200 for consulting, accounting, FP&A, compliance, insurance and others; median monthly earnings of $3,800; at least 20 hours a week for 6 weeks; about two weeks of self-paced training (15-20 hours) across five levels, with milestones paying $1,500-$4,000; four steps (resume review, 60-minute assessment, 30-minute call, 3-day mock problem build); three testimonials; an optional referral link, with an FAQ note that a referral link can be invalid.Source
  • 19

    Halluminate now has two finance benchmarks with findings the industry cares about: forecasting decides the verdict, and red flags reach the recommendation only 7-29% of the time. The Finance Researcher talks to leaders at banks, PE firms and hedge funds. The Series A got Fortune coverage, so the press is listening.

    High impactHigh effort First result in 2 quarters Member of Technical Staff

    First action

    Pick the five questions the first report will answer.

    How to measure it

    Press mentions in launch month

    Now
    –
    Target
    10
    By
    2 quarters

    Steps

    1. 1

      Pick questions

      Five that finance and labs both care about.

      Member of Technical Staff, Finance Researcher 1 week

    2. 2

      Run interviews

      20 finance leaders.

      Member of Technical Staff, Finance Researcher 2 months

    3. 3

      Rerun benchmarks

      Current models on both finance benchmarks.

      Member of Technical Staff, Research / Post-Training 1 month

    4. 4

      Survey experts

      What experts see while building tasks.

      People Platform Lead 1 month

    5. 5

      Write and design

      Report and landing page.

      Member of Technical Staff, Finance Researcher 1 month

    6. 6

      Launch

      Event, press briefings, digest.

      CEO 2 weeks

    First result
    2 quarters
    Cost
    Design, event and compute, roughly $20k-$40k
    Owner
    Member of Technical Staff, Finance Researcher

    Smallest test

    A short "state of" post using the two existing benchmarks.

    Why it works

    Publish a yearly report combining benchmark results over time, the task map, interviews with finance leaders on agent adoption, and what experts see while building tasks. Launch it with an event in New York and briefings for finance and AI press.

    An annual report makes Halluminate the default source on agents in finance. Labs, investors, finance firms and future experts and data partners all read it.

    Risks and how to handle them

    • Seen as favoring a lab

      Publish methods and rerun all models the same way.

    Builds on

    Evidence

    • Your websiteDealTrace: "a staged review of real private-equity deals" over ten deal packages and five stages. "Forecast quality is the capability that decides the verdict": supplying the true forecast lifts the judge score "from 53 to 80", and a flagged discrepancy reaches the recommendation "only 7 to 29 percent of the time". The post has no leaderboard, dataset, repository link or call to action, and is dated May 15, 2026 on /research but June 17, 2026 on /blogs.Source
    • HiringMember of Technical Staff, Finance Researcher (posted August 14, 2026): "Define and publish taxonomy of AI tasks/workflows in finance, and expanding that playbook to other industries"; "Talk with real-world financial services leaders at major banks, private equity firms, hedge funds"; define "state of the art benchmarks, leaderboards, new research directions".Source
    • PressFortune (October 1, 2026): a nine-person team; four of the top five closed-source U.S. AI labs are paying customers; the company is focusing on a small group of frontier labs rather than broad enterprise expansion for now; environment complexity needs to roughly double every six to eight months.Source
  • 20

    The CEO told Fortune that environment complexity needs to roughly double every six to eight months, and called the company's IP the ability to keep producing that complexity "generation after generation". Buyers can't see that progress today. The platform team is building analytics on environment cost and hours anyway.

    High impactHigh effort First result in 2 quarters CTO

    First action

    Agree five components of the index from existing environment data.

    How to measure it

    Lab RFPs or reviews using the index

    Now
    –
    Target
    2
    By
    3 quarters

    Steps

    1. 1

      Define components

      Five measurable components.

      CTO 2 weeks

    2. 2

      Backfill

      Score past benchmarks and environments.

      Member of Technical Staff, Platform Engineering 1 month

    3. 3

      Validate

      Check the index predicts model failure rates.

      Member of Technical Staff, Research / Post-Training 1 month

    4. 4

      Publish

      Method and curve on /research.

      Member of Technical Staff, Research / Post-Training 2 weeks

    5. 5

      Use in sales

      Each proposal states the index.

      Member of Technical Staff, Forward Deployed Engineer ongoing

    6. 6

      Update

      With each environment generation.

      CTO quarterly

    First result
    2 quarters
    Cost
    Research and engineering time
    Owner
    CTO

    Smallest test

    Score the two finance benchmarks and Westworld v1 internally.

    Why it works

    Define a simple, published index of environment complexity (steps per task, tools and apps involved, documents per task, verifier depth, expert hours per task) and report it for each benchmark and environment generation. Show the curve against the six-to-eight-month doubling target on the research page.

    It turns Halluminate's main promise into a number labs can track and compare, and it is hard for generalist vendors to match.

    Risks and how to handle them

    • Index is gamed or disputed

      Validate against model failure rates and publish the method.

    Builds on

    Evidence

    • PressFortune (October 1, 2026): a nine-person team; four of the top five closed-source U.S. AI labs are paying customers; the company is focusing on a small group of frontier labs rather than broad enterprise expansion for now; environment complexity needs to roughly double every six to eight months.Source
    • HiringMember of Technical Staff, Platform Engineering: "Build software that 10–100x's the quality and throughput of RL environment creation"; verifiers "for decks, Excel modeling, quantitative trading, and beyond"; "Build platform analytics for environment cost, hours, bottlenecks, and SME management".Source

Wildcards 2

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

  • 21

    Halluminate is rightly focused on a few frontier labs, with enterprise not a priority yet. But its forward-deployed engineers already serve "AI Labs and Enterprises", and it has worked with agent companies (Yutori on Westworld, Skyvern on Web Bench). Finance-agent startups need a credible score to show their own buyers, and the private Diligence Bench set already exists.

    Medium impactMedium effort First result in 1 quarter Member of Technical Staff

    First action

    Offer it to three finance-agent companies as a pilot.

    How to measure it

    Paid evaluations delivered

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Define the package

      Inputs, report, price, turnaround.

      Member of Technical Staff, Forward Deployed Engineer 1 week

    2. 2

      Automate runs

      Standard harness for outside agents.

      Member of Technical Staff, Platform Engineering 3 weeks

    3. 3

      Report template

      Scores and failure analysis.

      Member of Technical Staff, Research / Post-Training 1 week

    4. 4

      Pilot

      Three companies.

      Member of Technical Staff, Forward Deployed Engineer 1 month

    5. 5

      List on leaderboard

      Opt-in for evaluated agents.

      Member of Technical Staff, Research / Post-Training ongoing

    6. 6

      Decide

      Keep, price up or stop at 90 days.

      CEO 1 day

    First result
    1 quarter
    Cost
    Compute and engineering time
    Owner
    Member of Technical Staff, Forward Deployed Engineer

    Smallest test

    One paid pilot evaluation.

    Why it works

    Package a fixed-price evaluation: run the customer's agent on the private Diligence Bench and DealTrace sets, deliver a scored report with failure analysis, and (optionally) list them on the public leaderboard. Keep it standard, with no custom work, so it doesn't pull the team away from lab projects.

    It earns revenue from work already done, puts more agents on the leaderboard and creates relationships with companies that may later need custom environments.

    Risks and how to handle them

    • Distracts from lab work

      Standard package only, with a 90-day stop rule.

    • Leaks the private set

      Run agents on Halluminate infrastructure only.

    Builds on

    Evidence

    • HiringMember of Technical Staff, Forward Deployed Engineer: "Our FDEs will interface directly with our customers at AI Labs and Enterprises"; "Translate customer requirements into concrete technical specs our platform and project pods can build against".Source
    • PressFortune (October 1, 2026): a nine-person team; four of the top five closed-source U.S. AI labs are paying customers; the company is focusing on a small group of frontier labs rather than broad enterprise expansion for now; environment complexity needs to roughly double every six to eight months.Source
    • Your websitellms.txt: three offerings (RL environments for enterprise and financial software, expert-authored tasks graded with verifiable rewards, proprietary and open-source benchmarks) and one case study, Yutori, which "post-trained its Navigator web agent using Westworld environments". The sitemap has no case study page.Source
    • Your websiteWeb Bench (May 30, 2025), released "in partnership with Skyvern": an official leaderboard at webbench.ai, a Hugging Face dataset, a GitHub repository and a direct contact for teams interested in environments. Every leaderboard entry is still dated 2025-05-30.Source
  • 22

    Web Bench, still Halluminate's most-starred repository, was co-released "in partnership with Skyvern", and a partner doubles a launch's reach. The Chief of Staff owns "important business partnerships" and "high leverage industry initiatives". For finance, the strongest partners aren't AI companies but the institutions finance professionals already trust.

    Medium impactMedium effort First result in 2 quarters Chief of Staff

    First action

    List five candidate institutions and the task each could own.

    How to measure it

    Expert applications from the partner's audience

    Now
    –
    Target
    300
    By
    3 quarters

    Steps

    1. 1

      List candidates

      Five institutions and their fit.

      Chief of Staff 1 week

    2. 2

      Pitch

      Co-authored benchmark and shared launch.

      CEO 1 month

    3. 3

      Agree terms

      Ownership, branding and data rules.

      Chief of Staff 1 month

    4. 4

      Build tasks

      Partner tasks in Halluminate environments.

      Member of Technical Staff, Finance Researcher 2 months

    5. 5

      Launch

      Joint release and leaderboard.

      Member of Technical Staff, Research / Post-Training 2 weeks

    6. 6

      Invite members

      Partner audience to the expert network.

      Senior Recruiter ongoing

    First result
    2 quarters
    Cost
    Staff time and possible partner fee
    Owner
    Chief of Staff

    Smallest test

    A co-branded webinar on the existing benchmarks with one partner.

    Why it works

    Approach a finance professional body or a well-known financial modeling training provider to co-author the next benchmark. They contribute exam-style tasks and credibility; Halluminate contributes the environments, evaluation and leaderboard. Both promote it to their audiences, and the partner's members are invited to the expert network.

    The partner's name makes the benchmark authoritative to finance readers and the press, and its members are an ideal pool of experts.

    Risks and how to handle them

    • Slow institutional approvals

      Start with a training provider, which moves faster.

    Builds on

    Evidence

    • Your websiteWeb Bench (May 30, 2025), released "in partnership with Skyvern": an official leaderboard at webbench.ai, a Hugging Face dataset, a GitHub repository and a direct contact for teams interested in environments. Every leaderboard entry is still dated 2025-05-30.Source
    • Your websiteGitHub organization: 12 followers and 8 public repositories. WebBench has 100 stars and westworld 19; the DealTrace repository has 0 stars, one commit, no description and no README shown. The ballot study repository (November 2026) publishes all code and trajectories.Source
    • HiringChief of Staff: "scale from 10 to 25+ people in next 6 months", covering "important business partnerships" and "selling important hires". Strategic Project Lead: "Recruit and train our contractor finance staff" and "support $MMs in revenue this year". All 10 open roles are in San Francisco.Source
More example playbooks and reviews
ProgrammableDigital

Show Halluminate’s product working

Turn screenshots or the website into an animated product demo in its own brand, for the homepage, launch posts and outreach in this playbook.

Make a demo (opens programmable.digital in a new tab)

More growth playbooks

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

  • MaintainX growth strategyA mobile-first CMMS and enterprise asset management platform for industrial and frontline teams: work orders, preventive maintenance…21 growth ideas
  • Reddit growth strategyReddit runs reddit.com, a network of 100,000+ active communities where people post, comment and vote, funded mainly by advertising…22 growth ideas
  • Virtualitics growth strategyAI-native readiness software for defense: Virtualitics Iris and readiness applications for maintenance, materiel, personnel, supply chain…21 growth ideas
  • FitFeast growth strategyAn Indian protein snack brand: flavoured peanut butters, protein chips, protein shake sachets, protein bars and cookies, third-party…21 growth ideas
  • Autrun growth strategyBuilds and runs automation systems for growing organizations: integrations between ERPs, CRMs and cloud tools on Make, an on-premise Proxy…22 growth ideas
  • Bozzuto growth strategyA multifamily real estate company that develops, builds, manages and insures apartment communities, with about 130,000 apartments under…20 growth ideas
  • Centennial growth strategyA national owner and operator of retail and mixed-use destinations, now "Centennial, powered by Lincoln", providing property and asset…21 growth ideas
  • Scandinavian Biolabs growth strategyA Danish hair-care brand selling the drug-free Bio-Pilixin serum, Hairline Booster, shampoos, conditioners, supplements and routines…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. Halluminate did not commission, review or endorse it, and Programmable is not affiliated with Halluminate. 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 Halluminate runs its business.

Something wrong, or do you represent Halluminate? 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 3010 ideas

    Give the finance benchmarks a permanent home and finish the DealTrace repository; role-specific expert pages, a clearer path through the four-step process, and a referral program that works; a real data-partner offer, a research digest, the paper, the speed and careers basics.

  2. Days 31 to 607 ideas

    Recruit experts from the research itself, publish the finance task map, open campus and alumni channels, add public expert levels and community, instrument the expert funnel, pitch data rooms with DealTrace as proof, and publish the Yutori case study.

  3. Days 61 to 905 ideas

    Commit to the big bets and test the wildcards: the first adjacent vertical, an annual report, a public complexity index, private evaluations for agent companies and a co-released benchmark.

The research behind the plan

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

Customers

What customers say

Customers value long, realistic finance workflows written by real deal professionals. The most cited problem: benchmarks saturate, so environment complexity must keep rising.

  • Environments built on real (anonymized) deal data
  • Finding enough qualified finance experts to write and grade tasks
  • In their words: “RL environments”

Market

Competitive landscape

3 competitors compared: Mercor, Scale AI and Surge AI.

  • Gap to exploit: A trusted, permanent public leaderboard for long-horizon finance work
  • Mercor: Generalist; finance is one vertical among many
  • Scale AI: Broad focus; less depth in finance workflows

Hiring signals

Halluminate is hiring 10 roles, all in San Francisco, and they show where growth will come from.

  • Member of Technical Staff, Research / Post-Training: Public benchmarks, leaderboards and papers. Ideas 1, 2, 8.
  • Member of Technical Staff, Finance Researcher: A published finance task taxonomy. Ideas 12, 19.
  • People Platform Lead: Expert network built for a 1000x scale up. Ideas 4, 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