Follow-up · Nov 4 callTalent and hiring intelligence
Know the market before you scale to 4,000.
ElevenLabs goes from about 1,000 people to between 2,500 and 4,000 next year, and the sourcing data sits in tools that don't talk to each other. Since the call we mapped four crafts across 19 countries on your job board, traced the 907 people who joined you since last November, and measured talent density the way your job posts define it.
Forward deployed engineers: about 1 per posting in Japan
248 on public profiles, 220 FDE postings in 60 days. The US: 7,500+ for 4,597.
HotspotAdd to plan
Full-stack in Bulgaria: 70+ per posting
12,000+ engineers, 171 postings in 60 days. Bulgaria is on 6 of your full-stack listings.
Hires since Nov 2025
907
Talent competitors
25
Left them since Sep 6
1,600+
Olympiad ex-founders
743
Profiles per FDE posting, by hublast 60 days
Japan1.1
United States1.6+
Ireland1.8
Germany1.8
United Kingdom3.1+
Ask: which talent competitors lost FDEs this month?⌘K
Illustrative layout of a console you'd build · every number is from a live Crustdata pull · Nov 5, 2026
907
people joined ElevenLabs since last November, each traced to where they came from
1.1
forward deployed engineers in Japan for every FDE posting in the last 60 days, and 1.6+ in the US
1,600+
people left your 25 talent competitors since Sept 6
1 in 5
of your technical people on public profiles are ex-founders, top two of 22 AI and software companies (with Lovable)
What I heard
Connect the sourcing data, then use it to calibrate the business.
Data is really, really fragmented.
As we scale, there's going to be lots of crossover.
How are we hunting for white whales, as we call them, or talent density?
Where are the hotspots that the team can hire from?
Data will help us to calibrate, to get us there to know what we're looking for.
What it has to do
Four things the data layer has to get right for ElevenLabs.
A remote-first company with no job titles, hiring in 19 countries, can't map its market by title and city the way most talent tools do.
01
One person, one key, in every tool
Ashby, Juicebox, Stitch and the recruiter seats each hold their own copy of a candidate. Every copy needs the same person ID and profile URL, so "already in Ashby?" has an answer before anyone reaches out.
02
Map by craft and hub, not by title
Your job posts say "We don't have job titles," and 105 of the 907 people who joined since last November list theirs as just "GTM." The map has to run on career history and skills, across every country you'd hire in.
03
Talent density as a number
Your bar is already written down: "IOI medalists and ex-founders." Signals like those can be counted at any company, so density stops being a feeling and becomes an index the business can watch.
04
Learn from the funnel, then watch the market
Once Ashby outcomes show which backgrounds do well here, the same profile becomes an alert on the outside market, by craft. That's how the business gets calibrated.
What we'd build
Five things we'd build for ElevenLabs.
Ordered by how fast Nic's team could run them. Play 01 needs no integration and is ready this week. We show companies, titles and counts here; names come back in the pull.
01
The market map: supply, demand and movers for every craft and hub
The askRun it this weekSupply and demand
"Global team: we prioritize your talent, not your location" means the map is the sourcing plan. For four crafts you're hiring right now, we counted the people in the seat in each of the 19 countries on your job board, how many started their seat in the last year, and how many postings the market put up for the same craft in the last 60 days. Forward deployed engineers are the tightest market on it, and full-stack in Bulgaria, already on 6 of your listings, is one of the deepest.
Sort
Country
Profiles per posting
In seat
New in seat, 12 mo
Postings, 60 days
In seat: people on public profiles in that country with the craft's title today, local-language titles included (ML research: research titles at software, internet and IT companies). A "+" means at least that many; a range means two checks disagree and the truth sits between them. New in seat: started their current position in the last 12 months. Postings: job postings for the same craft and country since Sept 6, every employer, agencies included, and a posting can repeat across cities. "Your roles": listings on your public job board on Nov 5 (144 roles) that name the country. Pulled Nov 5, 2026.
How it runs
person_search_live counts each craft per country by current title and time in seat, in real time.
job_search counts postings for the same craft in the same country, across every employer, over a rolling 60 days.
A weekly job writes the table to a sheet and #talent-intel, and flags any hub whose ratio swings, so recruiters source where it's deep.
FDE runs at 1.6+ profiles per posting in the US and 1.1 in Japan. You won't source your way out of a market that thin, so the FDE card in "Beyond the five" builds them from the crafts your deployment hires already came from.
→ Sheet + #talent-intelWeeklyAny hub's ratio moves 20%+
02
Talent competitors, defined by where your hires actually come from
Competitive intelligenceAlready run
You're still defining who your talent competitors are. Your hires have already voted. We read all 907 people who joined ElevenLabs since last November, profile by profile. The biggest previous employers: Salesforce (23), Google (22), Palantir (15), Meta (13) and AWS (10). Split by craft, the lists barely overlap, which is the point: each craft has its own talent competitors.
How it runs
person_search_live pulls everyone at ElevenLabs with under a year at the company, and person_enrich_live re-reads anyone it misses, with full career history.
Each hire is grouped by craft from title and history (titles here are thin), and the employer right before ElevenLabs is counted.
Rerun monthly, plus the reverse (where leavers go), and the talent competitor list keeps itself current.
Talent competitors stop being a debate in a meeting. They're the companies your hires come from, by craft, refreshed every month.
Every profile read on Nov 5; 10 people who had already left ElevenLabs are out. Crafts are read from titles, so people who list no title sit only in "Everyone." Ambassadors and freelancers are left out of the crafts.
03
Market movers and exec moves at your talent competitors, every week
Signal · job changeExec insights
Take 25 talent competitors: the biggest employers from play 02, plus the AI labs and voice AI companies you compete with for people. More than 1,600 people left one of them since September 6, led by Google, Salesforce, AWS and Meta, and about a third already list their next employer. The moves below show where some of them went: OpenAI, Anthropic, CoreWeave and Sierra.
How it runs
A watch holds the talent competitor list, scoped to your crafts and seniority, and checks it on a schedule.
Each run returns who left, where they went, the new title and the start month, with full history behind each one.
Moves post to #talent-intel in Slack, tagged by craft, and exec moves go to a weekly digest for leadership.
A move is the one window when someone listens. The watch shows it that week, not at the next quarterly review.
→ Slack + weekly digestWeeklyLeft a talent competitor
Moves since Sept 6 · a sampleNov 5, 2026
Left
Joined
When
SalesforceDirector, Product Management for Voice and AI
AnthropicMember of Technical Staff
Oct
StripeHead of Marketing, Japan
OpenAIHead of Enterprise Marketing, Japan
Oct
GoogleHead of Partners, Public Sector
AnthropicGlobal Head, Public Sector Partner Growth
Oct
GoogleVP Engineering, Distributed Cloud
CoreWeaveSVP, Engineering
Oct
MetaSenior Machine Learning Engineer · UK
Isomorphic LabsSenior Research Engineer
Oct
MistralGTM, Enterprise · UK
OpenAIGTM, Startups
Oct
GoogleTechnical Account Manager · UK
SierraForward Deployed Infrastructure Engineer
Nov
Companies and titles only; names come back in the pull. Counts are people whose role at one of the 25 ended since Sept 6; some are between jobs.
04
White whales: IOI medalists and ex-founders, wherever they are
Only ElevenLabs can run this oneTalent density
Your job posts set the bar in your own words: "IOI medalists and ex-founders." A search can read that. More than 10,000 people in engineering or science seats list a STEM olympiad honor on public profiles (informatics, maths, physics and the like), and 743 of them have also founded something. 283 of those 743 have been in their current seat for two years or more. The most common employers today: Google (38), Meta (24), Microsoft (18), OpenAI (13) and Amazon (10).
How it runs
person_search on honors.title and past founder titles, scoped to your crafts and the countries you hire in.
Publications (research_profiles) and open source (dev_platform_profiles) add more signals; each person gets a density score and a reason.
A watch keeps the shortlist current. When one changes jobs or passes two years in seat, the craft's recruiter hears first.
Every company says it hires the best. ElevenLabs wrote its bar down in a way a search can read.
→ Ashby + #talent-intelMonthly, plus movesTwo or more signals
The 743, by country · outside ElevenLabsNov 5, 2026
Country
White whales
In seat 2+ years
United States
284
102
United Kingdom
62
20
India
46
18
Brazil
32
11
Canada
30
10
Singapore
26
11
Germany
13
3
Netherlands
13
6
Switzerland
12
7
Poland
9
3
Ireland
5
1
Bulgaria
2
0
White whale: an engineering or science seat today (engineer, scientist, technical staff, developer or CTO), a STEM olympiad honor (informatics, maths, physics, chemistry, biology, astronomy or computing) and a past founder title. Each of the 743 checked profile by profile on Nov 5. Other countries hold the rest.
05
Ashby, enriched and joined: one key for every candidate, kept current
Connects the toolsRediscoveryTagging
About 1 in 10 full-stack engineers changed seats in the last year: 12,000+ of 130,000+ in the UK and 5,000+ of 58,000+ in Poland started their current position in the last 12 months. That's how fast an Ashby record goes stale. Re-enriching gives every candidate their current employer, title and history, plus the same person ID and profile URL your sourcing tools work from, so a duplicate shows up before the reach-out, not after.
How it runs
Your script reads candidates from Ashby's API; person_identify matches each email or profile URL to one person.
person_enrich adds current employer, title, history, skills and honors; your script writes them back as Ashby fields and tags.
A watch re-checks the pool. A move, promotion or new seat at a talent competitor posts to the recruiter who owns the candidate.
The tagging strategy comes from fields, not from recruiters remembering to tag. And it's the base for funnel learning: which backgrounds pass which stage here.
→ Ashby fields + SlackNightlyCandidate moved
Ashbycandidates
→
Crustdataidentify · enrich · watch
→
Ashbyfields + tags
Slackmoved alerts
craftseniorityemployertalent competitorolympiadex-foundermoved this year
What one enrich returnsevery field is filterable
Who
Person ID, profile URL, current title, location
Career
Every employer and title with start and end dates, seniority, function
Signals
Honors, education, skills, publications, open-source repos and stars
Freshness
Refreshed every 30 days; live endpoints when it has to be current today
The open question · what does talent density mean
Talent density, measured the way your job posts define it.
We counted four public signals among the technical people at 22 companies: an olympiad honor, a past founder title, three or more published papers, and 100+ stars on open-source work. ElevenLabs and Lovable lead all 22 on ex-founder share, at about 1 in 5. 9 of your technical people list an olympiad honor. Click a column to re-rank.
Share of technical profiles with each signalNov 5, 2026
Company
Ex-founder ↓
Olympiad
3+ papers
100+ stars
Share of each company's technical people on public profiles (current title includes engineer, research, scientist, technical staff or developer). Ex-founder shows a range where two ways of counting the team give different answers. Many ElevenLabs people list no title or just "GTM", so small teams move more with each person. A missing signal isn't proof it isn't there; it means it isn't listed. Olympiad from listed honors, papers from Google Scholar, stars from GitHub. Pulled Nov 5, 2026.
The mix differs
Density isn't one thing
Google DeepMind's comes from publications (41% have 3+ papers). Hugging Face's from open source (17.5%). Yours and Lovable's from founders. Pick the signals that predict success here, not the ones that sound good.
Calibrate it
Weight it with your own funnel
Play 05 puts these signals on every Ashby candidate. Join them to stage outcomes and you learn which ones pass here, by craft. Those become the weights.
Run it monthly
One index, three uses
The same score ranks talent competitors (where density sits), the outside market (play 04's white whales) and your own pipeline (who to fast-track).
Where it plugs in
Build the console in-house. Don't build the index.
ElevenLabs builds its own tools, and this should be one of them. The part that doesn't pay to build is the crawl underneath: 1B+ people and 60M+ companies matched across sources, refreshed every 30 days, with live endpoints when a record has to be current to the minute. Your agent or app calls it through the API, the MCP or the CLI. Or we build it with you.
Yours · ATS
Ashby
Stays the system of record. We read through its API; your script writes fields and tags back.
Yours · sourcing
Juicebox, Stitch, recruiter seats
Keep sourcing. Each record carries the same person ID, so overlap shows up.
Yours · in trial
Noon
Logs the reach-outs across tools. We supply the person underneath them.
New · data layer
Crustdata
People, companies, jobs and watches through the API and MCP.
Yours · HRIS
HiBob
Who performs and the job architecture, joined to the same person records.
Yours · alerts
Slack
#talent-intel for movers, hotspots and white whales.
What your team builds
The console, the agent and who sees what
The density score and its weights
The Ashby sync and the tagging rules
Which signals route to which recruiter
What Crustdata supplies
1B+ people and 60M+ companies, from our own index, not resold data
Full career history with dates, honors, education, skills, publications, open source
Job postings across employers, for the demand side of the map
Watches that push changes to a webhook or Slack
Straight answers
The questions from the call.
Q1What data do you hold? Anything on interview processes?
Public data on people, companies and job postings, built as our own index (no data co-ops, no resold data), refreshed every 30 days, with live endpoints for anything that must be current. Past the profile: GitHub, Google Scholar, arXiv and the open web, matched to the same person.
What happens inside another company's interview loop isn't public, so we don't hold it. What is public, we can read in one pass: a single web search on Stripe's engineering onsite returns 8 public write-ups and guides naming its rounds (programming, a bug squash in an open-source codebase, and an integration round).
Q2Could we pull salary when it's written in different places in Ashby?
Yes, as a build with you. We've indexed ATS data for other recruiting teams: take the raw text from candidate records and notes, agree the structure with you (salary expectation, notice period, "reach out in six months"), parse it into fields, and make it searchable next to the external data.
Q3How does the API work?
A REST API, a CLI, and an MCP server for Claude or any agent. For recruiting teams we also run a hosted instance that keeps searches going overnight and remembers the team's calibration, so nobody re-enters the same inputs. Watches push changes to a webhook or Slack. Here's play 04's search, trimmed to fit:
POST /person/search
{ "filters": { "op": "and", "conditions": [
{ "field": "honors.title", "type": "[.]", "value": "olympiad" },
{ "field": "experience.employment_details.past.title",
"type": "[.]", "value": "founder" },
{ "field": "experience.employment_details.current.title",
"type": "[.]", "value": "engineer" } ] },
"limit": 100 }
// plus research, scientist and technical staff titles; returns total_count,// then each profile's career, honors, education and skills
Q4Can it connect Ashby, Juicebox, Stitch and the recruiter seats?
It gives them a shared key. We match every candidate to one person ID and profile URL and keep that record current, so any tool can check Ashby before outreach. We don't log into recruiter seats or read their messages; logging reach-outs is what the Noon trial covers. Whatever gaps the trial leaves are where we'd focus.
Q5What would you start with, for where we are today?
The market map and talent competitors first (plays 01 and 02): no integration, useful the week you turn them on, and they inform the interview redesign and the sourcing plan. Then talent density and white whales. Then Ashby enrichment and funnel learning, once the first calibrated profiles exist.
Q6Can it help with the inbound volume?
Yes. Enrich each applicant on arrival with history, tenure, employers, publications and open source, then route on what the public record shows rather than only on the resume. Join it to Ashby's source field and you can compare referrals with applications, and see which posting brings which backgrounds. "Beyond the five" has the details.
Beyond the five
Four more, once the first five are running.
Inbound
40,000 applications a month, sorted by the public record
Enrich every applicant, check the resume against their career history, and score on your density signals before a recruiter opens it. Compare referral and applicant quality by source and posting, so A/B testing the advert has an outcome to measure.
What your talent competitors publish about their loops
Engineering blogs, careers pages and candidate write-ups, read in one pass for each company on play 02's list. One search on Stripe returns 8 public write-ups and guides describing its programming, bug squash and integration rounds.
web_search_live · web_enrich_live
FDE, from adjacent crafts
Build forward deployed engineers instead of finding them
Your deployment hires came from Palantir and McKinsey. 1,500+ people who held a forward deployed title at Palantir now work somewhere else, and 173 Palantir alumni hold engineering seats in the UK. Add consultants and solutions engineers at AI companies and the FDE map gets much deeper.
For a company without titles, crafts and levels can come from history and skills. Cluster your own people and your talent competitors the same way, and HiBob gets a job architecture you can benchmark.
Everything below came out of the call. No dates we didn't agree to.
DougDone
This page, and the pulls behind it
The market map for four crafts in 19 countries, where your last 907 hires came from, density across 22 companies, and 743 white whales, each checked profile by profile.
DougNext
A shared Slack channel
One place for questions, feedback and anything the team wants pulled.
ElevenLabsNext week
The Noon trial starts
Tell us what it covers and what it doesn't. The gaps are where we'll focus.
NicWhen ready
Share it, then pick the first two plays
Your read on priorities decides what we build first.
The short version
Know where the talent is, who's moving, and who clears your bar, before you scale to 4,000.
Every number on this page came from a live pull this week. The same pulls can run every week, for every craft you hire.