Follow-up · Oct 15 callThe talent intelligence engine

Who can make the intro, and how they know them.

The engine already ranks leads by warmth. What it can't see yet is who actually knows whom. Since the call we ran Superhuman's 2,397 people in our index against one open req and one new hire: 654 of 906 sales leaders who fit Head of SMB and Commercial Sales have a path to someone at Superhuman, and 104 of the people who engage with Sierra's posts have no employer or school in common with Sierra. Below is how that gets into the engine, with nothing for employees to install.

A real candidate and real overlaps · live Crustdata pull · Oct 19, 2026 · candidate name withheld · your engine sets the final order
2,397
Superhuman people in our index, across seven company records
654 / 906
US commercial and mid-market sales leaders with a path to someone at Superhuman
104
people who engage with Sierra's posts and have no employer or school in common with Sierra
166
people who started a Superhuman role since July, each one a network session
What I heard

The engine works. The missing piece is knowing who actually knows whom.

It's giving them connections on paper.
If you look up someone you met at a conference or did an intensive with or went to high school with, that's not going to surface.
We would prefer to not have to ask them to do anything. The least barrier to entry possible.
How many actions they would have to take versus can RecOps manage it.
We want to have one-on-one conversations with employees coming in and working their networks with them.
What it has to do

Four things the answer has to get right for Superhuman.

The engine stays yours: the Superhuman Docs form, the logic in Superhuman Go, the Slack read. Whatever we add has to fit under it.

01

Go past connections on paper

Same company is a weak signal at a 4,000-person employer. The engine needs the years, the team and the size of the company, plus the ties no résumé shows: the conference, the intensive, the cross-functional project.

02

Zero lift for employees

No CSV exports, no browser extension, nothing to install. Employees across the company will use the engine, so anything that asks them to act first breaks it.

03

Fresh without a cadence

A quarterly "please re-export" is stale the day after. The candidate side has to be read live at lookup, and the employee side has to update as people join and leave.

04

RecOps runs it

Recruiters and sourcers keep the same form, the same Slack read and the same Noon workflow. The setup, the roster and the session outputs sit with RecOps.

The connections question

Three layers of warm. We bring two. Kenny's sessions bring the third.

Nobody outside a platform can hand you a clean first-degree graph without employees acting. So build warmth in layers, each with its own evidence, and let the engine's logic weigh them the way it already does.

Layer 1 · Crustdataevery lookup

On paper, done right

Every overlap between the candidate and Superhuman's people, with the dates, the team on each side and the size of the company. Schools with class years. Weighted so a shared mega-employer counts light.

13,457jobs at 7,045 other companies across Superhuman's 2,397 people, plus schools with class years
Layer 2 · Crustdataevery lookup

In public

Repeat reactions, comments and tags on posts: a public record of who actually talks to whom. It catches the conference friend and the community peer that overlap logic never will.

104 of 207repeat engagers on Sierra's posts have no employer or school in common with Sierra
Layer 3 · the sessiononce per new hire

Confirmed

The new hire walks into Kenny's 30 minutes with a map already built. They say yes, I know them, and yes, I'd make the intro. RecOps records it. That's true first-degree, with consent.

166people started a Superhuman role since July
Layers 1 + 2 + 3→your warmth logic in Superhuman Go→the Slack read, with the reason on every lead
Straight answer

We don't read anyone's connection list. Crustdata doesn't log in as employees, drop cookies, or ask anyone to install an extension or export a CSV, so there's no connection graph in our data. Layers one and two come from public career history and public posts. When the engine needs a confirmed first-degree tie, layer three is how it gets one, inside a session Superhuman already wants to run.

The plays

Five things we'd build under the engine.

Ordered by how fast Elyssa could run them. Play 01 is the missing layer in every lookup today.

01

Every lookup comes back with dated, same-team evidence.

The missing layerRun it this week

The form already collects a name and a profile URL. Crustdata reads that person live and matches every past job and school against Superhuman's 2,397 people in our index: who overlapped, where, which months, in which seats, at what size of company. A shared badge at a 100,000-person employer isn't a back channel, so those count light unless the two sat in the same function. At the reference stage the same rows show who sat in the same org at the same time, and in what seat.

How it runs
  1. The engine sends the profile URL to person_enrich. Live, so last week's job change is there.
  2. Match against Superhuman's roster, kept current with person_search across all seven company records.
  3. Return evidence rows (company, months together, seats, company size) to the engine's warmth logic and the Slack read.

Today's overlap says "both worked at Box". This says "same sales org, July 2012 to January 2017, 54 months side by side".

Destination Superhuman GoCadence every lookupTrigger form submit
One evidence row, as the engine gets itOct 19, 2026
At SuperhumanSolutions Engineer
Shared companyBox · about 4,400 people today
TogetherJul 2012 → Jan 2017 · 54 months
Their seatsLead Sales Engineer, while the candidate ran renewals and then enterprise inside sales
Same orgSales
Also overlapped22 more Superhuman people, at Box
→ your warmth logic decides the rank
02

Kenny's network sessions start from a map, not a blank page.

Closes the first-degree gapRan it on Sierra

166 people started a Superhuman role since July. Before each 30-minute session, Crustdata builds the new hire's map: who they worked beside, who joined while they were hiring, who engages with them in public, cut to the roles Superhuman has open. We ran it on Sierra, who joined in August. 6,829 people overlapped with Sierra at Box or Miro. 439 sat on the same recruiting and people teams, and 725 joined Miro in business roles while Sierra led business recruiting there. 1,376 of the strongest ties now work in a function Superhuman is hiring for, including 191 account executives and 135 sales leaders. The session turns that list into confirmed ties.

How it runs
  1. On the start date, person_enrich the new hire, then pull everyone who overlapped at each past employer with person_search: dates, team, hiring window.
  2. Add the people who engage with their posts (social_post_list_live), then cut to titles that match Superhuman's open reqs (job_search).
  3. The new hire confirms in the session. RecOps writes the confirmed ties into the engine as the top warmth tier.

The employee does one thing: the 30 minutes Kenny already wants with them. A list to react to beats a blank page and a memory test.

Destination RecOps + the sessionCadence once per startTrigger start date
Sierra's session mapOct 19, 2026
6,829

overlapped with Sierra at Box (2014 to 2019) or Miro (2019 to 2023)

1,376

strong ties now in a function Superhuman is hiring for

Eng & data480
People & recruiting209
Account execs191
Sales leaders135
Customer success135
Marketing128
Product & design50
Finance & legal48
Strong = same team, joined during Sierra's hiring years, or two or more years side by side. Names come back in the pull, for Sierra and RecOps only.
03

Catch the ties that never show up on paper.

Beyond overlap

Repeat reactions, comments and tags are a public record of who actually talks to whom. We read Sierra's last 23 posts: 207 people engaged more than once or left a comment. 104 of them have no employer and no school in common with Sierra, mostly recruiting and people leaders at other companies, plus founders and executives. Overlap logic can't find a single one. In the engine, the same check runs on the candidate: whose posts they engage with, and who at Superhuman engages with theirs.

How it runs
  1. social_post_list_live on the candidate's recent posts, with reactors and commenters.
  2. Resolve those people with person_enrich and match them to Superhuman's roster.
  3. Flag repeat engagers and commenters as a warmth signal next to the overlap rows.

It only sees people who post or engage in public, and recent posts at that. It's proof of interaction, not a confirmed connection, so it should sit under a layer-three tie in the engine's weighting.

Destination Superhuman GoCadence every lookupTrigger form submit
Who engages with Sierra's posts · 207 peopleOct 19, 2026
Worked with Sierra at Box, Miro or since84
Share a school19
No shared employer or school104
Recruiting, talent and people leaders elsewhere51
Founders and executives20
Sales and GTM8
Everyone else25
Engaged with two or more of Sierra's posts, or commented. Names withheld.
04

Open reqs, sorted warm-first.

Head of SMB and Commercial SalesSide by side with Noon

Superhuman posted Head of SMB and Commercial Sales on September 19. We pulled 906 US SMB, commercial and mid-market sales leaders at software and internet companies. 654 have a path to someone at Superhuman. 162 have a strong one: a year or more side by side, in the same function or at a company small enough that it means something. 146 different Superhuman people are a strong path to at least one of them. The most common strong paths run through Samsara, Yelp and Datadog.

How it runs
  1. person_search on the req's titles and segments for the pool (semantic mode works when the req reads loose).
  2. Score every candidate's paths against Superhuman's roster, weighted by company size and function.
  3. Hand the list to Noon or Ashby in warm-first order, with the path on every row.

Noon finds the people. This tells the sourcer who at Superhuman can open the door, before the first message goes out.

Destination Noon · AshbyCadence per open reqTrigger req opened
Strongest path per candidate · a sample of the 162Oct 19, 2026
Candidate todayPath at SuperhumanTogether
SamsaraDirector of Commercial SalesVP of Revenue Operationsboth at Samsara2021 → 2025
DocusignVP, Commercial SalesSenior Talent Partner, GTMrecruited commercial sales there2022 → 2024
SeismicVP, Commercial SalesManager, GTM Systemsboth at Payscale2016 → 2018
Augment CodeHead of Commercial SalesCommercial Sales Executiveboth at Datadog2020 → 2022
CyeraHead of Sales, NA CommercialCommercial Account Executiveboth at Okta2022 → 2024
HubSpotSenior Director of Sales, Mid MarketAlliances Partner Managerboth at HubSpot2021 → 2023
ProcoreSr. Director, Enterprise & CommercialSVP, Salesboth at Zendesk2022 → 2025
PagerDutyDirector, Commercial SalesGTM Recruitersales recruiter there2019 → 2021
Companies and titles only. Names come back in the pull. Some of these companies may be Superhuman customers.
05

Every team that joins brings its network.

Only Superhuman can run this one

Superhuman sits across seven company records in our index: Superhuman, Grammarly, Superhuman Mail, Coda, Rows, GPTZero and Fathom. Each team that joins adds its whole career history to the back-channel graph. Fathom's 142 people alone bring 875 past employers, and on their own they're a path to 268 of the 906 sales leaders above, 24 of them strong. A CSV process would need every one of them to export, then export again. Here the roster refreshes from public profiles, and a watch flags it when someone with a strong path changes jobs.

How it runs
  1. person_search across all seven company records each week: add starters, drop leavers.
  2. Re-score open reqs and saved candidates against the new roster.
  3. watch on strong-path candidates: a job change or a new title sends the engine a webhook.

Five teams joined in under two years. Each one should make every lookup warmer the week it happens, without anyone asking for a spreadsheet.

Destination RecOpsCadence weekly + on changeTrigger roster change
The roster, by company recordOct 19, 2026
RecordPeoplePast employers
Superhuman1,2104,060
Grammarly7361,652
Fathom142875
GPTZero112437
Superhuman Mail97511
Coda92205
Rows824
2,397 people in our index, each counted once under their main current record. Past-employer counts overlap across records.
Where it plugs in

Same form, same Slack read. More reasons on every lead.

The engine already calls a few MCPs. Ours plugs in the same way, or through the API. Nothing changes for recruiters, sourcers or employees.

Yours · unchanged

Superhuman Docs form

Name and profile URL in. Nothing new to fill out.

Yours · unchanged

Superhuman Go engine

Your warmth logic stays yours. It gets more evidence to weigh.

Yours · unchanged

Ashby

Interview history, how far they got. Still the first check.

Crustdata · API or MCP

Profile, paths, interaction

One source for the profile pull the engine gets from Clay and Apify today, plus the overlap and post evidence.

Yours · kept current

Employee database

Synced from the roster across seven records, plus confirmed ties from the sessions.

Yours · unchanged

Slack read · Noon

The same notification, with the reason on each lead. Noon keeps sourcing.

How to test us

One open req, side by side with Noon.

Pick the req. Run it in Noon and through us in the same week. Score the outputs, not the pitch.

Ask 01Count

How many people fit the req?

Ours for Head of SMB and Commercial Sales today: 906 US commercial and mid-market sales leaders at software and internet companies, from 4,273 title matches.

Ask 02Warm

How many come with a named path into Superhuman?

Ours: 654 with a path, 162 strong, each with the company, the months and the seats on both sides.

Ask 03Truth

Does the evidence match what your team knows?

Hand us 10 people your team already knows well, from Ashby or your own careers. Check our paths against reality. This is the cell we'd accept losing.

Ask 04Fresh

How many show today's job?

Take the top 25 on each list and check current titles and start dates against their public profiles. Every row of ours carries its start date.

Noon stays where your sourcers work. The test is about the data under the engine: whether the warm path is real, and whether the profile is current.

Straight answers

The questions from the call.

01Can you get each employee's real connection list?

No, and nothing in our data pretends to be one. We don't log in as anyone, drop cookies, or read connection lists. The two ways to get true first-degree are an employee export or an opt-in browser extension, and both ask employees to act.

What we can do is layers one and two above, which get much of the way without either, and the session map in play 02, which turns Kenny's sessions into confirmed ties with consent.

02What would employees, recruiters and RecOps each have to do?

Employees: nothing outside the session they'd already have. Recruiters and sourcers: the same form, the same Slack read, the same Noon. RecOps: owns the roster sync, the session maps and writing confirmed ties into the engine.

03Is it real-time?

The candidate is read live at lookup, so a job change from last week shows up. The Superhuman roster refreshes on a schedule you set (weekly is plenty), and a watch can flag job changes on the people who matter most.

04Does it go past professional profiles?

Yes. One record pulls together professional profiles, public posts, GitHub, Google Scholar, news and the open web. For back channels, the profile history and the posts carry most of the weight.

05Do we have to rebuild the engine?

No. The engine already calls MCPs, and ours fits the same slot, or Elyssa can call the API directly. Crustdata returns evidence rows; your logic keeps doing the ranking.

What happens next

Let's wire it into one lookup.

Everything below came out of the call. No dates we didn't agree to.

The short version

The engine already ranks by warmth. Now every lead comes with the reason.

Dated overlaps, public interaction and confirmed ties from Kenny's sessions, in the tools Superhuman already uses.