The GLMMR crystal GLMMR

Social reasoning data for AI labs.

Teaching AI when a person could change the outcome of a message,
and how to bring them in.

AI doesnโ€™t just write our messages.
Now it acts for us.

22 Apr

OpenAI

workspace agents

19 May

Google

Gemini Spark

16 Jun

Microsoft

Copilot Cowork

11 Aug

xAI

Grok Bot

8 Sep

Meta

Muse

14 Sep

Apple

Siri AI

29 Sep

OpenAI

dots

Routine messages, an agent can send. The ones that can change a relationship need a person, and the hard part is knowing which, and where in the message. How you bring them in decides what you get: a quick answer, or their judgement.

AI has the context.
The writer has the read.

The meeting notes say the CFO pushed back on price. A model reads that as a price objection. The writer, with the same notes, reads a CFO who was never going to say yes in front of his own team.

the context:
emails, meeting notes, history

the model

its read

their read

so they decide how hard to push, what to leave out, how much to say

We tested it on 9 Oct: records written by an open model didnโ€™t lift its own scores on two social-reasoning tests. The read has to come from a person.

Labs already pay billions a year
for expert reasoning.

CompanyWhat it suppliesPublic signal, as reported
Scale AILabelled and expert dataValued over $29bn after Meta’s investment, June 2025
MercorExperts hired to make lab data$10bn valuation, Oct 2025; $2bn+ gross run rate
AfterQueryExperts’ step-by-step reasoningFounded 2025; reported $3.2bn valuation, Sept 2026
EthosAI-run expert network, London$22.75m Series A led by a16z, May 2026
GLMMRSocial reasoning from real work, built to record outcomesPre-seed. Records start with the first cohort

A labโ€™s own logs canโ€™t capture it: an edit shows what changed, not why. What a person sees in the other person exists only as they write, inside the product they write with. That product is GLMMR.

GLMMR records social reasoning.

How people read the person they write to.

Relational

How they read the person theyโ€™re writing to, and what the relationship can take.

Tacit, lived experience

What they have seen happen before, never written down.

27 of 37

of the responses in our own early sessions drew on one of these: 15 on lived experience, 12 on the relationship.

Juliana did this by hand for seven years.
Peter built the product on her method.

Juliana Camargo

Juliana Camargo

Co-founder and CEO

Ran her own consultancy, grown by word of mouth: 1,500+ one-to-one sessions drawing out the judgement of founders, CEOs and YPO leaders, and turning it into writing people trust. Moved to the UK from Brazil at 20, after placing top three nationally in a L’Oréal marketing competition.

Peter Jin Hong

Peter Jin Hong

Co-founder and Chief Product Architect

Fourteen years at Google and Alphabet, across Search, Assistant and Trust & Safety, working with behavioural science leaders Dr Dacher Keltner and the late Dr Albert Bandura. Speaks on trust and Human+AI collaboration, including at Harvard’s AI for Human Flourishing think tank.

Edwin Cubillos

Edwin Cubillos
Engineer
(ex-Mercado Libre, Uber, Nokia)

Line Hjartarson

Line Hjartarson
Design Lead / Lovable Ambassador
(former Design Director: Omny, Cognite)

AI research scientist (name withheld)
Staff-level AI research scientist at a large technology company. Trains AI agents to read situations and follow the unwritten social rules of interacting with people. Interviewed and aligned; open to joining when the round closes.

Our product finds where a person could change
the outcome of a message, then brings them in.

Built for the messages at work where the relationship matters.

The product on a real outreach email: three lines marked as needing the writer's judgement

It marks the lines that need the writerโ€™s judgement.

A session on the same email, at the warming up stage of four (warming up, going deeper, the mirror, the line): the product gets the writer to exercise their judgement on why they picked this person

A four-stage method (warming up, going deeper, the mirror, the line) draws out the writerโ€™s judgement.

Every session is graded by hand against Julianaโ€™s method: 200+ since July.

What each record holds: when the person beat AI,
and how they were brought in to do it.

1

Where in the message human judgement is needed

2

How they were brought in

3

Their judgement: how they read the person theyโ€™re writing to (social reasoning)

4

The reply they expected, and the one the AI expected

5

What came back

whose read
was right

Captured inside our product. Only writers who opt in; a lab gets labels, never the message or a name.

A product people want.

16 of 35

people in demos asked to use it on the spot.

340

on the waitlist, with no paid marketing.

Message after a demo: It was great! Thanks for inviting me. I want to shar your tool with our team members!

after a demo

Each record should cost us ~$4 to make.
Expert reasoning suppliers pay experts to make theirs.

What it costs to make the data, at a planning price of $100 a record

Mercor

$60โ€“70

GLMMR

~$4

Mercor, for example, pays its experts $60โ€“70 of every $100 it earns. Ours comes from a product people want to use for their work.

Mercor: Sacra; Dealroom, July 2026. GLMMR: our estimate of model spend, $3.30 to $4.20 a record, from first reads measured on test drafts ($0.19 to $0.41 each); a full session not yet measured.

A public test first, so labs can check
the data before they buy it. Q2 2027.

500 records: the public test

the rest: for sale to labs

It measures how often the person beat the model on the same message. Any lab can run it. AfterQuery started the same way.

Eighteen months
to prove labs will pay.

Late Oct 2026

Launch. Records start.

Q2 2027

Benchmark published. First lab pilot.

Second half 2027

First lab contract, about $3M.

End of 2028

Three labs buying. About 90,000 records a year, from about 1,200 people who opt in.

Dates are targets.

The round pays for the proof:
people, data and compute.

It takes us to the published benchmark and the first lab pilot, spent in the order of the plan.

People

Our AI research scientist first

Owns the dataset, the benchmark and the statistics, and talks to labsโ€™ data teams with Juliana. Then a platform engineer.

Data collection

Writers and consent

Inviting and supporting the writers who opt in, and privacy counsel for consent and for any release of records to labs.

Model compute

Evals and a first model

Running the product for the cohort, a grader trained on hand grades, frontier models tested on our benchmark, and a tuned open model.

We build the product with AI coding agents and have merged 628 changes since July, so the round goes to research and data, not a large product team.

The GLMMR crystal GLMMR

Letโ€™s scale Human+AI brilliance.

Juliana Camargo

Juliana Camargo

Co-founder, CEO

juliana.camargo@glmmr.me

Peter Jin Hong

Peter Jin Hong

Co-founder, Chief Product Architect

The GLMMR crystal GLMMR

Social reasoning data for AI labs.

Teaching AI when a person could change the outcome of a message,
and how to bring them in.

AI doesnโ€™t just write our messages.
Now it acts for us.

  1. 22 AprOpenAI workspace agents
  2. 19 MayGoogle Gemini Spark
  3. 16 JunMicrosoft Copilot Cowork
  4. 11 AugxAI Grok Bot
  5. 8 SepMeta Muse
  6. 14 SepApple Siri AI
  7. 29 SepOpenAI dots

Routine messages, an agent can send. The ones that can change a relationship need a person, and the hard part is knowing which, and where in the message. How you bring them in decides what you get: a quick answer, or their judgement.

AI has the context.
The writer has the read.

The meeting notes say the CFO pushed back on price. A model reads that as a price objection. The writer, with the same notes, reads a CFO who was never going to say yes in front of his own team.

the context:
emails, meeting notes, history

the model

its read

their read

so they decide how hard to push, what to leave out, how much to say

We tested it on 9 Oct: records written by an open model didnโ€™t lift its own scores on two social-reasoning tests. The read has to come from a person.

Labs already pay billions a year
for expert reasoning.

CompanyWhat it supplies
Public signal, as reported
Scale AI

Labelled and expert data

Valued over $29bn after Metaโ€™s investment, June 2025

Mercor

Experts hired to make lab data

$10bn valuation, Oct 2025; $2bn+ gross run rate

AfterQuery

Expertsโ€™ step-by-step reasoning

Founded 2025; reported $3.2bn valuation, Sept 2026

Ethos

AI-run expert network, London

$22.75m Series A led by a16z, May 2026

GLMMR

Social reasoning from real work, built to record outcomes

Pre-seed. Records start with the first cohort

A labโ€™s own logs canโ€™t capture it: an edit shows what changed, not why. What a person sees in the other person exists only as they write, inside the product they write with. That product is GLMMR.

GLMMR records social reasoning.

How people read the person they write to.

Relational

How they read the person theyโ€™re writing to, and what the relationship can take.

Tacit, lived experience

What they have seen happen before, never written down.

27 of 37

of the responses in our own early sessions drew on one of these: 15 on lived experience, 12 on the relationship.

Juliana did this by hand for seven years.
Peter built the product on her method.

Juliana Camargo

Juliana Camargo

Co-founder and CEO

Ran her own consultancy, grown by word of mouth: 1,500+ one-to-one sessions drawing out the judgement of founders, CEOs and YPO leaders, and turning it into writing people trust. Moved to the UK from Brazil at 20, after placing top three nationally in a Lโ€™Orรฉal marketing competition.

Peter Jin Hong

Peter Jin Hong

Co-founder and Chief Product Architect

Fourteen years at Google and Alphabet, across Search, Assistant and Trust & Safety, working with behavioural science leaders Dr Dacher Keltner and the late Dr Albert Bandura. Speaks on trust and Human+AI collaboration, including at Harvardโ€™s AI for Human Flourishing think tank.

Edwin Cubillos

Edwin Cubillos
Engineer (ex-Mercado Libre, Uber, Nokia)

Line Hjartarson

Line Hjartarson
Design Lead / Lovable Ambassador (former Design Director: Omny, Cognite)

AI research scientist (name withheld)
Staff-level AI research scientist at a large technology company. Trains AI agents to read situations and follow the unwritten social rules of interacting with people. Interviewed and aligned; open to joining when the round closes.

Our product finds where a person could change the outcome of a message, then brings them in.

Built for the messages at work where the relationship matters.

The product on a real outreach email: three lines marked as needing the writer's judgement
It marks the lines that need the writerโ€™s judgement.
A session on the same email, at the warming up stage of four (warming up, going deeper, the mirror, the line)
A four-stage method (warming up, going deeper, the mirror, the line) draws out the writerโ€™s judgement.

Every session is graded by hand against Julianaโ€™s method: 200+ since July.

What each record holds: when the person beat AI, and how they were brought in to do it.

1

Where in the message human judgement is needed

2

How they were brought in

3

Their judgement: how they read the person theyโ€™re writing to (social reasoning)

4

The reply they expected, and the one the AI expected

5

What came back

whose read
was right

Captured inside our product. Only writers who opt in; a lab gets labels, never the message or a name.

A product people want.

16 of 35

people in demos asked to use it on the spot.

340

on the waitlist, with no paid marketing.

Message after a demo: It was great! Thanks for inviting me. I want to shar your tool with our team members!

after a demo

Each record should cost us ~$4 to make.
Expert reasoning suppliers pay experts to make theirs.

What it costs to make the data, at a planning price of $100 a record

Mercor$60โ€“70
GLMMR~$4

Mercor, for example, pays its experts $60โ€“70 of every $100 it earns. Ours comes from a product people want to use for their work.

Mercor: Sacra; Dealroom, July 2026. GLMMR: our estimate of model spend, $3.30 to $4.20 a record, from first reads measured on test drafts ($0.19 to $0.41 each); a full session not yet measured.

A public test first, so labs can check the data before they buy it. Q2 2027.

500 records: the public test

the rest: for sale to labs

It measures how often the person beat the model on the same message. Any lab can run it. AfterQuery started the same way.

Eighteen months
to prove labs will pay.

  1. Late Oct 2026

    Launch. Records start.

  2. Q2 2027

    Benchmark published. First lab pilot.

  3. Second half 2027

    First lab contract, about $3M.

  4. End of 2028

    Three labs buying. About 90,000 records a year, from about 1,200 people who opt in.

Dates are targets.

The round pays for the proof:
people, data and compute.

It takes us to the published benchmark and the first lab pilot, spent in the order of the plan.

People

Our AI research scientist first

Owns the dataset, the benchmark and the statistics, and talks to labsโ€™ data teams with Juliana. Then a platform engineer.

Data collection

Writers and consent

Inviting and supporting the writers who opt in, and privacy counsel for consent and for any release of records to labs.

Model compute

Evals and a first model

Running the product for the cohort, a grader trained on hand grades, frontier models tested on our benchmark, and a tuned open model.

We build the product with AI coding agents and have merged 628 changes since July, so the round goes to research and data, not a large product team.

The GLMMR crystal GLMMR

Letโ€™s scale Human+AI brilliance.

Juliana Camargo

Juliana Camargo

Co-founder, CEO

Peter Jin Hong

Peter Jin Hong

Co-founder, Chief Product Architect

juliana.camargo@glmmr.me

โ† โ†’ or tap to move