The Names We Carry: Épée, Gallaudet, Clerc

Three names, three jobs, one mission.

Today, September 23, is the International Day of Sign Languages. The date is Deaf history in itself: on September 23, 1951, in Rome, Deaf associations founded the World Federation of the Deaf. The United Nations adopted the Day by consensus in 2017, and the world has observed it every year since 2018. This year the WFD turns 75, and its theme is "Declaring Deaf People's Human Rights."

I want to dedicate this day to the people who marked our history. Not as a nostalgic gesture. Their names are written into every part of CLERC:

  • our dataset is called Épée,
  • our sign recognition model is called Gallaudet,
  • and the company itself is called CLERC.

Every time we train, test or ship something, we say their names. Here is why.

Why look back when AI moves this fast

We are living through the fastest decade technology has ever known. AI is the biggest revolution of our time, and for the Deaf community it is the biggest opportunity in our history, as I wrote in our manifesto.

In a world this fast, it is tempting to believe that everything starts now. It does not. The people behind these names worked in the 18th and 19th centuries, with paper, ink and sailing ships that could take 52 days to cross the Atlantic. They faced the question we face today: how do you make a language that lives in hands and faces visible, teachable and respected by everyone else?

I have spent nearly a decade on that question, through Keia, Spokhand and now CLERC. Every founder needs role models, even ones born more than 300 years ago. These are mine.

Épée: the dataset

Charles-Michel de l'Épée was born in Versailles in 1712. Around 1760, in Paris, he agreed to teach two deaf twin sisters who had lost their teacher. He turned his own home into a school, and his lessons were free. He wrote that his school was for the poor first.

He did not invent sign language. Deaf Parisians were already signing, and in 1779 the Deaf author Pierre Desloges wrote it plainly: Épée learned the language from Deaf people. What Épée did was make it visible. From 1771 he held public demonstrations, and anyone could come and question his Deaf pupils. He called those pupils "living dictionaries." He began a dictionary so that other teachers could learn signs. It was unfinished when he died in 1789, and it was printed only a century later.

That is exactly what a dataset is. Living dictionaries: Deaf signers whose language is recorded, annotated and made usable. A dictionary that is never finished, only versioned. Épée v0.3 is public on Hugging Face: 1,200 clips from 6 Deaf signers, with a DOI so researchers can cite it.

Gallaudet: the model

Thomas Hopkins Gallaudet was a hearing minister from Hartford, Connecticut. In 1815 he sailed to Britain to learn how deaf children were taught. The Braidwood schools kept their method secret. In London he watched a public demonstration given by the Abbé Sicard, head of the Paris school, with two of its Deaf teachers, Jean Massieu and Laurent Clerc. In Paris, Massieu and Clerc gave him lessons in signs.

Then came the crossing. In 1816, aboard the Mary Augusta, 52 days at sea: Clerc taught Gallaudet signs, and Gallaudet taught Clerc English.

Gallaudet learned the language from Deaf teachers, sign by sign. Our recognition model learns the same way, from real signers, video by video. Today, Gallaudet 1.3 powers our live demo: sign one of 75 ASL signs to your webcam and it reads the sign. Your video is never recorded or uploaded: only anonymous skeleton points are sent to compute the sign, then discarded.

The model carries his name as a tribute to Thomas Hopkins Gallaudet (1787-1851). It is not affiliated with or endorsed by Gallaudet University.

CLERC: the company

Laurent Clerc was born in 1785 in La Balme, a village near Lyon. He was Deaf. At 12 he entered the Paris school created to carry on Épée's work. His first teacher, Jean Massieu, was Deaf too. Clerc became a teacher there himself.

In 1816 he signed a contract to teach in America for three years. He stayed for more than fifty. On April 15, 1817, he and Gallaudet opened the school in Hartford with seven students: the first permanent school for the Deaf in the United States. Clerc became America's first Deaf teacher of Deaf students, and he trained many of the teachers of the schools that followed. American Sign Language grew from the French signing he carried and the signs Deaf Americans already used.

A Deaf Frenchman who crossed the Atlantic to build something that would outlive him. I did not choose this name by accident. I am a Deaf Frenchman building an American company, and I carry his example: when Deaf people lead, the whole field moves forward.

Why pixels, why data

For today's video, we drew their portraits in pixels. Not as decoration. A pixel is the smallest unit of image data, and data is precisely what their generations could not keep. For decades after 1817, early ASL went almost unrecorded in print. The language lived in classrooms and homes, in hands and faces, and most of it disappeared with the people who signed it.

Our generation has the tools they never had.

The Deaf Digital Heritage

This is the vision behind CLERC: the Deaf Digital Heritage. A safeguarded digital record of our languages. Not of one country, but of sign languages across the world. According to the United Nations, citing the WFD, there are more than 70 million deaf people in the world, and together they use more than 300 sign languages. Each one deserves to be preserved, studied and understood by the technology that is now learning the world's spoken and written languages.

The Deaf Digital Heritage is sourced from the Deaf community. It is structurally protected so that it stays in Deaf hands and is never exploited. And it is designed as a gift to the world: the foundation that AI needs to understand sign languages accurately.

Our turn

Épée, Gallaudet and Clerc did their part with what they had. The tools have changed. The mission has not.

Now it is our turn. Not CLERC alone: with the Deaf community, with our partners and with our clients. Every signer who records, every researcher who tests a model, every company that builds on this data adds a page to a dictionary that Épée started nearly 250 years ago and never finished.

Happy International Day of Sign Languages.

Try Gallaudet 1.3 in the live demo at clerc.io/demo, explore Épée on Hugging Face, and follow CLERC to see what comes next.

Portraits in the video: Charles-Michel de l'Épée (engraving, 1833), Thomas Hopkins Gallaudet (photograph, c. 1842), Laurent Clerc (photograph, 1869), all in the public domain. The names Épée, Gallaudet and CLERC are tributes and imply no affiliation with or endorsement by any institution.

MORE FROM CLERC

September 14, 2026

A Four-Second Avatar Is Not a Benchmark. That Is the Point.

SonZo AI just published the first external evaluation of CLERC data, under the 7 Production Gates. The avatar is the best result so far, and the review says it is a pilot, not a benchmark. Here is why that honesty is the whole collaboration.

August 18, 2026

Nobody Talks About the Language

Interpreters will not disappear, and fixing deafness is not the point. The real risk with sign language AI is quieter: a flattened, biased version of the language becoming the standard.

August 11, 2026

The Deaf Digital Heritage

For centuries, sign language lived only inside physical communities. The Deaf Digital Heritage is the structured, permanent corpus that changes that: built by the Deaf community, protected in Deaf hands, opened to the world.

August 3, 2026

Six Months That Beat Five Years

The last six months of CLERC produced more than the five years before them combined. The tools arrived for everyone at the same time. The difference is usage, and for Deaf founders the shift runs deeper than speed.

July 9, 2026

The Sign Language AI Dataset Landscape in 2026

WLASL, How2Sign, ASL Citizen, YouTube-SL-25, GoSign.AI: there have never been more sign language datasets. So why can't AI labs ship? An honest map of the landscape, and where the real gaps are.

July 6, 2026

Four Months, Less Than $1,000, and One Conviction: Data Comes First

CLERC's first public sign language recognition demo reached up to 71% accuracy on an unseen signer after four months, less than $1,000, and a Deaf-led data build.

June 22, 2026

Why Foundation Models Need ASL Training Data

Every major foundation model lab is missing the same thing: a sign language modality. Here is why ASL training data is not optional for the next generation of multimodal AI, and why generic video datasets do not close the gap.

June 1, 2026

Deaf People Use the Future First

Texting, captions, video calls. Deaf communities used them before the rest of the world caught up. Sign language is next, and this time people are not the only ones following.

April 27, 2026

Deaf Community First

Why CLERC exists because of, not for, the Deaf community and why that is a technical requirement, not a values statement.

April 22, 2026

We Are Not an Accessibility Company. We Are a Technology Company.

CLERC builds sign language AI infrastructure: structured, Deaf-led sign language data for ML labs, translation engines, and research teams. We're a tech company, not an accessibility company, and here's why that distinction matters.

April 13, 2026

From Raw Video to Labeled Signs

Raw footage in, production-ready sign language data out. How the CLERC pipeline transforms video into structured, linguistically grounded data.

April 7, 2026

Why Data, Not the Model

Sign Language AI does not have a model problem first. It has a data problem, and that changes everything.

April 1, 2026

THE CLERC DEMO: From Our Database to Your Use Case

Sign language has never had an ImageNet moment. CLERC is building the data foundation that makes sign language AI viable — and the demo shows exactly what that means in practice.

March 29, 2026

Manifesto of CLERC: The Foundation for the Next Generation of Sign Language AI

For nearly a decade, I have been obsessed with one challenge: breaking the communication barrier through technology.