EN
Back to the archive

The archive · Hardware & Devices · Product decision · 2024–2026

Pocket bets on AI recorder: 10,000 units, $27M ARR, Accel leads $11M

Viral concept video brought preorders; launched 2025-10 with 10,000 units; by 2026-03, 35,000 units, $27M ARR; raised $11M led by Accel.

Open Vision Engineering · Pocket

The betBet people buy dedicated hardware for meeting notes instead of apps; if novelty only, 35,000 units is ceiling.Scaling

What the business is

Pocket builds AI hardware: a dedicated conversation recording device that automatically turns meetings, calls, and in-person conversations into speaker-labeled summaries, drafted follow-up emails, and trackable to-dos; it records only when the user explicitly enables it, uses end-to-end encryption, complies with HIPAA and SOC 2, and does not train models on customer data.

Starting capitalStarted with preorders from a concept video; completed an $11M round led by Accel on 2026-06-29 (follow-on investments included YC, Vercel CEO Guillermo Rauch, ElevenLabs co-founder Mati Staniszewski, Opendoor CEO Kaz Nejatian, and others).

How it started

At the end of 2024, Open Vision Engineering released a concept video: a small device that could record conversations and automatically generate notes. The video quickly went viral, and thousands placed preorders before the product existed. Founders Gabriel Dymowski and Akshay Narisetti officially launched Pocket in 2025-10, shipping more than 10,000 units on the first day.

What happened

By 2026-03, Pocket had cumulatively shipped more than 35,000 units and reached $27M in annualized revenue, with month-over-month growth exceeding 50% in some months, and became the most-watched launch of the Y Combinator Winter 2026 batch; enterprise customers such as DoorDash began using it. On 2026-06-29, it announced the completion of an $11M round led by Accel, with follow-on investments from Y Combinator and multiple operator investors; the funds were used for expanded design/engineering hiring and exploration of new hardware form factors.

How it ended up

Still running: the AI hardware sector previously had multiple high-profile failures, and Pocket's $27M annualized revenue is currently a rare example of real revenue, but the company has not disclosed data on whether 35,000 units can scale to 350,000 units or whether retention works.

Background

Pocket builds AI hardware: a dedicated conversation recording device that automatically turns meetings, customer calls, and one-on-one conversations into speaker-labeled summaries, drafted follow-up emails, and trackable to-dos; it records only when the user explicitly enables it, uses end-to-end encryption, complies with HIPAA and SOC 2, and promises not to train models on customer data.

The story began with a concept video at the end of 2024: a small device that could record conversations and automatically generate notes. The video went viral, and thousands placed preorders before the product had even been built; founders Gabriel Dymowski and Akshay Narisetti officially launched Pocket in 2025-10; it shipped more than 10,000 units on the first day and became the most-watched launch of the Y Combinator Winter 2026 batch.

By 2026-03, Pocket had shipped more than 35,000 units and reached $27M in annualized revenue, with month-over-month growth exceeding 50% in some months; enterprise customers such as DoorDash began using it. On 2026-06-29, the company announced the completion of an $11M round led by Accel, with follow-on investments from Y Combinator, the Vercel CEO, an ElevenLabs co-founder, the Opendoor CEO, and others; the funds were used for design and engineering hiring expansion and exploration of new hardware form factors.

The industry backdrop is multiple high-profile failures in the AI hardware sector: Pocket's $27M annualized revenue became one of the few figures with real revenue, while software competitors Otter.ai, Fireflies.ai, and the wearable Limitless were also competing for the same demand. The company has not disclosed retention/repeat purchase data, and whether 35,000 units can scale remains unknown.

What has to be true

  • Concept video validated demand at low cost, bringing preorders before product existed.
  • Dedicated device with HIPAA/SOC 2 and no training on data passed legal review for enterprises.
  • 35,000 units and $27M ARR showed real revenue amid AI hardware funding bubble.
  • YC W26 most-watched launch and Accel round turned viral moment into scalable company.

What can be applied

Validate demand with concept video, use YC distribution, and design for compliance; but revenue doesn't prove category—35,000 units must become 350,000.

Aftermath

After $11M round on 2026-06-29, team hired and explored new form factors; product still centers on recorder; 35,000 units and $27M ARR remain latest figures; DoorDash is customer; competitors include Otter.ai, Fireflies.ai, Limitless; retention data unpublished.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases