Pocket: Revenue, team and how it was built
Pocket (AI note-taking device that attaches to the back of a phone and records conversations and calls, USA): US$100M a year run rate as reported by Business Insider, August 2026. Evidence grade D: reported, not verified. Team: 2 co-founders, Gabriel Dymowski and Akshay Narisetti, who started it in 2024; 13 full-time staff in July 2026 (XYZ), 12 employees in August 2026 (Hakky Handbook, citing Business Insider) and 35 employees on the Y Combinator page in November 2026. Launched October 2025; still running, revenue growing.Added 6 November 2026
- Outcome
- US$100M a year run rate as reported by Business Insider, August 2026Revenue growing
- Evidence
- D reported, not verified
- Launch
- October 2025
- Status
- Still running
- Team
- 2 co-founders, Gabriel Dymowski and Akshay Narisetti, who started it in 2024; 13 full-time staff in July 2026 (XYZ), 12 employees in August 2026 (Hakky Handbook, citing Business Insider) and 35 employees on the Y Combinator page in November 2026
- Outside funding
- US$11M seed round led by Accel with Y Combinator, Peak XV Partners and angel investors, as reported by XYZ in July 2026; US$11.5M in total according to the Lean AI Leaderboard
- Group
- Voice and audio
- Main channel
- Not stated
What is not verified
The headline US$100 million run rate in August 2026 rests on one source: a Business Insider article that this pipeline could not open, because businessinsider.com disallows AI agents. It is read here through Hakky Handbook's summary of 19 August 2026 and the Lean AI Leaderboard, which both cite that article, so they are not independent. The earlier US$27 million run rate is one statement by co-founder Gabriel Dymowski to XYZ. A run rate is a projection from recent months, not a year of revenue, and no payment-processor profile, investor statement or filing confirms any figure. A reader could look for the revenue in a later funding announcement or an investor's own statement. The headline number is grade D: one claim or one estimate, with no second source. Aiolum shows it because the story is useful, and marks it so that it is not mistaken for a verified figure.
Pocket is an AI note-taker, a product that records speech, turns it into text with a speech-recognition model, and then asks a large language model (an AI model trained on very much text that can write summaries and answer questions) to write notes and task lists. Pocket's version is a hardware device with two microphones and a contact microphone, a microphone that picks up vibrations of the phone body, so it can record a phone call without the loudspeaker. It has no screen and stores audio offline until the phone app syncs it.
Gabriel Dymowski, a Polish founder who earlier built DoxyChain, and Akshay Narisetti met in 2024 at The Residency, a hacker house in San Francisco, and worked on Pocket full-time from 2025. A concept video in late 2024 brought thousands of pre-orders, and the co-founder told XYZ that more than 10,000 devices shipped on launch day in October 2025. He told XYZ that by the end of March 2026 the company had delivered over 35,000 devices with an annualized revenue run rate of US$27 million, and the company raised US$11 million led by Accel. Business Insider reported in August 2026, as summarized by Hakky Handbook, that Pocket had reached a US$100 million annual run rate with 12 employees and over 200,000 devices sold.
The headline is grade D: a company statement in one press article that this pipeline could not open (Business Insider blocks AI agents), read through a Japanese news summary and a leaderboard that both cite that article. A run rate is recent revenue multiplied up to a year; it is a projection, not a year of sales. The team was two co-founders at the start and is 35 people in November 2026, so it is shown here as a team that grew after funding.
How much revenue does Pocket make?
Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.
| Fact | Value | Grade | How we know | Sources |
|---|---|---|---|---|
| Annual run rate, August 2026 | US$100 million | D | Business Insider, August 2026, as summarized by Hakky Handbook on 19 August 2026 ("announced that it has reached a $100 million annual run rate just eight months after launch"). The Lean AI Leaderboard lists the same figure with snapshot time August 2026 and cites the same article. The article itself could not be read. | [1, 2] |
| Run rate per month, August 2026 | about US$8.3 million | calc | Calculated: US$100 million divided by 12. | [1] |
| Annualized run rate, end of March 2026 | US$27 million | D | Co-founder Gabriel Dymowski to XYZ, published 5 July 2026: "That is the projected annual revenue based on the first months of results." | [3] |
| Growth implied by the two run rates | about 30% a month from March to August 2026 | calc | Calculated: (100 / 27) to the power 1/5, minus 1. The company told XYZ and Business Insider it grows about 50% a month; 50% a month from US$27 million for five months would be about US$205 million, so the two claims do not fit together exactly. | [1, 3] |
| Devices | more than 10,000 shipped on launch day (October 2025); over 35,000 delivered by end of March 2026; over 200,000 sold by August 2026 | D | Company statements to XYZ (first two) and to Business Insider as summarized by Hakky Handbook (last). | [3, 1] |
| Share of revenue from devices | over 90% | D | Gabriel Dymowski to XYZ, July 2026: "the vast majority of revenue – over 90% - comes from devices". | [3] |
| Largest market | United States, 80 to 90% of sales | D | Co-founder to XYZ. Main channel is heypocket.com; also an Amazon US seller from spring 2026. | [3] |
| Founding team | 2 co-founders | C | XYZ and the Y Combinator page name Gabriel Dymowski and Akshay Narisetti as founders; the company was founded in 2024 according to Y Combinator. | [3, 4] |
| Team size | 13 full-time (July 2026); 12 employees (August 2026); 35 employees (November 2026) | C | XYZ; Hakky Handbook citing Business Insider and the Lean AI Leaderboard; the Y Combinator page, which also lists 16 open roles. FundedIQ still shows the 1-10 employee band. | [3, 1, 2, 4, 6] |
| Funding | US$11 million seed led by Accel | D | XYZ, July 2026: with Y Combinator, Peak XV Partners and angels including Mati Staniszewski, Guillermo Rauch and Kaz Nejatian. Hakky Handbook dates the round to June 2026. The Lean AI Leaderboard lists US$11.5 million in total. | [3, 1, 2] |
| Accelerator | Y Combinator, Winter 2026 batch | C | Y Combinator company page (batch W26) and XYZ. | [4, 3] |
| AI models | GPT-5, Claude and others; user can choose ChatGPT, Claude or Gemini for summaries | C | Home page: "Powered by GPT-5, Claude, and more. Pocket is model-agnostic". Co-founder to XYZ on choosing the model. | [5, 3] |
| Hardware | 2 microphones and 1 contact microphone, 64 GB storage, 4-day battery, 52 g, MagSafe | C | Specifications on the home page; recording range up to 15 m. | [5] |
| Time from launch to the August 2026 figure | 10 months | calc | Calculated: October 2025 to August 2026. Business Insider's headline, as summarized, says eight months; the launch date it counts from is not given. | [3, 1] |
How to read this table: a grade D value is a company statement reported once, or a directory figure, with no independent check. A grade C value comes from the company's own page or a register-like page such as Y Combinator's. A "calc" value was computed on this page from the rows above it. No value is grade A or B.
| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Pocket device | 129 one-time | shown as reduced from US$199; free plan included | |
| Free | 0 | 0 | unlimited transcription at standard accuracy, summaries kept 30 days, 3 Ask Pocket chats a day |
| Pro | 19.99 | billed yearly at US$239.88; summaries kept forever, best transcription, speaker names, unlimited chats | |
| Enterprise | custom | admin controls, API keys, HIPAA mode and BAA |
The device is a one-time purchase shown at US$129 with a crossed-out US$199, with free shipping. The home page says the first 30 days of the subscription are free from delivery and that after that US$19.99 a month keeps the advanced features; the plan table shows Pro at US$19.99 a month billed yearly (US$239.88 a year, "Save ~33%") and does not print a separate monthly price. Hakky Handbook's summary of Business Insider gave the premium plan as from US$16.59 a month in August 2026. Enterprise prices are by quote.
When did Pocket launch, and how did it grow?
From a concept video in late 2024 to a reported US$100 million run rate ten months after launch. Each entry carries its grade.
- 2024CGabriel Dymowski and Akshay Narisetti meet at The Residency, a hacker house in San Francisco. Y Combinator gives 2024 as the founding year. [3, 4]
- Late 2024DA concept video brings thousands of pre-orders, according to the company. The founders first built a meeting-minutes app, then moved to hardware. [3, 1]
- 2025CBoth founders work on Pocket full-time. [3]
- October 2025DLaunch. The company says more than 10,000 devices shipped to users on launch day. [3]
- Winter 2026CPocket joins Y Combinator's W26 batch. [4, 3]
- End of March 2026DThe co-founder told XYZ: over 35,000 devices delivered and a US$27 million annualized run rate. [3]
- June to July 2026DUS$11 million seed round led by Accel, reported by XYZ on 5 July 2026; 13 full-time staff. [3, 1]
- August 2026DBusiness Insider reports, per Hakky Handbook, a US$100 million annual run rate, 12 employees and over 200,000 devices sold. [1, 2]
- November 2026CThe Y Combinator page lists 35 employees and 16 open roles. [4]
Why did Pocket work?
Four causes, in the order they mattered. Each one is a reusable lesson.
A dedicated device for a job phones do badly
The co-founder told XYZ why customers buy it instead of a phone app: it records without network coverage, an incoming call does not stop the recording, and its own battery does not drain the phone. The contact microphone records both sides of a phone call without the loudspeaker.
Hardware first, subscription later
Over 90% of revenue came from devices in mid-2026, according to the co-founder. The free plan keeps the device useful without paying; the Pro plan sells better transcription, unlimited chats and longer storage. In his words, there is no subscription without a product in the customer's hands.
Many models, no single provider
Pocket passes transcripts to models from several providers, and the user can choose ChatGPT, Claude or Gemini. The device and the app are Pocket's own; the language intelligence is rented.
From one buyer to a company order
The co-founder described the path: a manager buys a Pocket, likes it, promotes it inside the company, and the company places a corporate order. He named DoorDash as such a customer.
How it was built
These details come from the co-founder's interview with XYZ and from the company's home page. They are grade C or D. [3, 5]
- Software is built in San Francisco; the devices are manufactured in Asia, and the founders split their time between the two.
- The device records offline to 64 GB of storage and syncs to the app over Bluetooth or Wi-Fi.
- Summaries use large language models from several providers; the company says customer data is not used to train models and claims HIPAA and SOC 2 compliance.
- The shop runs on Shopify; the company also sells on Amazon in the US.
- The funding pays for production runs and inventory, which the co-founder said hardware needs before it can sell at scale.
Tools named in the sources
Each one opens its page in the Tools tab, with every story that uses it.
Who tried this before Pocket, and why did they fail?
Four companies in our database of failed startups sold a dedicated gadget or wearable that had to beat the phone or bigger rivals. The values below are from that database and are not independently verified.
Humane
The promiseA US$699 screenless AI pin to replace the smartphone.
Why it diedCustomers did not want to give up the phone, and the device did little that a phone could not do.
Jawbone
The promiseBluetooth headsets and fitness trackers.
Why it diedFitbit and Apple outcompeted it while product recalls and poor service cost it customers' trust.
Wimm Labs
The promiseA smartwatch with small apps, so users need not take out the phone.
Why it diedThe market was not ready, and better-funded rivals took the early buyers.
Woolet
The promiseA leather wallet with Bluetooth tracking, made by a Polish hardware startup.
Why it diedHigh manufacturing cost per unit against Tile and later Apple AirTags at lower prices.
The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.
What changed
Humane asked people to replace the phone; Pocket attaches to the phone and does one task, recording, that the phone does badly during calls and long meetings. Jawbone and Wimm Labs lost to large companies that added the same function to their own devices, and that risk is open for Pocket too: meeting software and phone makers already offer AI notes.
What differs is that the expensive part, the speech and language models, is now rented from providers by the minute, so a two-person start could ship useful AI on a cheap device. Like Woolet, Pocket carries the cost of manufacturing and inventory, which is why the co-founder told XYZ it raised money to pay for production.
Can one person copy Pocket now?
Not alone. The idea of a one-job device on top of rented AI, maybe.
A solo founder today would need a manufacturer, certification, inventory money and customer support for physical returns, which Pocket paid for with an US$11 million round according to XYZ. The open part is software: a single-purpose recorder or capture tool whose value is the AI notes, sold first as pre-orders after a video.
Where it is still open: Field workers, clinicians and salespeople who need notes of in-person work, and languages or regions where the large note-taking products are weak.
- 1Find a task the phone does badly during real work: calls without speaker, offline recording, long battery life.
- 2Show the product in a concept video and take pre-orders before paying for a production run.
- 3Keep a free plan so the device is useful alone, and sell the AI extras as a subscription.
- 4Do not depend on one model provider: let the user choose.
What could go wrong for Pocket?
The risks are the mirror image of the causes above.
What is not verified
The headline US$100 million run rate in August 2026 rests on one source: a Business Insider article that this pipeline could not open, because businessinsider.com disallows AI agents. It is read here through Hakky Handbook's summary of 19 August 2026 and the Lean AI Leaderboard, which both cite that article, so they are not independent. The earlier US$27 million run rate is one statement by co-founder Gabriel Dymowski to XYZ. A run rate is a projection from recent months, not a year of revenue, and no payment-processor profile, investor statement or filing confirms any figure. A reader could look for the revenue in a later funding announcement or an investor's own statement.
Team size and growth
The company was started by two people in 2024. XYZ counted 13 full-time staff in July 2026 and Business Insider, as summarized, 12 employees in August 2026. The Y Combinator page lists 35 employees and 16 open roles in November 2026, after the US$11 million round. FundedIQ still shows the 1-10 band, which is out of date.
Recording other people
The device records people in a room up to 15 metres away and both sides of a phone call. The home page shows a testimonial that says "Phone call transcription (with consent)". Hakky Handbook's summary says AI note-takers face criticism for recording without explicit consent, and that Pocket answers that a button press and a device on the table are more visible than recording with a phone. Recording laws differ by place, and the user is responsible for consent.
Trust signals on the home page
The home page shows "400,000+ Creators, Founders, and Thinkers", twice the more than 200,000 devices reported in August 2026; it may count users rather than devices, but no source says. A "Trusted by professionals at:" logo row has no source. Many reviews carry a first name, initial and city (for example "Lucas W. - Boston, MA") and dates within a few days in October 2025.
Competition
On this site, Wave AI, Risenote and Voibe sell AI meeting notes as software; Pocket is the only one that sells its own device. Hakky Handbook's summary lists Google Meet, Zoom, Microsoft Teams, Granola and a planned OpenAI device as competition.
Dependence on model providers
The home page says it is "Powered by GPT-5, Claude, and more", and the co-founder told XYZ that users choose between ChatGPT, Claude and Gemini. Several providers lower the risk from any one of them, but the AI part of the product is rented.
Hardware economics
Over 90% of revenue came from devices in mid-2026, according to the co-founder. Device sales must be earned again every month, and production must be paid before sales, which is why the round went to inventory.
Sources
The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.
- D Hakky Handbook, 19 August 2026: Pocket, the smartphone-attachable AI notetaker, sees rapid growth, hitting $100 million in ARRRead in full, downloaded 6 November 2026. A news summary that links Business Insider as its source. Source for the US$100 million run rate, 12 employees, over 200,000 devices, the premium price, the consent debate and the competitors.
- D Lean AI Leaderboard, published sheet (snapshot August 2026)Read the CSV of the published Google Sheet that the page loads, downloaded by the harvester on 5 November 2026. Lists US$100 million revenue, 12 employees, US$11.5 million funding, founded 2025, and cites the Business Insider article.
- D XYZ (Poland Unpacked), 5 July 2026: Polish-founded AI device Pocket raises USD 11m as it scales in the USRead in full, downloaded 6 November 2026. Interview with co-founder Gabriel Dymowski. Source for the founding story, the US$27 million run rate, devices, revenue split, funding, team of 13 and model choice.
- C Y Combinator company page: Pocket, Take Notes in the Real WorldRead the page data embedded in the HTML, downloaded 6 November 2026. Source for the founders, founding year 2024, batch W26, 35 employees and 16 open roles.
- C Pocket home page, read 6 November 2026Read in full. Source for the product, specifications, prices, plans, model providers, the consent wording, the user counter and the reviews.
- D FundedIQ: Pocket (heypocket.com) fundingRead in full, downloaded 6 November 2026. A directory; source only for the 1-10 employee band and the founder listing.
- C Database of failed startups, local exportLocal export. AI-written summaries, not independently verified. Source for the failed predecessors.
Run this company? Send corrections or a verified figure to aiolum@livyden.com