Senior Software Engineer

Pocket raises $11M in bet on rising demand for AI note-taking devices. 6:16 AM PDT · June 29, 2026 Unlike AI gadgets like Rabbit or Humane, companies building dedicated gadgets for recording and transcribing meetings have actually seen some traction. The market is huge - though a bit constrained, as smartphones work fine for such tasks when paired with note-taking apps - and startups like Plaud, Mobvoi, Anker, Viaim, and Vibe have jumped in to take advantage. In this crowded market, Y Combinator-backed Pocket thinks it can win with its design, packaging, and pricing. The company sells a $129 credit card-shaped puck, which sticks to the back of your phone, and promises unlimited recordings, transcriptions, and to-do items, no subscription required. The startup says it has sold more than 130,000 units since launching last year, and that momentum has now helped it score $11 million in funding from Accel, Y Combinator, and ElevenLabs CEO and co-founder Mati Staniszewski. Pocket's core idea isn't novel: You stick the puck to the back of your phone, turn on recording during a meeting, and it'll record and transcribe your conversations. Users can then ask the accompanying phone app to generate summaries of meetings, ask an AI assistant questions about meetings, create mind maps, and transform the text into different templates. While the basic transcription comes free with the puck, the company sells a $200-per-year plan to unlock unlimited AI summaries, queries to the AI assistant, daily highlights, and file attachments. "You can record on the go, offline, and in the field, which is exactly how lawyers, salespeople, doctors, real estate agents, construction workers, and students use Pocket today," Accel partner Cecilia Wang said. "Not only are people present rather than shifting focus to take notes, but more information and insights also get captured than ever before, which otherwise would have been lost. Over time, that accumulation of insights is really valuable: one central place where your ideas, conversations, and thoughts live, rather than scattered and lost," Wang said. Lightspeed is building its edge on followers, not just funds| Equity Podcast 0 seconds of 30 minutes, 16 seconds Volume 0% Pocket was founded by Akshay Narisetti, who was a founding member of rival note-taking startup Omi; and Gabriel Dymowski, who previously founded a blockchain-based document management startup. "We thought every meeting notetaker was built for online conversations, but nothing was geared towards real-life talk. AI really needs a lot of context to work better for us, and a lot of that context exists offline," Narisetti told TechCrunch. For its enterprise customers, Pocket offers custom workflow management, webhook support, and integration with apps such as Google Calendar, OneDrive, Google Drive, Obsidian, Claude, and Cursor. Plus, there is a model context protocol (MCP) server to connect its AI assistant to other databases. Like other meeting notetakers, Pocket wants to help people automate tasks like drafting emails, updating CRMs, and creating action items based on meetings. The company is banking on shipping software fast to enable these integrations. Devices like Pocket undoubtedly face competition from software players such as Granola, Zoom, Fireflies, Otter, and Read AI. However, device-first companies like Plaud, which is on track to generate annual revenue of $100 million via software sales, are also building enterprise capacity along with desktop apps for digital meetings. When you purchase through links in our articles, we may earn a small commission. This doesn't affect our editorial independence. Ivan Mehta Ivan covers global consumer tech developments at TechCrunch. He is based out of India and has previously worked at publications including Huffington Post and The Next Web. You can contact or verify outreach from Ivan by emailing [email protected] or via encrypted message at ivan.42 on Signal. October 13 - 15 San Francisco Scale faster. Grow your portfolio. Gain practical expertise. No matter your goal, Disrupt can empower you. Save up to $330 today!

In the fight with big techs, Read AI prepares office in Brazil after revenue jumps 291%. With 90% of its Brazilian base concentrated in SMEs, startup accelerates seven-figure investment to attract the public sector and large companies, said global CEO David Shim to Bloomberg Línea June 16, 2026 | 03:39 PM Bloomberg Línea Brasil - The accelerated advancement of generative artificial intelligence tools after the turn of 2022 brought with it a side effect known to ten out of ten executives: information overload and task overload. In the productivity ecosystem, the race of technology startups is focused on moving from being a mere data repository to becoming an integrated action platform. It is on this thesis that American Read AI, a platform that positions itself as a digital executive assistant by monitoring meetings, emails, and work messages, is based. Founded at the beginning of the pandemic by David Shim - a serial entrepreneur who sold his previous startup, Placed, to Snapchat for $135 million in 2017, later acquired by Foursquare - the company has the Brazilian market at the top of its global priorities. "Brazil is currently a 'top 4' market for us in the world," global CEO David Shim told Bloomberg Línea in an interview during Web Summit. "It was an initial surprise, but now we are investing heavily, in the seven-figure dollar range." The numbers are a response to the platform's growth in the country. Last year, Read AI surpassed the mark of 1 million registered Brazilian users on the platform. In the most recent balance sheet, the country recorded a 200% jump in users and a 291% expansion in revenue year-over-year. Additionally, the local market boasts the highest conversion rate from free to paid subscriptions in the entire global operation. To support the pace, the startup is preparing to open its first physical office in the country in the next 12 months, likely in São Paulo. "It was very important for us to commit to the market by introducing Brazilian Portuguese. It was a big step to come in and say 'we accept pix.' There's a whole setup to get that running, to be the pioneer in doing so. And I think the next step is to structure a team here and ensure that people feel they are receiving localized support in this market," he said. According to Shim, the investment also meets a growing demand from large companies and public entities for a local support team. Today, Read AI operates with 90% of its customer base concentrated in small and medium-sized enterprises (SMEs) and 10% in large corporations - a mirror of the go-to-market model it adopted in the US before closing large-scale enterprise license contracts. "In the US, we were also 90% SMEs and 10% corporate. When we started to advance, these larger companies began to say: 'Look, we have 40 people using it, we need to get an enterprise license' and implement Read AI more and more broadly. We expect something very similar to happen in Brazil," he stated. 'Reaction factor' In a market saturated with meeting transcription tools - where competitors from Zoom and Microsoft Copilot to language models like OpenAI and Gemini offer text summaries - Read AI tries to redesign the defense strategy through what it calls analytical multimodality. Of the company's processing costs, which total tens of millions of dollars annually, 90% run on proprietary models and only 10% depend on third-party LLMs like Anthropic and OpenAI. That 90% is dedicated to the "narration layer." "Anyone can do a transcription or a text summary today," says Shim. "What big techs don't do is capture the reaction to words. Our models analyze whether you nodded positively, looked away out of boredom, or accelerated your speech rate from 150 to 200 words per minute because you were excited. This reaction layer is our differentiator." In Brazil, the startup calibrated its models to differentiate local Portuguese from European Portuguese and to create a behavioral baseline. "Brazilians are historically more expressive and gesture more in video calls. The model needs to understand that this is the cultural norm, not necessarily that the person is in an atypical peak of euphoria. We normalized this data," explains the founder. From minutes to 'Corporate Tinder' The flight plan for Read AI envisions transitioning from a recording system (minutes and summaries) to an action system. Recently, the company globally launched Ada, a "digital twin" focused on agentic productivity. By connecting email, management platforms like JIRA, Notion, Confluence, and CRMs like Salesforce and HubSpot, the tool starts predicting the professional's next steps. "If you receive an email chasing a project, the AI cross-references data from your latest meetings, analyzes your delivery volume, and suggests a response saying the text will be out in two weeks, without you having to open a chat and ask anything," says Shim. The interface logic, according to him, mixes TikTok's recommendation algorithm with Tinder's mechanics. "The AI will suggest actions based on your history, and you'll swipe right or left, approving or discarding. If you approve, it automatically updates the final system." Asked about the risk of agents fully automating interactions to the point of emptying the human role, Shim dismisses the scenario in the short term. "The human in the loop will be essential for at least the next five years. Someone needs to be the judge to assess whether the agents' suggestions make sense," he states.