--- title: Sabi raises $50M seed for a brain-reading cap that turns... description: Sabi raised a $50M seed round at a reported $600M valuation to build the Sabi Cap, a non-invasive wearable that reads brain signals and converts them into text... canonical: https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Sabi raises $50M seed for a brain-reading cap that turns thoughts into text | daily.dev og:description: Sabi raised a $50M seed round at a reported $600M valuation to build the Sabi Cap, a non-invasive wearable that reads brain signals and converts them into text... og:url: https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz og:image: https://api.daily.dev/og/posts/kCd66Pidz.png og:image:alt: Sabi raises $50M seed for a brain-reading cap that turns thoughts into text og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Sabi raises $50M seed for a brain-reading cap that turns thoughts into text **[Collections](https://daily.dev/sources/collections)** · 3 min read · 0 upvotes · 0 comments ## Summary Sabi raised a $50M seed round at a reported $600M valuation to build the Sabi Cap, a non-invasive wearable that reads brain signals and converts them into text prompts for AI models. The round was led by Khosla Ventures, with Accel, DST Global, Initialized Capital, Collab Fund, and former OpenAI product chief Kevin Weil participating. The cap can hold up to 100,000 sensors feeding a brain foundation model trained on 100,000 hours of labeled neural recordings collected by Sabi, and the company claims it can predict a user's next 3-4 keystrokes before typing. Sabi designs its own chips, including a non-contact EEG chip fabricated by TSMC, which it calls the first chip purpose-built for consumer brain-computer interfaces. Reactions were mixed: one researcher welcomed the hardware investment given past BCI limitations, while another was skeptical of the gap between the modest keystroke-prediction claim and the bold pitch of thought-based agent control. ## Content Sabi has come out of stealth with a $50 million seed round to build a wearable that turns brain signals into text and AI prompts. The reported valuation is $600 million. ## The money Khosla Ventures led the round, with Vinod Khosla named as the lead. Accel, DST Global, Initialized Capital and Collaborative Fund (Collab Fund) also took part, along with Ascend and angels including Kevin Weil, the former OpenAI chief product officer. ## The hardware The Sabi Cap is a non-invasive wearable, built into headwear like baseball caps and beanies. No surgery is involved. Its sensors are chip-scale EEG units, 1 to 5 millimeters across, woven into the lining. Reports put the count at somewhere between 70,000 and 100,000. They are non-contact, meaning they pick up electrical activity through hair across a gap of up to 5 millimeters rather than touching the scalp. Sabi designed its own EEG chip and neuroimaging sensors. TSMC fabricated the chip, which Sabi describes as the first purpose-built for consumer brain-computer interfaces. Sabi's internal benchmarks claim very low power use compared with medical-grade equipment, which is what would let the array sit in ordinary fabric without much heat or electrical noise. ## The model The signals go to what Sabi calls a Brain Foundation Model, trained on more than 100,000 hours of labeled non-invasive neural recordings. Sabi collected the data itself from volunteers doing mental tasks and imagined speech. The idea is to learn patterns shared across many brains, so users don't need a long recalibration session every day. That recalibration has been a big reason non-invasive interfaces stayed in labs. Sabi says the cap can predict your next three or four keystrokes from brain signals before you type them. The company projects initial speeds of about 30 words per minute, across three modes: imagined speech to text, thought-driven prompts for AI assistants, and predictive commands. It plans to ship early commercial products. ## My take Thirty words per minute from thought alone is a claim I'd want to see demonstrated outside a company video. Still, the reaction from people who know the field is telling. @omarsar0 did BCI research during his PhD and said the technology wasn't great then. He called the investment in better hardware a big deal, since current AI could unlock interesting research and personal agent applications. @Hesamation's joke was that brain foundation models would soon be reading his thoughts. Both reactions fit: it's exciting and a little unnerving at once. Whether the sensors can really read imagined speech through hair is the open question, and a $600 million valuation on a seed round makes the answer matter. ## Questions this post answers ### How much funding did Sabi raise for its brain-reading cap and who led the round? Sabi raised a $50M seed round at a reported $600M valuation, led by Khosla Ventures. Other participants included Accel, DST Global, Initialized Capital, Collab Fund, and former OpenAI product chief Kevin Weil. The company is building the Sabi Cap, a non-invasive wearable that reads brain signals and turns them into text prompts for AI models. _Readers tracking AI hardware funding rounds follow startup milestones like this one on daily.dev._ ### What does Sabi's brain-computer interface cap actually do? The Sabi Cap can hold up to 100,000 sensors, each 1 to 5 millimeters, feeding a brain foundation model trained on 100,000 hours of labeled neural recordings collected by Sabi itself. The company claims the system can predict a user's next 3-4 keystrokes from brain signals before typing, with the longer-term pitch being direct thought-based control of AI agents. _Developers evaluating emerging AI hardware interfaces follow breakdowns like this on daily.dev._ ### What makes Sabi's EEG chip different from typical brain-computer interface hardware? Sabi's EEG chip is non-contact, meaning it reads brain signals without touching the scalp, and was fabricated by TSMC. Sabi describes it as the first chip purpose-built specifically for consumer brain-computer interfaces, pairing custom silicon with a company-owned dataset of neural recordings, a hardware investment that past BCI efforts have generally lacked. _Those weighing custom silicon bets in emerging hardware categories track stories like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 4 discussions and 24 comments across x (as of 2026-10-14). **TL;DR:** Reaction is largely cautious curiosity: people are intrigued by owning chips and a huge proprietary dataset but repeatedly question whether accuracy, noise, and session-to-session drift can live up to the pitch, while others raise privacy and consent worries about thoughts becoming training data. **Sentiment:** 20% positive · 45% mixed · 35% skeptical **The case for** - Owning the custom chip and 100,000 hours of labeled recordings is seen as a defensible moat nobody can easily replicate. - The modest next-keystroke prediction framing is viewed as an honest, achievable first milestone rather than overhyped mind-reading. - Some find the idea of 'thinking at' an agent appealingly faster than typing. **The pushback** - Many doubt the signal-to-noise and spatial resolution of 100,000 scalp sensors can reliably decode intent versus noise. - Comparisons to Meta's Brain2Qwerty highlight high character error rates, casting doubt on the leap to 30wpm inner-speech decoding. - Session-to-session drift from hair, sweat, and cap fit is flagged as an unsolved calibration problem. - Privacy/consent concerns about intrusive thoughts or private rehearsed sentences being captured and fed to an LLM. - Skepticism about where the 100,000 hours of training data actually came from and who consented to it. **By community** - x (mixed): Replies swing between excitement about chip/data ownership and sharp skepticism about accuracy, noise, drift, and privacy implications. **Hottest debate:** Whether the touted sensor count and keystroke-prediction demo represent real decoding progress or just an easier motor-signal trick dressed up as thought-reading. **Open questions** - What is the actual error rate or independent benchmark for the claimed 30 words-per-minute inner-speech decoding? - How is spatial resolution achieved with 100,000 non-contact sensors, and what signal type are they really measuring? - How was the 100,000-hour labeled neural dataset collected, and under what consent terms? - How will the system handle drift across sessions without daily recalibration? - What safeguards prevent unintended or private thoughts from being transmitted to an agent? **Highlights** > @rohanpaul_ai The keystroke demo is the easier half. Meta's Brain2Qwerty decoded typing from motor signals: 67% character error on EEG, 32% on MEG, 35 volunteers. Predicting keys you're about to press rides that motor prep. Sabi's 30 wpm target is inner speech, with no keypress to lock onto. > — [neuralloot on x](https://x.com/neuralloot/status/2108575264284234204) > @omarsar0 100k sensors is the headline. the hard part is drift: eeg shifts with hair, sweat, how the cap sits. a brain model that works across sessions without recalibrating every morning is the real milestone > — [i\_Am\_Snow\_Flake on x · 1 points](https://x.com/i_Am_Snow_Flake/status/2108584173124415900) > @testingcatalog Owning the chip and the 100,000 hours of recordings is the part I'd bet on, since nobody can buy that data off a shelf. The 30 words per minute claim still needs an independent measurement before it's more than a demo. > — [DoDataThings on x](https://x.com/DoDataThings/status/2108585720638267800) > @testingcatalog Predicting keystrokes from brain signals? That's not autocomplete, that's reading over my shoulder inside my skull. If it goes wrong, prosecute the deployer — responsibility ties back to a person. Who consented to their thoughts being training data? > — [NWAPCHELP on x](https://x.com/NWAPCHELP/status/2108582597215596947) > @omarsar0 Wearable BCI demos will need published evidence on task accuracy, calibration time, false activations, and performance across users. The sensor count alone doesn’t tell us how reliably intent can be decoded. > — [CKPillaiAI on x](https://x.com/CKPillaiAI/status/2108572734003282028) **Source threads** - [x](https://x.com/testingcatalog/status/2108569887585022109) · 1 points · 8 comments - [x](https://x.com/Hesamation/status/2108561804645323102) · 1 points · 4 comments - [x](https://x.com/rohanpaul_ai/status/2108569695586754920) · 0 points · 5 comments - [x](https://x.com/omarsar0/status/2108565653682536887) · 0 points · 7 comments ## Similar posts on daily.dev - [BCI startup Neurable looks to license its ‘mind-reading’ tech for consumer wearables](https://daily.dev/posts/bci-startup-neurable-looks-to-license-its-mind-reading-tech-for-consumer-wearables-9mtg5t3oa) · TechCrunch · 0 upvotes · 0 comments - [China's BrainCo bets brain tech is a headband, not surgery](https://daily.dev/posts/china-s-brainco-bets-brain-tech-is-a-headband-not-surgery-meawb8wdj) · The Next Web · 0 upvotes · 0 comments - [Science Corp raises $230M as it races to bring its brain implant to market](https://daily.dev/posts/science-corp-raises-230m-as-it-races-to-bring-its-brain-implant-to-market-hrlpnlopk) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#brain-computer-interface](https://daily.dev/tags/brain-computer-interface) [View this post on daily.dev](https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Sabi raises $50M seed for a brain-reading cap that turns thoughts into text","url":"https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz"},"datePublished":"2026-10-09T14:47:16.320Z","dateModified":"2026-10-14T16:15:43.093Z","description":"Sabi raised a $50M seed round at a reported $600M valuation to build the Sabi Cap, a non-invasive wearable that reads brain signals and converts them into text...","image":"https://pbs.twimg.com/media/HUMk5i3XYAABKJa.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HUMk5i3XYAABKJa.jpg","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,brain-computer-interface","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Sabi raises $50M seed for a brain-reading cap that turns thoughts into text"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/sabi-raises-50m-seed-for-a-brain-reading-cap-that-turns-thoughts-into-text-kcd66pidz#faq","mainEntity":[{"@type":"Question","name":"How much funding did Sabi raise for its brain-reading cap and who led the round?","acceptedAnswer":{"@type":"Answer","text":"Sabi raised a $50M seed round at a reported $600M valuation, led by Khosla Ventures. Other participants included Accel, DST Global, Initialized Capital, Collab Fund, and former OpenAI product chief Kevin Weil. The company is building the Sabi Cap, a non-invasive wearable that reads brain signals and turns them into text prompts for AI models. Readers tracking AI hardware funding rounds follow startup milestones like this one on daily.dev."}},{"@type":"Question","name":"What does Sabi's brain-computer interface cap actually do?","acceptedAnswer":{"@type":"Answer","text":"The Sabi Cap can hold up to 100,000 sensors, each 1 to 5 millimeters, feeding a brain foundation model trained on 100,000 hours of labeled neural recordings collected by Sabi itself. The company claims the system can predict a user's next 3-4 keystrokes from brain signals before typing, with the longer-term pitch being direct thought-based control of AI agents. Developers evaluating emerging AI hardware interfaces follow breakdowns like this on daily.dev."}},{"@type":"Question","name":"What makes Sabi's EEG chip different from typical brain-computer interface hardware?","acceptedAnswer":{"@type":"Answer","text":"Sabi's EEG chip is non-contact, meaning it reads brain signals without touching the scalp, and was fabricated by TSMC. Sabi describes it as the first chip purpose-built specifically for consumer brain-computer interfaces, pairing custom silicon with a company-owned dataset of neural recordings, a hardware investment that past BCI efforts have generally lacked. Those weighing custom silicon bets in emerging hardware categories track stories like this on daily.dev."}}]} ```