The Work After the Signal

Sabi

Sabi's cap has to recognize the moment a wearer means to communicate. An interface built around imagined speech also needs to leave people free to think without sending an instruction.

That responsibility sits inside the company Rahul Chhabra, CEO, and Atmadeep Banerjee, CTO, co-founded. Sabi announced a $50M Seed round on October 9, led by Khosla Ventures, with Accel, Initialized Capital, Kevin Weil, DST Global, Collaborative Fund and Ascend participating. They are financing a brain-AI interface designed to fit inside an ordinary baseball cap.

The intended experience is quiet: form words internally, send a trained mental command, or confirm a suggested action. Sabi is developing custom EEG electronics and miniature sensors to capture signals through hair without gel or surgery. A phone or computer would receive the result wirelessly.

Banerjee brings a concrete research record. He co-first-authored MindEye, the NeurIPS 2023 work reconstructing viewed images from fMRI recordings. Sabi's cap uses a different sensing approach, and the published paper is evidence of relevant experience rather than a performance certificate for the device. The company says its Brain Foundation Model was trained on more than 100,000 hours of labeled neural recordings.

The commercial pressure goes beyond whether a model can decode a signal. A wearable has to earn the right to stay on someone's head, then make its output useful without demanding constant supervision. An unwanted suggestion and an unwanted action carry different consequences. The interface needs to make that difference understandable to the person wearing it.

Sabi says it is building controls for when recording happens, which applications receive commands and which actions need confirmation. It also describes encryption starting at the sensor and inference on encrypted data. Those are development claims, with independent security and everyday performance still to be established. Permission is already part of the engineering work, before an AI assistant gets a new way to hear from its user.

The funding supports a larger neural dataset, completion of the custom chip, early-device manufacturing and hiring across neuroscience, hardware and machine learning. Sabi plans an early prototype demonstration at CES 2027 and is preparing units for its early-access waitlist. As those caps reach people, the interaction will have to give a wearer a clear way to move from a private intention to an instruction they chose to send.

Oxide Computer Company

Oxide's founders had a tax bill worth telling investors about. The company says selling computers generated taxable income this spring, after components, manufacturing, salaries and the rest of the business were paid for.

Then came the problem a profitable hardware company can still have: customers wanted more machines than it could build. An order arrives long before the finished rack, and somebody has to pay for the parts in between. Growing that gap takes cash, even when the business works.

On October 9, Oxide Computer Company announced a $445M Series D led by Eclipse. US Innovative Technology Fund, Riot Ventures, Jane Street, Friends and Family Capital and Counterpart returned. Atreides Management and AMD Ventures joined as new investors, with AMD extending an existing technical relationship into the cap table.

CEO Steve Tuck says manufacturing capacity grew 20x over the past 12 months and demand still exceeds supply. That is Oxide's account of capacity, not a revenue growth figure. Tuck and CTO Bryan Cantrill say existing cash, previous funding and debt facilities could cover the current backlog. Taking on additional demand would have required more caution.

This round changes how much of that demand they can accept. More working capital means committing to components and manufacturing before the customer gets a computer, with room for the supply disruptions the founders know can turn a promising order book into an uncomfortable calendar.

Tuck, Cantrill and Jessie Frazelle cofounded Oxide in 2019 around a specific proposition: enterprises should be able to own cloud infrastructure built as a complete system. Compute, storage, networking and open-source software arrive together. Developers provision resources through APIs while the organization keeps the hardware and the operating responsibility.

AMD EPYC processors are already part of that system. AMD's new investment follows its support for Oxide's platform software, giving the chip relationship a longer financial commitment. Eclipse is backing a company it helped launch, now buying materials for machines customers want delivered.

Oxide raised $100M in July 2025 and $200M in February 2026. The founders framed the latter around independence, because an enterprise planning years of infrastructure needs confidence its supplier will stay around. The Series D carries that promise onto the manufacturing floor, where the next customer order becomes a component purchase before it becomes a rack in somebody else's data center.

OptimizeGEO

Michael Kassan is putting his own money into a marketing problem that starts before the customer reaches a brand. At OptimizeGEO, the AI-generated answer becomes something people can examine, challenge and work on.

The MediaLink founder and 3C Ventures CEO joins the company as an investor and strategic advisor. Adweek reports a pre-Series A led by Young Capital at an $80M valuation. Kassan's investment amount is undisclosed; $80M describes the company's reported value, not cash raised.

Cofounders Kirthiga Reddy, CEO, and Saurabh Doshi have lived through a discovery change before. Reddy was Facebook India's first employee and MD, then a SoftBank partner. Doshi led teams at Meta APAC. Their company is built around the next place a customer forms an opinion: the response an AI assistant puts together.

That response has a supply chain. Product pages, reviews, reporting and retailer listings can all shape what a buyer hears. A brand can own its website and still have somebody else's description become the introduction. The marketing work spreads well beyond the page the team knows how to edit.

OptimizeGEO measures visibility, share of voice, sentiment and citations across major AI platforms. It then prioritizes changes and uses agents to carry approved work into content, technical updates, media and commerce, with human review. The company says those changes connect to traffic, conversions, pipeline and revenue.

The October 1 announcement puts its reach at more than 100 brand clients across 20 countries. Its Glance case reports a 2.5x lift in AI visibility and a 6.6x lift in AI-shopping share of voice from April through July 2026. Those are company-reported results; they describe a specific engagement, not a promised return for every buyer.

Kassan's disclosed remit fits the distance between a useful dashboard and a business that people will buy: global go-to-market strategy, enterprise relationships, partnerships and category development. Existing institutional backers include f7 Ventures and Micron Ventures. The announcement names them as part of the investor base, without assigning them a new check in this investment.

For Reddy and Doshi, the interesting customer conversation now reaches the team responsible for making the change. Someone has to approve a product correction, decide which retailer deserves attention and connect a better AI answer to what the business actually sold. Kassan joins while that work is becoming a commercial relationship of its own.

Jaca

Pete Ferrari's newest investor arrives with airtime attached. For Jaca, the allulose brand he founded through Healthy For Life Foods, that puts a different resource behind a familiar ambition: become something people keep in the kitchen.

Mercurius Media Capital announced a $1M strategic investment on October 8, 2026. MMC supplies premium advertising inventory in exchange for equity, helping consumer businesses reach larger audiences while preserving cash. The announcement gives Jaca access to that network; it leaves the ownership terms and media schedule undisclosed.

Ferrari, also Founder and President of Global Foods Group, wants national television to help introduce the product. MMC Founding Partner Piyush Puri is backing the opportunity to make allulose a more familiar category. Their commercial handshake reaches beyond getting another food brand recognized.

Jaca sells 100% pure allulose to consumers and serves wholesale and co-brand partners. The kitchen promise is practical: a sugar alternative people can cook and bake with. Its website puts recipes beside the product, which is where an unfamiliar ingredient starts becoming useful instead of merely interesting.

A television introduction has a short window to explain a purchase that may alter a person's usual recipe. The customer needs to understand what allulose is, what to do with it and where to buy it. A clever ad can start that conversation. The bag and the next batch of baking have to carry it further.

There is a useful financing distinction here. The $1M headline describes a strategic investment through a media-for-equity model, rather than a disclosed pile of unrestricted cash. Advertising access can preserve a company's money for other work, but the commercial value still depends on reaching people who care enough to try the product.

Global Foods Group reported 470% year-over-year Amazon revenue growth for June 2026. The dollar base was not disclosed, so keep the scale of that claim in view. It gives Ferrari a growth story to bring into a larger campaign while leaving plenty of work in understanding repeat demand.

For food operators, the deal puts category education inside the financing conversation. Jaca gets more room to explain allulose, and MMC gets an ownership interest in the brand doing the explaining. The next introduction travels from a screen toward a shopping decision, then into a kitchen where Ferrari's company has to earn its place beside the familiar ingredients.

Kapta Space

Kapta Space has customers waiting for radar that still has to earn its flight history. That puts manufacturing decisions inside the founders' engineering work, with classified missions giving the hardware a destination before its first orbital demonstration.

Milton Perque, Founder and CEO, and Adam Bily, cofounder and CTO, are taking that work forward with a $24M Series A announced October 9. Washington Harbour Partners leads, with Bison Ventures co-leading. MetaVC, MYDA Capital, Entrada Ventures, Aurelia Foundry and Fitz Gate Ventures also participated. Kapta reports $30M in total funding.

The Seattle company's mission centers on keeping track of moving threats across air, sea and land. A radar observation becomes useful to a tracking mission when the system can keep returning to the target as its location changes. Coverage, beam steering and revisit speed all enter the customer's buying decision.

Kapta calls its modular design a Metamaterial Electronically Scanned Array, or MESA. Electronic steering lets the radar redirect its beam; the company is designing the architecture for lower cost, power consumption and manufacturing complexity than conventional active electronically scanned arrays. Those advantages remain company claims, with flight qualification ahead.

Perque came from Echodyne. Bily brought antenna engineering experience from Apple and Astranis; investor StoryHouse Ventures also identifies Tom Driscoll as a cofounder. The product has to live inside somebody else's spacecraft, with the mission carrying its own demands for integration, power and operations.

Kapta says it secured multiple classified moving target indication contracts earlier this year. Customer names and contract values remain undisclosed. The Series A will support production, manufacturing expansion, flight qualification and preparation for an orbital demonstration, giving the capital a physical job.

Its commercial ambitions extend to radar imagery and infrastructure monitoring. The partnership approach on Kapta's website draws on existing satellite buses, data pipelines and mission services. Selling a payload can therefore make the team responsible for a handoff: delivering an instrument another organization can put to work in orbit.

Ramp

Karim Atiyeh's title at Ramp changed as the finance department's job expanded. The technical cofounder became co-CEO alongside Eric Glyman in June, formalizing a partnership that already reached across risk, operations and marketing.

That is a revealing management choice for a company moving deeper into the work finance teams perform. When software starts carrying out accounting tasks and tracking what AI consumes, decisions about the product become decisions about who gets to spend, approve and review.

Bloomberg now reports that Ramp has raised approximately $1.85B at a roughly $60B pre-money valuation. Dragoneer Investment Group and Thrive Capital reportedly led, with Founders Fund participating. Ramp declined to comment. The $60B figure excludes the new capital, and the report leaves the round's exact structure undisclosed.

Eric Glyman, Karim Atiyeh and fellow cofounder Gene Lee built Ramp around saving businesses time and money. The New York company now reports more than 70,000 customers. Corporate cards, expense management, bill payments and procurement gave it a place inside financial decisions before AI enlarged the assignment.

Take the AI invoice. A provider's total can tell a finance team how much was consumed while leaving the business owner searching for what the work achieved. Ramp's token spend management brings costs into view by provider, model and user. That gives the budget conversation a more useful starting point than a single swelling charge.

Then there is Ramp Stack, its AI accounting platform. Ramp describes reconciliations, schedules and reporting that can run through to the ledger, with outputs linked to source material. Accountants still need to review the work. Getting software closer to execution brings the reviewer closer to the consequences of an error.

For a buyer, those products belong in the same discussion. The permission to make a purchase, the record of its payment and the explanation at month-end all have to survive the trip. A beautifully automated step can still leave somebody hunting for the evidence required by the next one.

The reported financing follows a $750M raise at a $44B valuation in June. Ramp has not supplied a public allocation of the October proceeds. Its existing product direction is visible, though: more of the decisions around company money can happen inside its software, leaving finance teams to decide which work they are ready to entrust to it and which approvals they need to keep close.

Sonilo

Keli Li lost part of a winter holiday to a music library. A video for her Stories needed the right track, and the search kept going. The future Sonilo COO recognized the product's purpose before there was a funding headline to explain it.

Sonilo has now raised $11M in Seed funding, led by B Capital with Redpoint participating. Announced October 8, the round puts compute, creator and developer distribution, and music-industry partnerships behind a familiar creative frustration: footage is finished, but the sound still needs work.

CEO and co-founder Shawn Song led multimodal AI work at TikTok. CTO Alex Yin worked alongside him there, while Keli Li built and ran TikTok's global music operation. CMO Trista Taylor brings consumer-AI product and launch experience. Understanding the picture, the product and the people who own the music gives this team several different jobs to finish together.

Sonilo calls its technology the Sound World Model. Upload video, describe the sound in text, or use both; the system generates music and effects around the scene's pacing and emotion. The practical ambition is to let an editor judge what fits instead of spending the afternoon making a library track fit.

There is a second customer in that transaction: the person who has to approve the finished work. Sonilo licensed Shutterstock's music catalog for model training, a partnership Shutterstock has publicly confirmed. Commercial output rights still follow the applicable plan or API agreement. Licensed training and a client's permission to publish belong in the same conversation, with separate answers.

B Capital General Partner Daisy Cai sees generative audio becoming part of the AI-native video stack. TapNow and ComfyUI give Sonilo early distribution into creative workflows, while the web product and API already give users ways to work with it. The announced consumer app is planned for Q4; wider rights agreements remain work in progress.

That leaves Sonilo building on both sides of the editor's timeline. Better models need computing capacity. More useful distribution needs tools creators return to. Music partnerships need people who understand what the rights holder is agreeing to and what the production team is allowed to ship.

Keli Li's next music-industry conversations carry an unusual assignment: help make a creator's own footage useful to the musicians whose catalogs train the model. Sonilo's licensing business has to give both sides a reason to keep working together.

Hello Robot

Sixty people will help Hello Robot decide what useful assistance looks like. Some will be older adults, some care partners, some professional staff. They can live with the same robot and have very different reasons for wanting it in the room.

That is a product conversation worth funding. Hello Robot has received nearly $3M in a Phase IIB SBIR grant from the National Institute on Aging, part of the NIH, to study Stretch with the University of Illinois Urbana-Champaign. Announced September 29, the project began September 22 and is expected to run through May 2029.

Aaron Edsinger, cofounder and CEO, and Charlie Kemp, cofounder and CTO, built a mobile manipulator around a compact wheeled base and a telescoping arm. Stretch 4, introduced in May, puts that design in the hands of researchers and developers. The company manufactures its robots in Martinez, California, and also has an Atlanta office.

The new research focuses on older adults with mild cognitive impairment or mild dementia. Reminders, finding everyday objects, staying socially connected and support for daily routines are among the proposed uses. Language, planning, perception and voice interaction will contribute to the work.

A reminder arrives inside a relationship. The older adult has preferences about being prompted; a family member may be looking for reassurance; a member of the care team has responsibilities that continue after the robot speaks. Getting those expectations into the design is part of making the product usable.

Wendy A. Rogers and Aaron Edsinger are the principal investigators. Initial testing will take place at the McKechnie Family LIFE Home, where Illinois researchers can work in a home environment. Lindsey Communities remains a core research partner, connecting the project to the settings where assistance has to fit people's lives.

Extended evaluations will move into homes and care settings, including an 8-week period involving professional staff. Researchers will assess usefulness, reliability, trust, acceptance and safe operation. Those are study questions, with results still to come.

For a future care buyer, that work could make the buying conversation more specific: which routines deserve help, who controls the interaction and how much attention the system asks of the people around it. The grant puts older adults and their care partners inside the process of deciding what Stretch should do when it enters their day.

BioConsortia

A seed treatment spends part of its working life waiting. Before a grower puts it in the ground, the biology has to survive the commercial routine that gets the seed there. That waiting time belongs inside the product.

BioConsortia, Inc. has raised $15M to carry more of its microbial science through that routine. The October 9 equity investment came from an entity affiliated with John Pasquesi, managing member of existing investor Otter Capital. Founder and CEO Marcus Meadows-Smith is taking familiar investor backing into manufacturing, regulatory work and commercial launches.

The Davis, California company says its Always N nitrogen-fixing seed treatments have more than 2 years of shelf life, plus another 2 years on treated seed. For an industrial seed business, that is a practical invitation: biological nitrogen support that can travel through the handling and distribution systems already serving the crop. Farmers have enough jobs without becoming custodians of a fragile microbe.

Always N uses microbes to make atmospheric nitrogen available around the roots. BioConsortia offers wild-type and gene-edited versions for different regulatory environments. The commercial plan has to travel with the science, because a product's route into a market depends on what that market permits.

Zaffre, the company's biostimulant progressing toward launch, addresses emergence, establishment and crop performance under temperature stress. BioConsortia reports an average gain of 4.4 bushels per acre across 3 seasons of corn trials. That is a reported trial result, with the crop and conditions attached, rather than a universal yield promise.

The company reports growing Always N adoption through H&T in New Zealand and expanding commercial activity across several regions. The Mosaic Company is a commercial partner in the Americas. Those relationships put the product alongside people who already understand how agricultural inputs reach a grower.

Inside BioConsortia, Damian Curtis leads R&D and Sarah Reiter leads business development. September's addition of Matthew Mouw as VP of Business Development brings experience at Ball Horticultural Company and DuPont Pioneer into the work of matching microbial products with partners, vegetables and row crops.

The investor knows the platform. The commercial team now has more funding for the less photogenic work around it: keeping the organism useful through production, registration and the journey to a bag of seed that somebody is prepared to buy.

HyLight

A 12-metre airship has a remarkably earthbound job at HyLight: help somebody decide where to send a maintenance crew. The aircraft gets the photograph. The business earns its place when that photograph becomes an instruction a network operator can use.

Martin Bocken, CEO, Josef Rokusek, CTO, and Thomas Laporte, CBO, have built the French company around that entire journey. Aircraft, sensors, analysis and field operations sit inside the same organization. A buyer gets fewer handoffs between the flight and the answer, while the founders own more of the work needed to make the answer useful.

HyLight announced its Series A on October 12: €13M, approximately $14.5M at the day's ECB reference rate, led by Rubio Impact Ventures with DRW and Teampact.ventures. Y Combinator and Ring Capital also participate, according to the company-distributed release. Private Equity Magazine reports that the financing is split equally between equity and debt. Keep that accounting attached to the headline.

The company's announcement puts its starting position at 45 people, 4 HyLighters operating daily and 7,000 km inspected. Those are company-reported figures. Its target is 15 aircraft in operation in 2027, an expansion that brings production, pilot coordination, maintenance and data analysis into the same planning conversation.

HyLight's hydrogen-powered aircraft carry optical, LiDAR and gas-leak sensors over power lines, pipelines and railways. The company describes zero direct emissions in flight and near-silent operation. The precise qualifier matters: a flight claim tells us something useful about the inspection mission, without pretending to settle the whole manufacturing and hydrogen supply chain.

Enedis, NaTran, Téréga, SNCF Réseau and VINCI appear in the company's customer account. These operators need to identify faults and vegetation risks along infrastructure that stretches far beyond a convenient demonstration site. A clear image is valuable; knowing which stretch deserves attention gives that image a job inside somebody else's working day.

Rubio's investment case joins cleaner inspection flights to earlier fault discovery. HyLight plans to expand its fleet and automated analysis together, with Germany and the Netherlands among its priority markets. For the founders, every additional aircraft extends a promise already made on the ground: deliver observations clearly enough that the next repair crew knows where to go.

Primitive Labs

A customer can walk away from a perfectly functioning product. Primitive Labs is building around that freedom: the hesitation, changing priorities and lost patience that a successful software test can miss.

Sony Innovation Fund has invested in the San Francisco startup, with the announcement dated October 5. The investment amount remains undisclosed. The money supports behavioral modeling research and Instinct, its platform for simulating how customer populations move through digital experiences before launch.

CEO Rohit Talluri, CTO Jean Farmer and COO Gabriel Fong founded Primitive Labs in 2026 after working together at AWS. Talluri later worked in Amazon's AGI Autonomy Lab; Farmer worked on Amazon AGI and the Nova launches. They bring experience building agents that operate software to a different question: how the people using it decide to keep going.

Instinct builds customer models from an organization's research, analytics and behavioral data. Those models navigate journeys such as onboarding, subscription and checkout, giving product, design and research teams a way to examine where different audiences hesitate or abandon the task.

That changes the conversation before release. A checkout can satisfy the specification while asking for trust the customer hasn't offered. An onboarding flow can explain everything the company wants to say while exhausting the person it hopes to welcome. Primitive Labs wants teams to investigate those choices while they still have room to change the experience.

The company says it is working with partners across retail, e-commerce, media, entertainment, advertising and gaming. It also has backing from a16z speedrun, Olive Tree Capital and Pack Ventures. Those are disclosed company backers; the announcement identifies Sony as the new investor without spelling out a wider financing syndicate.

Sony Ventures' Atsushi Kitagawa points to the customer journey as an area his team has been evaluating as enterprises become more agentic. The connection is practical: faster software creation gives teams more versions to consider, each carrying assumptions about the people expected to use it.

For a product leader, the useful output is a decision worth revisiting. The model has to stay close to actual customer behavior, and the finding has to reach the designer or engineer who can act on it. Instinct's place in the release process will grow through those handoffs, one customer journey at a time.

Freshworks / Torii

Uri Haramati built a company around finding software nobody had a complete view of. He is now putting its future beside the place where employees ask IT for permission to use it.

Freshworks has signed a definitive agreement to acquire Torii, extending Freshservice toward software management and AI spend governance. Announced October 12, the deal is expected to close in Q4 2026, subject to customary closing conditions. Financial terms were not disclosed.

Uri Haramati, CEO, started Torii with Uri Nativ, CPO, and Tal Bereznitskey, CTO. The company launched in 2017 around a problem familiar to anyone inheriting a software estate: applications accumulate, ownership gets scattered, and the people responsible for the bill have only part of the picture.

Torii brought discovery, licenses, access, contracts and spend into the same working context. Its official funding history includes a $10M Series A led by Wing Venture Capital in 2021 and a $50M Series B led by Tiger Global in 2022. Those are historical financing rounds, separate from today's undisclosed purchase price.

AI gives that old inventory problem a variable bill. A license tells you someone can use a tool. Consumption tells you how much that person or team is actually spending. Torii says it tracks AI usage to the token, connects cost to people and teams, and forecasts where spending is heading.

Dennis Woodside, Freshworks' CEO and President, wants that evidence inside Freshservice. The service desk already sits in the path of access requests, employee departures and operating decisions. Connecting software cost to those moments gives IT and finance a chance to work from the same account of what is happening.

Freshworks has been assembling more of that operating picture. Its Device42 acquisition closed in 2024, adding infrastructure discovery and dependency mapping. Torii brings the software and AI layer to the proposed combination, with a different set of ownership and consumption questions attached.

For Torii customers, the immediate instruction is continuity. The companies remain independent before closing, and Uri Haramati says instances, integrations, workflows, support contacts and renewal terms are unchanged. The combined product's delivery timetable and commercial packaging have not been announced.

The interesting handoff starts when an employee asks for access and somebody has to approve both the work and the expense. Torii's next home could put more of that conversation inside the system already handling the request.