What the New System Has to Keep

Rocker

Alex Vogenthaler's description of running a large product organization sounds like a job nobody remembers putting in the job description: keep the company's unfinished business in your head. Know which decision is waiting, which teams are about to collide, and who needs the history before they act.

After leading product at Google and Stripe, Alex Vogenthaler built Rocker with cofounder and CTO Peter Terrill, whose background includes leading engineering at Stripe. Their starting point is familiar to anyone who has watched a capable team lose a week between two perfectly reasonable meetings.

Rocker announced $12.85M in funding on October 28, led by Greycroft, with participation from Anthropic and Menlo Ventures' joint Anthology Fund, Matchstick and Stripe. Jeanne DeWitt Grosser, Fran Ryan and David Singleton also invested. The announcement leaves the round label and valuation undisclosed.

The assistant is designed to hold an evolving picture of projects, people, decisions and dependencies. An employee can use it individually; deployed across a company, those assistants connect to shared context. The useful intervention arrives when somebody needs to act, with enough background to understand why the action matters.

That puts a fairly expensive management habit inside the product brief. People spend hours reading conversations because one of them might contain the detail that changes their assignment. Rocker's proposition is to carry that context forward so the next person can work with it, subject to the permissions around the information.

Peter Terrill has made customer control part of the design. Rocker says it can run inside a customer's own cloud, and that private context is shared only with approval. Its website displays SOC 2 Type II compliance. Those are company representations; the announcement supplies no independent measurement of the product's operating impact.

Greycroft's investment thesis centers on enterprises owning the memory their agents accumulate. That makes the deployment choice part of the commercial proposition: a business can want help connecting its decisions while caring deeply about where the record of those decisions lives.

Rocker is opening early access for individuals and companies. The founders are inviting employees to find value in their own workday, then asking their organizations to decide how much of that understanding can travel between colleagues. Somewhere inside that decision is the next commitment a manager has been carrying around alone.

Serval

Serval plans to hire more than 100 people to help customers figure out what to automate. For a company selling software that builds automations, that is a revealing place to put the money.

Jake Stauch, CEO, and Alex McLeod, CTO, have raised a $150M Series C led by Sequoia at a $3B valuation. Elad Gil, Evantic, First Round, General Catalyst and Redpoint returned; Conviction and Headline joined. Serval says the round brings its total funding to $277M; its disclosed Series A and Series B were $47M and $75M.

The San Francisco cofounders worked together at Verkada before starting Serval in 2024. Their customers know the awkward middle of enterprise automation: somebody understands the process, somebody else can build it, and everybody has a queue. A request to get an employee working can become a small project of its own.

Serval's Catalyst turns natural-language requirements into code-based workflows. Admins review, test and approve what gets published. The employee-facing agent uses the approved tools. That separation gives the person who knows the business a way to help build its software while IT retains authority over what can run.

The work also lives beside tickets, access and assets in Serval's service-management platform. Customers can start with an automation layer over their existing ITSM, then consider moving the system of record. That gives the sales conversation a practical sequence: show useful work before asking someone to move the filing cabinet.

Serval's SeatGeek case study describes 132 automations built in 8 weeks and more than 50% of IT tickets automated in 60 days. It also describes engineers moving into business units to build tools. Those are vendor-hosted customer results, with a useful operating consequence: resolving the queue can change the assignment of the people who used to manage it.

Now Serval is building ProServal, a services arm whose Automation Engineers will work alongside enterprise teams to identify opportunities and redesign processes. They sit inside a plan for more than 500 hires over the next 12 months. Roughly half of the automation engineers are expected to be recent graduates trained through a 4-week program.

The new hires will arrive with AI skills. Their customers will bring the history, exceptions and departmental habits that determine whether an automation belongs in the business. Serval is funding the time those two groups spend together, with the next process waiting for somebody to explain why it works that way.

NewEdge Power

SVP is committing to the people who secure a battery project's place on the grid. The investment structure gives those people a longer assignment than finding somewhere to put the equipment.

NewEdge Power and Strategic Value Partners have announced a partnership combining a platform investment with a utility-scale power joint venture. Its seed portfolio targets up to 1.2 GW in PJM, the grid operator serving 13 states and Washington, D.C. Financial terms were not disclosed.

Joe Nano, CEO and co-founder, leads strategy and capital formation. Steve Barna, CDO and co-founder, leads development. Their work meets at a difficult commercial handoff: turn a promising site into a project somebody can finance and eventually connect.

NewEdge will lead development under the venture, with SVP funding selected projects as they progress. SVP's investment in the company itself will be funded over the investment period to expand the development team and its capabilities. The arrangement supports both the projects and the organization moving them forward.

That distinction carries weight in power. A developer has to keep paying people to work through sites, permits and grid questions while individual assets move at different speeds. More demand can make the opportunity larger without making any particular parcel ready sooner.

NewEdge already develops utility-scale storage, solar and distributed power across U.S. markets. SJF Ventures led its Series A in 2025, following initial backing from Segue Sustainable Infrastructure and North Sky Capital. This is a new capital relationship around an existing development business, with a specific PJM portfolio to advance.

The timing has substance, and it requires care. PJM's 2026 forecast still projects substantial long-term electricity demand growth, while lowering near-term expectations after updating assumptions and examining large-load requests more closely. A data center projection can change while the work on a power site continues.

For landowners and development partners, that makes the staying power of the organization across the table commercially relevant. For NewEdge, the partnership brings an investor into the work between identifying an opportunity and deciding which project earns its next allocation of capital.

Barna's team now carries that work into a portfolio targeting up to 1.2 GW, with SVP deciding which advancing projects it will fund along the way.

Wabi

Getting friends to agree on dinner can become a small unpaid job. Wabi wants an AI assistant to take a shift, alongside the people who will actually sit down to eat.

Founder and CEO Eugenia Kuyda is building around that shared responsibility. After founding Replika, she is putting an agent inside plans, interests and routines that involve other people, with personalized mini-apps they can keep using together.

Wabi announced $55M in financing on October 29 at a $1.25B valuation, led by IVP. Andreessen Horowitz, Forerunner and Amazon Alexa Fund participated, alongside Tobi Lütke, Marc Benioff, Marissa Mayer, Dick Costolo and Alex Pall. The announcement leaves the series label unspecified.

The product combines an AI agent, shared spaces and tools built around what people want to do. Wabi describes families turning weekly meal plans into grocery lists, couples keeping restaurant lists and friends making book-club trackers. A conversation can leave behind something the group returns to next week.

That return visit carries the business story. A helpful answer has an audience of 1; a shared plan has people waiting on one another. Whoever normally sends the reminder, gathers the preferences and checks whether anyone booked the place has a reason to want the software to stick around.

Kuyda's earlier work matters here. Replika put her inside consumer AI companionship, where an ongoing relationship gives the product its place in somebody's day. Wabi brings that experience to software that participates in relationships people already have, while giving the group something practical to do.

a16z led Wabi's $20M pre-seed, announced in November 2025. Its investment thesis described consumers creating and remixing personal software, with creation and distribution living together. The current shared-assistant positioning gives that idea a familiar invitation: bring the person you already make plans with.

IVP's Alex Lim and a16z's Anish Acharya both point to Kuyda's consumer judgment in the financing announcement. Wabi says the money will support product development and expansion. It supplies examples of use, while leaving active-user totals, retention and revenue undisclosed.

The grocery list is a useful place to follow this company. Somebody has to update it after the dinner plan changes, and the other people have to find the update before they shop. Wabi is asking for a recurring place in that exchange, with the household still deciding what belongs on the table.

Crux Labs

The pull request is where Crux Labs' pitch gets interesting. The company wants a business expert's criticism of an AI agent to reach the engineer as a proposed fix, already tested against earlier interactions.

That puts the person reviewing the agent closer to the work of changing it. For enterprises asking software to handle sensitive customer conversations, the distance between those jobs can become expensive.

Crux Labs emerged from stealth on October 29 with $6M+ in an oversubscribed funding round co-led by Benchstrength and Misneach, with Analog Ventures and Pareto Holdings participating. Co-founder and CEO Julie Kheyfets and co-founder and CTO Krishna Kulkarni plan to expand research and engineering while accelerating enterprise deployments.

The New York startup begins with the people who know the business. Its platform captures their reasoning on actual agent interactions, calibrates evaluations to their judgment, and says it can run those evaluations across 100% of production traffic. Alerts surface problems; proposed changes arrive in the customer's repository.

There is a useful commercial choice inside that sequence. An engineering team can buy an evaluation tool and still inherit the job of persuading busy specialists to explain what went wrong. Crux is selling the work around the tool, positioning its agents as a dedicated evaluation team.

The specialists still have something substantial to contribute. Crux's own site acknowledges that experts disagree. Getting software to reproduce a score also means deciding whose judgment the score should represent, and whether that judgment fits the case in front of it.

That is a consequential assignment in healthcare, cybersecurity and financial services, sectors where Crux reports use by high-growth businesses and Fortune 500 companies. The customers remain unnamed, and the announcement gives no revenue or measured improvement figures. Its coverage claim describes how much traffic it evaluates; accuracy needs its own evidence.

Kheyfets previously led Tractable's North America business and served as Block Renovation's CEO. Kulkarni's background includes Meta, Checkr and Healthify. Their new company puts a familiar enterprise problem into the AI budget: the people building the software need a dependable way to learn from the people who understand the exceptions.

As Crux adds enterprise deployments, each customer brings another definition of an acceptable answer into that engineering conversation.

StimCardio

An ablation ends. The patient still has a week to get through, and a heart that can return to an unwanted rhythm. StimCardio is building a treatment assignment for that ordinary stretch of life after the procedure.

StimCardio Medical has closed an oversubscribed $30M Series A led by John Slump's Metis MedTech, with Catalyst Health Ventures and other new and existing investors participating. The Los Angeles company plans to use the financing for its NeuroPulse 2 pivotal study and preparation for an FDA De Novo submission.

Founder and CSO Amos Ziv, co-founder and CTO Tsahi Holand, and CEO Amir Soltanianzadeh are developing Rhyvive, an investigational wrist-worn therapeutic. It delivers non-invasive median nerve stimulation, targeting autonomic pathways involved in atrial fibrillation. The intended role is alongside catheter ablation, extending care into the home.

That makes the patient's participation part of the product. A wearable therapy needs someone to put it on, follow the treatment schedule and keep using it after the hospital visit has faded into the calendar. The company combines stimulation with physiological sensing, a companion app and cloud-based analysis, bringing the treatment routine and the information around it into the same development effort.

There is clinical substance behind the financing. The randomized, double-blind, sham-controlled NEUROPULSE pilot reported fewer recurrent atrial arrhythmia events with active stimulation. Its modified analysis included 59 patients, and a larger trial is needed to confirm the findings.

The endpoints deserve their own sentence: 6-month arrhythmia-free survival was 90.0% with stimulation and 80.8% with sham, a difference that was not statistically significant. A lower count of episodes and a demonstrated reduction in the number of people who experience recurrence are different claims. The study gives the team a reason to investigate further, with a precise question still open.

John Slump already sits on StimCardio's board. Catalyst managing partner Joshua Phillips does too, placing people familiar with medical-device development close to the clinical and regulatory decisions this capital will support.

Rhyvive remains investigational, without regulatory clearance or approval and unavailable for sale. For the team, the next stretch includes organizing the multicenter study and turning a treatment delivered at the wrist into evidence physicians can assess when discussing life after ablation with their patients.

DoorDash / Instacart

DoorDash put a promise about its competitors inside its Instacart acquisition announcement. Local grocers would remain free to use other delivery platforms.

That is a revealing commitment from a company proposing to put more of a retailer's online business under one roof. The grocer still has to trust the supplier, even when that supplier can bring the delivery network, storefront technology and advertising tools to the same conversation.

DoorDash co-founder and CEO Tony Xu and Instacart CEO Chris Rogers are proposing an approximately $17B deal, announced October 29. Instacart shareholders would receive $65 per share in value. At least 75% of the aggregate consideration would be cash, and shareholders' cash and stock elections are subject to proration.

Instacart arrives with considerably more than a recognizable shopping app. Its Marketplace, Enterprise platform and Ads business serve the people buying groceries, the retailers selling them and the brands paying to reach those customers. In Q3, Instacart reported $1.095B in revenue, including $331M in advertising and other revenue, up 23% from a year earlier.

For a grocer, those are connected decisions. Who runs the online storefront? Who fulfills the order? Who helps sell a brand's next placement? A wider supplier can make those decisions easier to coordinate. It also earns a more consequential place in the retailer's working day.

DoorDash brings a business that reported 993M orders and $4.616B in Q3 revenue. Its proposed combination would extend the relationship from a delivery request into more of the technology supporting the sale. The companies expect to retain Instacart's brand, app and enterprise products, while offering DashPass and Instacart+ through one membership fee. For households already paying for both services, the proposed membership bundle would give the combined company a more regular place in the shopping routine.

All of that sits beyond a substantial closing window. The companies expect the transaction to close at the end of 2027 or in the first half of 2028, subject to stockholder and regulatory approvals, including in Canada and Israel. They remain separate businesses until closing. The membership benefits are promises about the combined business, with implementation still ahead.

Xu and Rogers have made the retailer's freedom part of the proposition. Under a combined owner, the grocer would still be deciding which parts of its business to entrust to DoorDash and which relationships to keep alongside it.

Cohere

Cohere's founders want a bank's difficult AI assignment to improve the models its next customer gets. That is a more demanding business than selling each institution the same clever answer.

Aidan Gomez, Nick Frosst and Ivan Zhang describe a loop between domain-specific work and general model development. Customers expose failures that public benchmarks miss. Improvements can then travel into other products, giving the company another reason to stay close to the work inside regulated organizations.

Cohere announced the initial close of its Series E on October 29, targeting up to C$7B. The companies of Schwarz Group lead, with Bell Ventures, Bullhound Capital and the Government of Canada joining. Inovia Capital, NVIDIA and Prosperity7 Ventures return. The amount secured at this initial close was not disclosed; C$7B remains the target.

There is a physical side to that ambition. Compute capacity, model training and industry-specific products all need funding before the resulting capability reaches a customer's desk. Cohere says the proceeds will support that buildout, while its founders plan both frontier models and smaller models suited to different costs and deployment requirements.

CTO Phil Blunsom's product organization has a particularly revealing piece of work in public beta. Confidential Computing on Model Vault lets customer-side software verify evidence about the environment handling a request before sending sensitive data. The service uses NVIDIA technology on Microsoft Azure, making control something a buyer can examine inside the deployment.

That belongs in the financing story. A security team evaluating AI also has to decide who may operate it, where information will go and what evidence justifies permission. North brings search and agent workflows into that conversation. The University of Toronto partnership gives Cohere a concrete institutional setting for privately deployable AI under local data control.

Cohere reports 165% year-over-year ARR growth through Q3 and 300 customers globally. Those figures describe the company's account of demand. The founders' research loop now has more organizations supplying the difficult assignments, while the platform has to keep their information within the boundaries each organization accepts.

For Aidan Gomez, Nick Frosst and Ivan Zhang, building a model business means staying useful to the people who decide whether that model gets inside the institution at all.

Salesforce / Rune Technologies

David Tuttle and Peter Goldsborough made unreliable connectivity a design condition for Rune Technologies. The military logisticians using its software still have to decide where supplies go while the network is having a very bad day.

Salesforce has signed a definitive agreement to acquire Rune, announced October 29. The proposed combination would bring that field-level judgment closer to the enterprise workflows inside Missionforce, extending Salesforce's responsibility into places where waiting for a server is a poor operating plan.

David Tuttle, Rune's co-founder and CEO, and Peter Goldsborough, co-founder and CTO, started the company in July 2024 after working at Anduril. Their product, TyrOS, brings inventory, personnel, equipment and distribution into the same logistics picture. Knowing supplies exist is useful. Knowing which vehicle, crew and route can actually move them gives that information a job.

Rune describes an architecture that keeps critical data available locally and exchanges prioritized updates when connections return. That recovery deserves attention. Headquarters needs a usable account of decisions made while a formation was disconnected, and the people in the field need software that lets them keep working in the meantime.

The company says the Army selected TyrOS for I Corps under its Next Generation Command and Control production effort in October, following prototypes with the 4th and 25th Infantry Divisions. Rune also supplies the logistics application for the Marine Corps' Project Dynamis. Those relationships give the acquisition an existing operational assignment, with users already involved in shaping the software.

Earlier investors financed that work: a $6.2M seed led by Andreessen Horowitz's American Dynamism fund, then a $24M Series A led by Human Capital in July 2025. The acquisition price remains undisclosed. Salesforce expects closing in fiscal Q4 2027, subject to customary conditions, so the integration described in the announcement is still ahead.

Kendall Collins, President and CEO of Missionforce and Government Cloud at Salesforce, would take on a broader handoff if the transaction closes. A supply decision made away from a reliable connection has to find its way into the records used to maintain equipment, coordinate people and plan the next movement. Rune's founders built for the disconnected stretch. Joining Salesforce would carry their work further into the institution that has to act when those updates arrive.

Ozlo

Bose stopped making Sleepbuds. The people who depended on them still had to get through the night, which gave 3 former Bose colleagues a rather personal product brief.

N.B. Patil, Brian Mulcahey and Charles Taylor acquired and licensed sleep-related technology and built Ozlo around that unfinished relationship. Patil is now CEO, and the company says it has passed 260,000 customers in under 2 years. Somebody has to persuade those people to put electronics in their ears, turn onto a pillow and leave them there.

That is a demanding customer interview. The person conducting it is trying to fall asleep.

Ozlo announced a $22M Series B on October 29, led by Modi Ventures. The total includes previously issued convertible instruments, so the headline needs to travel with that qualification. LifeArc and ARTIS Ventures, which co-led its $10M Series A in 2023, are among the returning participants.

Resmed, Scrum Ventures, Drive by DraftKings, Wheelhouse 360, WISE Ventures, Niremia Collective, Accelerate Ventures and World Solver Syndicate also participated. Nick Foles, Danny Davis, Melissa Urban and Michael Gervais joined the round. There is a broad group financing a very intimate piece of consumer behavior.

Sleepbuds 2 uses noise masking and a passive seal to soften interruptions such as snoring and street noise. Ozlo advertises up to 14 hours of battery life, with results depending on use. The low-profile design has to accommodate side sleepers, and the software has to make the evening easier without giving the owner another evening job.

The investment reaches beyond selling the next pair. Ozlo plans to expand globally and put more into hardware, software, engineering and research. Its Calm collaboration connects the device to sleep content. The company also describes a Dreem Health partnership that offers a route from the bedroom to sleep specialists.

Those relationships give the founders different assignments. Content has to fit a nightly routine. Clinical care has to recognize when a person needs evaluation beyond what a consumer product can offer. Ozlo is also preparing regulatory work toward submitting a tinnitus product to the FDA for review; that remains development work, with the medical claim still ahead.

Patil's team now has a larger audience living with the consequences of its smallest design decisions. As Ozlo adds services around Sleepbuds, that tired customer still gets to decide, every night, whether the whole arrangement stays in their ears.

Mintive

An employee leaving the company should not take a useful AI agent's access with them. Mintive has built ownership transfer into its agent identities, giving IT a way to keep the work running after the person who set it up has moved on.

That is an unusually practical place to enter the agent conversation. Somebody has to own the connections, decide which tools the agent can reach and answer for the actions it takes. A clever assistant becomes an organizational responsibility surprisingly quickly.

CEO Jiquan Ngiam, CTO Vijay Vasudevan and fellow co-founders Bozhi See and Phillip Bensaid are building Mintive around that responsibility. The San Francisco Bay Area company, formerly MintMCP, announced a $7M round led by Cowboy Ventures on October 29, bringing its total funding to $11M.

Operator Collective and Burst Capital joined as new investors. Coatue, Hustle Fund and Maven Ventures participated again. The capital backs a company selling the infrastructure around agents: identities, permissioned tools, model access, monitoring and coworker agents inside Slack and Teams.

Mintive's design gives each agent its own identity while delegating access through the person asking, a team credential or a service account. Administrators can narrow a tool's permissions or revoke an agent. The person who originally clicked create can hand over ownership without taking the whole arrangement apart.

Its LLM gateway also assigns approved models and team budgets. The same IT organization can decide which model gets used and how much a team can spend.

There is a customer story underneath those controls. Modern Treasury describes using MintMCP for a bank-operations agent that checks documentation, code, production data and error reports before a support specialist responds. Its engineering team reports saving technical account managers 30 minutes to 4 hours per case. Customer-facing actions still require human approval.

That division of labor deserves attention. The support specialist receives a case with more of the research assembled, while keeping responsibility for what the customer is told. Access to several systems becomes useful because the agent has a defined assignment across them.

Mintive reports more than 50 paying customers and over 2.5M agent calls a day. As those deployments grow, its customers will keep changing teams, models and permissions. Ngiam's team is selling the continuity that lets an agent keep doing its assigned work while the people responsible for it change seats.

Verisma

Verisma gives its clients a say in when the AI has to stop. An incoming document that falls outside a healthcare provider's operating rules goes to a human reviewer, keeping a person responsible for documents that need judgment as they move through the clinical workflow.

That detail deserves a place in the financing story. On October 29, the Alpharetta, Georgia company announced senior-secured financing from Golub Capital, including a delayed-draw term loan facility, to support technology investment and acquisitions. The amount was undisclosed. Existing private equity backer NewSpring remains part of the picture.

CEO Marty McKenna is expanding a business that follows health information through release, retrieval, care coordination and archiving. Founder and Executive VP Andy McManus is still there, building on a history the company dates to 2002. This is an established healthcare operation taking on a broader assignment as AI reaches further into its customers' daily work.

Inbound Document Intelligence receives and routes documents, supports follow-up actions and extracts information for quality reporting. Verisma describes a workflow in which the provider's procedures govern the automation and exceptions come back to people. A care team needs the information in the right place, with the unresolved work visible to somebody who can act on it.

Chief Technology & Product Officer Anupriyo Chakravarti has made a related argument about interoperability: a long list of integrations can still leave important records out of reach. His examples include archived material and less common clinical systems. The company has to work with the information environment its customer actually has, including the parts nobody would choose for a new installation.

NewSpring already helped fund Verisma's 2023 combination with ScanSTAT. Golub is now providing debt financing for the next phase of platform development and potential M&A. The two relationships have different jobs, and the delayed-draw structure leaves the amount actually borrowed outside the public announcement.

For healthcare buyers, Verisma's expanding scope puts more of the information handoff inside one supplier relationship. Each addition to that scope gives McKenna's team more work to coordinate: where a record lives, who may use it, which action follows and who gets the exception when a client's rules say the machine has gone far enough.

Dren Bio

Benjamin Kozub joined Dren Bio to lead commercial work in May. The company's lead medicine is still investigational, which tells you something about how early the work of reaching patients has to begin.

On October 29, the San Carlos biotech closed an oversubscribed $350M private financing co-led by Juvia Group and Lightspeed Venture Partners. Founder and CEO Nenad Tomasevic is advancing dibotatug, an antibody targeting CD94, toward regulatory submission and a planned commercial launch.

The lead studies concern cytotoxic T-cell lymphoma and large granular lymphocytic leukemia, rare blood cancers with substantial unmet need. Dren reports more than 300 patients dosed in its Phase 1/2 program. Its descriptions of responses and tolerability remain company-reported; the financing announcement supplies no numerical response rates.

That leaves a demanding assignment for the people joining the business. A commercial team has to prepare for a medicine whose approved use, timing and eventual place in care still depend on clinical evidence and regulatory decisions. Building too late can waste precious time. Building early requires paying people to prepare for an outcome nobody can guarantee.

Tomasevic and President and COO Amit Mehta have another route running alongside that work. Sanofi acquired the separate DR-0201 program, now SAR448501, in 2025, under a deal with $600M upfront and contingent milestones. Dren continued independently, carrying dibotatug and a broader antibody pipeline forward.

The distinction matters to anyone reading a biotech deal headline: buying one program can leave its creator with more programs to develop, partnerships to manage and a commercial organization to build. The same scientific company can supply a pharmaceutical partner and prepare to take its own medicine further.

Juvia's Sujal Shah and Lightspeed's Shelley Chu join the board with this financing. New participants include Mirae Asset Capital Life Science, General Catalyst, Cormorant Asset Management and OrbiMed, alongside returning backers. The company calls this a private financing and discloses no valuation or series letter.

For Kozub, who previously led commercial work at CymaBay, the assignment now sits beside clinical development, manufacturing and regulatory colleagues already on Dren's team. The $350M supports their overlapping calendars as dibotatug moves through development and the company prepares for the possibility of putting its own medicine into doctors' hands.

Neuron Industries

The old controller stayed intact. After Neuron Industries replaced a 1991 Siemens S5 at CureWood, the customer still had a way back to the machine's previous setup.

That is a revealing detail for a startup selling speed. In Neuron's account of the retrofit, the hardwired safety chain also stayed in place. Getting a lumber line running on new controls involved deciding what to preserve, right down to the equipment somebody might need if the change went badly.

Co-founders Dennis Ren, CEO, and Kenneth Rhee, CTO, are building Neuron around that contact with the physical world. Ren built custom controls at Freeform. Rhee founded WorkIPO, developing industrial enterprise software, after working at Amazon. They have taken on the controller and the software around it together.

Their $10M Seed, announced October 29, was co-led by FCVC and Obvious Ventures, with Y Combinator, Thin Line Capital, Zee Prime Capital and Founder Factor participating. The El Segundo company plans to put the capital into its industrial-control product family, team and domestic production capacity.

Cortex AIC and Synapse let engineers describe machine behavior, inspect generated control code and test it before deployment. The useful ambition is to let an engineer spend more time deciding how a machine should behave, while software handles more of the work of expressing that decision.

CureWood gives that ambition some sawdust. Neuron reports about 3 hours of human engineering and an 8-hour cutover for a line with 189 hardwired inputs and outputs. Those are different clocks. Moving the wires still took a shift, even when preparing the logic moved quickly.

The case study also describes using recorded operating data to trace an intermittent sensor problem to dust on a lens. That puts a modest, practical benefit beside the headline speed claim: help the person responsible for the line understand a fault while the evidence is still available.

Early deployments are company-reported, and Neuron's current site lists estimated releases of Cortex Mini in Q1 2027 and Cortex in Q2 2027. Pilots give engineers a place to start while the team prepares the wider product rollout, with production work still on its calendar.

FCVC and Obvious are backing a supplier whose work continues after the new controller powers up. At CureWood, the next conversation was already about additional instrumentation and changes the owner had wanted for years. Ren and Rhee now have more capital to stay in those conversations.

Dulo

Sebastian Thrun is moving upstream from the machine that drives to the work of designing the machine. With Adrian Aoun, he is building Dulo around foundation models for hardware design, taking another run at the space where AI has to answer to the physical world.

The San Francisco startup's $130M Seed round surfaced on October 29, 2026. Aoun confirmed the financing himself, giving a very short public company description a substantial financial commitment behind it.

The backers include Eric Schmidt, John Doerr, Marc Benioff, Yuri Milner and Ram Shriram. Construct Capital, Lowercarbon Capital, Bling Capital, BoxGroup, SV Angel, Haun Ventures and Scribble Ventures also participated, according to TokenPost; The Information separately named DST Global. The sources reviewed identify no lead investor or valuation.

Thrun's history includes the self-driving project that became Waymo, Google X and Google Brain. Dulo's website points to leaders from Waymo, Google Brain and SAIL. That background puts the founders close to a demanding engineering question: how much of engineering judgment can a model make useful to someone building an actual object?

A generated design still needs a manufacturing method. A shape that works for one process can be an expensive suggestion for another. Autodesk's existing generative-design tools already make those constraints part of the assignment, which gives Dulo a market with established habits and people who know exactly where the work gets difficult.

Dulo is early in explaining its approach. The site says coming soon; the public record reviewed offers no customer list, performance benchmark or supported CAD integration. Specific hardware targets and a product release date remain undisclosed. Putting those details into a financing story would mean writing a product launch the company has yet to give us.

The opportunity is to make a broader range of engineering possibilities worth examining without asking the engineer to inherit an unmanageable pile of checking. That is the commercial relationship taking shape beneath the foundation-model ambition: help people explore further while giving them work they can evaluate and carry into production.

Thrun and Aoun now have the backing to develop that relationship. Somewhere in the future product is a decision about which engineering task to put in front of a customer first, and how much of that customer's working day Dulo can earn.