The Name That Comes Up

Cherubic Ventures

The easiest way to make an AI fund look modern is to make it enormous. The harder move is deciding which founder deserves the first check before the market has built a spreadsheet around the answer.

Cherubic Ventures closed Fund VI at $68.88M on September 1, taking firm-wide AUM across 6 funds above $500M. The repeating 8s nod to prosperity and good fortune in East Asian culture. Less sentimental allocation problem.

Founder and Solo GP Matt Cheng is taking that capital into AI infrastructure, developer tools, enterprise software, healthcare, physical AI, and robotics. Those categories can consume extraordinary amounts after a company proves itself. Cherubic's job happens earlier, when the product is unfinished and a founder needs the first serious believer more than another observer.

The firm says Fund VI companies have raised more than $500M in subsequent funding. That figure belongs to the portfolio companies, not Cherubic's AUM, but it shows the handoff: one fund identifies the opportunity, then a larger capital stack arrives when the evidence catches up.

Sudo AI makes the thesis tangible. Cherubic says it was the robotics company's earliest institutional investor. Two years after founding, Sudo AI has reached a company-reported valuation near $2B. Its simulation-trained system is designed to handle unfamiliar objects without new real-world data for every task. If that holds outside a curated demonstration, the prize is a shorter path to useful physical work.

Fund VI also includes Entire, founded by former GitHub CEO Thomas Dohmke, plus Patlytics, Max AI, Generation Lab, and therapiAI. The portfolio runs from code and patents to healthcare and embodied intelligence. Different markets, same wager: the first useful version may appear before a category agrees on its buyer or budget.

Cherubic has history behind that posture. Earlier investments include Hims & Hers, Flexport, Calm, Paidy, 91APP, and Astranis. Fund V closed at $110M in 2022. Fund VI is smaller by headline amount, but Cherubic has not disclosed check sizes, reserves, ownership targets, named LPs, or return data. The discipline has to show up in selection, not in a story invented around missing numbers.

AI will keep attracting vehicles large enough to sound like infrastructure projects. Cherubic is defending the moment before scale, when conviction has to arrive without consensus. Its next decade depends on whether those early checks keep finding founders the later market wishes it had met sooner.

Outcomes One

Inside a pharmacy, the clinical opportunity rarely arrives with extra time attached.

The same team is processing prescriptions, handling insurance questions, managing inventory, answering patients, and keeping the counter moving. A medication review or adherence intervention may be valuable, but it still has to fit inside that operating reality.

That is the system Outcomes® is trying to connect.

H.I.G. Capital has completed its acquisition of Outcomes One, an Orlando-based pharmacy technology and clinical services company. The purchase price and transaction structure were not disclosed.

Outcomes sits across two layers of pharmacy work. Rx30 and Computer-Rx support the mechanics of dispensing, billing, inventory, and workflow. Its clinical network and Outcomes Premium connect pharmacies with health plans and pharmaceutical companies around pharmacist-led interventions.

The scale is substantial. H.I.G. says Outcomes reaches approximately 85% of U.S. pharmacies. In 2025, the company facilitated more than 10 million clinical interventions across 130 million covered lives.

Those numbers explain why this is more than a conventional healthcare software acquisition.

A clinical opportunity can be identified by an algorithm, funded by a program, and still disappear if it reaches the pharmacist as one more disconnected task. The product is not only the software or the network. It is the handoff between an identified need and a moment when the pharmacy can act.

Jude Dieterman, Outcomes One's CEO, says the partnership will support continued investment in the pharmacist-led clinical network and platforms including Rx30, Computer-Rx, and Outcomes Premium. H.I.G. Managing Director Arjun Mohan framed the deal around expanding technology, services, and innovation while elevating the pharmacy's role in healthcare.

The distribution is already there. The harder work is proving that reach can become repeatable care capacity.

The acquisition announcement does not disclose intervention completion rates, patient outcomes, pharmacy economics, or customer retention. Those are the measures that will determine whether a national footprint creates durable clinical value or simply more tasks behind the counter.

Healthcare keeps asking pharmacies to become a larger front door for care. H.I.G. is now backing the operating layer that may decide how wide that door can open without overwhelming the people standing behind it.

Medici Brands

One successful food brand can be instinct. A portfolio of them needs plumbing.

That is the operating bet inside Medici Brands' $250M Series B. Existing investors Greenoaks and Valor Equity Partners co-led the round, with CEO Peter Rahal, ICONIQ, and Imaginary Ventures participating. Axios reports a $2.25B post-money valuation.

Medici is the new parent company behind David Protein, HallPass, and Epogee. David launched direct-to-consumer in September 2024 with a bar packing 28g of protein into 150 calories and 0g of sugar. The company now says the brand reaches more than 35,000 retail locations, including Walmart, Target, and Costco, and is on track to pass $300M in 2026 revenue.

Those are company-reported numbers, but the velocity explains the size of the check. David has moved from bars into frozen dessert and ready-to-drink shakes. HallPass launched nationwide at Walmart in August with candy designed to carry fewer calories and less sugar at a mass-market price. Medici says Rowdy will follow later this year.

The interesting part is underneath the labels.

David used much of its $75M Series A in 2025 to acquire Epogee, the company behind EPG, a plant-based fat system used to preserve familiar texture with fewer absorbed calories. That deal secured a critical ingredient and widened the product map. It also moved the company from buying formulation leverage to owning part of it.

Now the capital is moving one level higher. Medici says the Series B will expand HallPass, extend David into new formats, advance product innovation, and build the infrastructure needed to launch and scale more brands.

That infrastructure is where a clever product thesis becomes an institution. Formulation has to travel across categories. Retail relationships have to support different price points and buying occasions. Manufacturing, quality, packaging, inventory, and customer trust have to scale without making every new label feel like David wearing another outfit.

Peter Rahal already knows how much brand clarity can move a product from building RXBAR. Medici is asking a harder question: can the instincts that built one breakout company be converted into a repeatable operating system for food?

The $250M buys room to answer it across protein, candy, snacks, and whatever comes after Rowdy. The next chapter will show up on shelves before it appears in another funding announcement, one distinct brand at a time, all depending on machinery the customer never sees.

Archive

Creator marketing gets expensive in a peculiar way. The audience can grow faster than the team’s ability to see what creators made, secure the rights, and prove that any of it moved the business.

The content is everywhere. The control room is usually a spreadsheet with trust issues.

Archive has closed an undisclosed strategic funding round to push deeper into that operating gap. Anti Fund and Florida Funders led the financing, with Battery Ventures, Stripe, Tiger Global, Lux Capital, Human Capital, and more than 50 founders and executives participating.

Co-Founders Paul Benigeri, CEO, and Geoffrey Woo, Chairman, started Archive in 2021 after working as e-commerce operators. Stories disappeared, posts scattered across platforms, and teams spent hours saving evidence their customers and creators were already producing.

Archive gives brands social listening plus creator-program software to find creators, run campaigns, manage content, and turn posts into paid partnership ads. The company says its AI understands short-form video across TikTok, Instagram, and YouTube, including untagged and disappearing content.

Archive reports that more than 1,000 brands use the platform, including Crocs, DoorDash, and Pinterest. It says it processes tens of millions of videos each day and recently signed 2 seven-figure enterprise contracts with unnamed global consumer brands. Those are company-reported figures.

IAB projected U.S. creator ad spend would reach $44B in 2026 after estimating $37B for 2025. As that budget becomes a recurring channel, brands inherit a less glamorous obligation: find the right creator, catch content before it vanishes, manage permission, and connect attention to a commercial result.

That is where the round gets interesting. AI can search more video, surface more creators, and automate more campaign work. The valuable product is the visibility to know which human voice fits the brand and which moment is worth amplifying.

Archive did not disclose the investment amount, valuation, security, ownership, or an updated total raised. It also did not name the 2 large AI labs it says use the platform. A strategic round is not permission to invent the arithmetic.

Creator programs are becoming too large to run as a pile of tabs and human memory. The category will be shaped inside the decisions automation cannot make cheaply: whose credibility travels, which content carries the brand, and when a fleeting post becomes an asset the business can keep using.

Capacity

Enterprise AI has developed an expensive habit: every solved problem arrives with another vendor, another training loop, and another version of the customer truth.

The web bot knows one policy. The voice agent knows another. Quality assurance and analytics sit elsewhere, leaving the human representative to play air-traffic controller while the customer waits.

Capacity has raised more than $54M in Series E funding to bet against that fragmentation. The St. Louis company says the round brings total capital raised above $159M. Kathy Ireland is the only investor named in the closing announcement and is described as a lead investor; Capacity did not disclose the valuation, terms, or the rest of the syndicate.

David Karandish and Chris Sims founded Capacity in 2017 around repetitive support work in disconnected systems. Enterprises have since filled those systems with AI, letting the machines repeat the fragmentation faster.

Capacity's answer is a shared knowledge layer spanning omnichannel AI agents, agent assist, automated quality assurance, conversational intelligence, and outbound campaigns. Its “train once, use everywhere” argument is simple: a customer conversation should not lose context whenever it changes channels.

Enterprise AI budgets are becoming more selective. The first cycle rewarded experimentation. The next has to explain integration, governance, measurable return, and why several tools should survive when one platform can carry the same knowledge.

Capacity enters with company-reported scale. It says ARR passed $100M in June after growing 20x in 3.5 years, more than 20,000 organizations use the platform, and 813% three-year growth placed it at No. 432 on the 2026 Inc. 5000. The figures are not independently audited here, but this is beyond demo-stage theater.

The capital will support R&D, go-to-market, customer success, partnerships, integrations, and international expansion. Capacity has also expanded through acquisitions, which makes integration more than a sales line. A platform can put many capabilities on one invoice; the harder work is making voice, chat, QA, analytics, campaigns, and human escalation behave as though they learned the customer together.

Capacity is financing that handoff while enterprise buyers count how many systems their AI strategy created. The next chapter sits inside the moment when a conversation changes channels and nobody has to ask the customer to begin again.

Critical Materials Group

A missile can have better software, better sensing, and a cleaner procurement slide. It still needs the material that makes it move.

That upstream reality is where Critical Materials Group is putting a $10.3M Seed led by Overmatch Ventures, with Victory Six Advisors and Fulcrum Ventures participating. The Texas manufacturer is building modular, automation-ready production for energetic materials that sit inside munitions long before anyone debates the guidance system.

Kevin Capozzoli, CEO and co-founder, came to the problem after Army Special Forces service and defense-technology operating roles. Director Travis Swanson brings decades around energetic-material formulation, testing, qualification, and private laboratories. One knows the cost of a thin industrial base from the user side. The other knows what can go wrong inside the process.

CMG is designed to add capacity beside government-owned, contractor-operated plants and established defense manufacturers. Smaller, distributed production nodes could give primes and government customers more flexibility without forcing proven facilities to stop carrying today's load.

The financing will support platform development, facility commissioning, engineering and operations hiring, regulatory milestones, and existing contract commitments. Tectonic Defense reported 27 employees, with company plans to reach 80 by the end of 2026 and 200 by the end of 2027. Those are targets, not finished capacity, but they show how quickly this becomes a recruiting, compliance, and execution story.

The capital record deserves precision. An August SEC Form D reports $7.8M sold to 15 investors, while the company announced the broader $10.3M Seed on September 2. The filing is a regulatory snapshot, not a reason to quietly round the facts until they behave. No valuation was disclosed, and customers and contract values remain private.

Overmatch General Partner Morgan Hitzig described energetics as infrastructure upstream of nearly every U.S. kinetic capability. That thesis is less glamorous than another autonomy montage and more consequential when inventories get tested. Advanced systems cannot outrun missing inputs.

CMG now has to turn distributed production into commissioned facilities, regulatory progress, and measured throughput without giving safety the usual startup treatment of a slide near the appendix. The next chapter will be written in customer deliveries and whether one more production node makes the wider network harder to exhaust.

Elucid

Cardiology has spent decades getting better pictures of the artery. The harder question is whether the picture explains which lesion can hurt the patient.

Elucid has raised a $55M Series D to work on that decision. The oversubscribed round includes IAG Capital Partners, Elevage Medical Technologies, and a newly disclosed but unnamed large medical-technology company that Elucid says is now its fourth publicly traded strategic investor.

The company has built its platform around coronary CT angiography. Plaque-IQ quantifies and characterizes plaque so a physician can look beyond how narrow an artery appears and examine the disease burden inside it. The FDA cleared Plaque-IQ under 510(k) K241524 in 2024.

Elucid is also developing FFR-CT to estimate whether a specific narrowing is restricting blood flow. That product remains under FDA review. The intended platform combines plaque composition with physiology, giving clinicians two different views of the same lesion.

That sounds like a product story. The Series D makes it an execution story.

The money will support commercial expansion, continued AI research and development, and landmark clinical studies. Elucid reports that positive coverage policies now reach approximately 70% of insured U.S. lives. The next job is turning that coverage, a cleared product, and a growing evidence base into routine clinical use.

One part of the evidence program is AI-PREDICT, a retrospective international multicenter study planning approximately 1,000 subjects across more than 20 sites. The study is designed to examine lesion-level risk, but it has not produced the outcome proof that adoption will ultimately require.

That boundary matters. Healthcare AI earns trust when the claim stays aligned with the evidence. Plaque-IQ is cleared. FFR-CT is under review. The coverage figure is company-reported. A planned study is not a result.

Kelly Huang, PhD, is leading Elucid into the phase where those distinctions become commercial obligations. The company now has to fit the technology into cardiology and radiology workflows, advance reimbursement, complete the clinical work, and show physicians what changes because the measurement exists.

Cardiovascular disease caused an estimated 19.8M deaths in 2022, according to WHO. The scale of that burden creates room for better tools, but it also raises the standard for what better must mean.

A more detailed image is useful. A more confident treatment decision is the outcome that matters.

Newlight

The highest-value investor in maritime hardware may be the one willing to lend you a ship.

Newlight has raised a $9M Seed round from lomarlabs, BIRD Energy, Undeterred Capital, CiRi Ventures, and Fusion VC. The financing followed an 8,500-nautical-mile commercial voyage from Singapore to Ghana aboard a 650-foot Lomar Shipping bulk carrier, where Newlight says its hydrogen-hybrid retrofit cut fuel use 24%, CO2 emissions 28%, and CO emissions 22%.

That sequence matters. Haran Cohen Hillel, co-founder and CEO, and Evyatar Cohen, co-founder, VP and COO, put the hardware inside a working engine room, kept the ship moving for a month, and brought operating data back to the capital conversation.

Newlight is building for the fleet already earning money. Its system adds precisely controlled hydrogen to marine diesel engines, reading load, pressure, temperature, flow, and safety signals. If a reading moves outside the approved range, hydrogen delivery stops and the engine continues on diesel.

Newlight says installation takes 1 to 2 weeks without drydocking or taking the vessel out of service. For a similar ship, it projects up to $500K in annual savings and payback below 18 months. Those economics are company projections, not fleet-wide audited results, but they address costs shipowners already understand.

RINA previously completed a four-day Factory Acceptance Test covering safety layers, controls, electrical integration, engine behavior, emergency stops, fire and leak detection, and conventional-fuel changeover. Vessel-specific class review and commissioning still apply. The argument has moved from a lab promise toward the operating conditions where marine technology earns trust slowly.

The investor list tells the same story. lomarlabs is the venture arm of the shipowner that supplied the vessel. The syndicate backed a company trying to make decarbonization useful before the wider fuel and vessel infrastructure catches up.

For industrial climate technology, that combination is unusually useful. Capital pays for the next build; operating access reveals whether the build belongs anywhere near a customer's balance sheet.

Newlight says it now has commercial agreements covering 12 vessels across multiple shipping companies, although only Lomar is named publicly. The next stretch belongs to those engine rooms: different loads, routes, crews, hydrogen supply conditions, and class reviews. Every installation will add another line to the record shipowners actually buy.

Aslan

An analyst can watch an underground forum for months and still remain outside the room.

The messages are visible. The usernames are visible. What is missing is presence: the ability to stay inside a network long enough to understand who controls it, how participants move, and which signal changes an operator's next decision.

Aslan has raised a $20.8M Seed co-led by Khosla Ventures and XYZ Venture Capital, with 2048 Ventures, BoxGroup, Liquid 2 Ventures, Alumni Ventures, and others participating. The Washington, D.C. company also used the announcement to come out of stealth.

Chase Reid, CEO and co-founder, and Ansel Tessier, CTO and co-founder, are building human-supervised AI operatives for national-security and law-enforcement missions. The agents enter underground forums, Telegram channels, and other digital environments where foreign adversaries and transnational criminal organizations recruit, finance, coordinate, and hide.

Axios reports the FBI, Homeland Security Investigations, and other agencies have worked with Aslan, often through shorter experiments while they resolve questions around guardrails, data collection, storage, and operational use.

The company says its operatives have mapped a live smuggling network, uncovered sanctions-evading cyber fraud, surfaced technology-transfer pathways tied to U.S. AI infrastructure, and exposed a foreign recruitment effort targeting U.S. defense professionals. Those are company-reported examples, not public case files or independent benchmarks.

Chase Reid also drew a public line against domestic operations that target or surveil Americans. As capability expands, customers will test whether that boundary, human supervision, and data governance remain part of the operating system.

Aslan had 10 employees when the round was announced and expects to nearly double headcount by the end of October. The capital will expand the technical team, including forward-deployed engineers who can sit beside agencies when the model meets a live mission.

That may be the sharper investment thesis. National security needs teams close enough to understand why an answer matters, what it permits, and where the machine must hand judgment back to a person.

The hostile network keeps moving while institutions debate the next tool. Aslan now has $20.8M to see whether supervised digital presence can move with it without surrendering the trust required to stay in the room.

Physical Superintelligence

A machine can search more possibilities than a scientist can hold in a lifetime. It still needs a human to decide which question deserves the electricity.

That division of labor sits underneath Physical Superintelligence, a Cambridge public benefit corporation launching with a $58M Seed led by Breakthrough Energy Ventures. A broad group of deep-tech and individual investors joined the round.

CEO and co-founder Matt Pines is building the institution. President and co-founder Alex Klokus carries the operating mandate. Co-founder Alex Wissner-Gross, Chief Scientist and Chief of Strategy, brings the physics thesis. CTO Alessandro Morari brings systems experience from NVIDIA, IBM Research, and U.S. national laboratories.

Their premise has nerve: physics discovery can become a resourceable process. Humans choose the mission and define success. PSI's "virtual physicists" decompose problems into hypotheses, run them in parallel, and force results through conservation laws, formal proof, simulation, digital twins, and physical measurement.

Reality gets the final vote, refreshing in a market where the model usually gets another prompt.

The first commercial proving ground is AI infrastructure. PSI is productizing Emmy to optimize power, cooling, networking, compute, materials, and geometry inside terrestrial and orbital data centers. Changing one constraint pushes on the others, so design still moves through slow human iterations.

PSI is betting a machine can search that design space faster without grading its own homework. A plausible answer can survive a demo. A thermal design has to survive the heat.

The company has one unusually patient proof point. The Fermi Explorer Mission says PSI devised the perihelion-pump trajectory for a proposed Alpha Centauri spacecraft within its mass and budget limits. The mission is planned, and the trip is measured in tens of thousands of years. Startup timelines have met their match.

The Seed will fund recruiting, Emmy, verification infrastructure, data-center productization, and research across the energy-to-compute stack. PSI reports more than 25 researchers and engineers, but no customers, revenue, valuation, or audited commercial performance.

The useful question stays inside the lab: can machine-scale search reduce physical uncertainty faster than experiments, facilities, and customer trust can absorb the result? PSI is betting verified physics can make the physical world less stubborn, one campaign at a time.

Locus Robotics

Warehouse autonomy stops being a software story when the robot meets a soft bag, an irregular package, or the one SKU nobody designed the demo around.

That physical exception is where Locus Robotics is placing its next bet. The company raised $41.6M from existing investors as part of a Series G, CEO Rick Faulk told Axios. An SEC Form D filed August 24 records the full $41,599,804 equity offering sold to 18 investors, with $0 remaining. The filing names no lead investor, valuation, or specific use of proceeds.

Locus began with warehouse operators Bruce Welty and Mike Johnson confronting a problem created when Amazon bought Kiva Systems and outside customers needed another robotics path. Mike Johnson remains President and COO. Rick Faulk is CEO, and Sean Johnson is CTO. The company learned the market from inside the building, where a technically elegant machine still has to survive changing order volume, labor gaps, tight aisles, peak season, and inventory that refuses to behave like a product sheet.

The original Locus model paired autonomous mobile robots with people to reduce walking and keep work moving. LocusONE became the orchestration layer across robots, tasks, demand, and inventory. The company reported more than 6B cumulative picks by October 2025, with the latest billion completed in 24 weeks, across more than 350 sites serving 150-plus brands. Those figures are company-reported, but show the operating surface investors are backing.

Now Locus Robotics is pushing from assisted fulfillment toward more autonomous work. Locus Array moves picking, putaway, induction, drop-off, slotting, and replenishment into the aisle. The May acquisition of Nexera Robotics added a compliant gripper designed for the awkward reality of soft goods, porous materials, irregular packages, and varied weights. Navigation gets a robot to the shelf. Manipulation determines how much of the shelf it can own.

CB Insights identifies G2 Venture Partners, GS Growth, and Scale Venture Partners as Series G participants. No lead has been publicly confirmed. The more important capital signal is that existing backers are funding Locus while the company expands the boundary between collaborative robots and autonomous fulfillment.

Warehouse operators are buying flexibility under pressure. The next layer of value will belong to systems that can absorb a new SKU, another peak, and a different building without turning every exception into a person walking over to rescue the machine.

N-Power Medicine

Cancer research has a geography problem.

The therapies may be designed in global laboratories, but many patients eligible to help prove them are treated in community oncology practices far from the academic centers where trials concentrate. The science can move quickly while the research system asks a patient to travel farther, wait longer, or never enter the study at all.

N-Power Medicine is building around that gap. The Redwood City company closed a $32M Series B with new strategic investor Labcorp Venture Fund and continued participation from Merck Global Health Innovation Fund, Innovatus Capital Partners, and an unnamed U.S.-based biotech investor.

Co-founder and CEO Mark Lee and co-founder and clinical-science leader Christer Svedman are attacking a stubborn design problem: how to create trial-grade comparison data inside routine care without pretending retrospective records can answer every question a randomized control arm was built to settle.

N-Power's answer is ProECA, a prospective external control arm created before and alongside a study. Instead of searching backward through whatever reached the chart, the program aligns patient eligibility, assessments, imaging, sequencing, biosamples, and follow-up with a defined protocol while people receive standard care.

Drug developers may be able to enroll smaller control groups and make earlier development decisions. Community oncologists gain research infrastructure. Patients gain another path into the evidence system without leaving the practices where they are already being treated.

N-Power says its first ProECA, focused on non-small cell lung cancer, is operating at more than 40 sites. Pharma and biotech partners are using it in early-development and Phase 4 programs. Colorectal and prostate cancer programs are planned later this year.

The caution belongs inside the opportunity. FDA guidance and a recent Journal of Clinical Oncology paper make clear that external controls require validation, traceability, endpoint alignment, and honest treatment of residual confounding. Prospective collection can improve the evidence without erasing the variables nobody measured.

N-Power now has capital to make that model operational across more clinics and indications. Its future will be written in the handoff between protocol and practice, where a cleaner trial design meets the oncology visit that still has to happen on time, close to home, and with enough trust for a patient to say yes.

AIR

An AI agent can inherit more power from one plugin than a new employee gets in a month. The uncomfortable part is that nobody may have checked who wrote it, what it calls, or whether its instructions changed overnight.

Yair Saban and Niv Hoffman built Air around that handoff. The Tel Aviv cybersecurity company emerged from stealth with $50M across 2 Seed rounds closed within weeks. Sequoia Capital led the first $10M. Greenoaks led the next $40M, with Swish, Netz, Zach Frankel, Yinon Costica, Ofir Ehrlich, Anne Neuberger, Omer Adam, Varun Anand, and other angels participating.

The founders came from Israel's Unit 8200. Their product targets an enterprise-adoption problem: agents now connect to skills, plugins, MCP servers, websites, internal data, and sub-agents. Each connection expands what an agent can do and adds another place where trust can expire.

Air calls its answer a context firewall. Air Control discovers and governs sanctioned and shadow agents. Air Filter evaluates add-ons before installation. Air Defend watches actions at runtime. Air Marketplace supplies vetted components. Approval becomes continuous because a package, repository, domain, or maintainer can change after yesterday's scanner said everything looked fine.

The evidence is early and company-reported. Air says its research found 17,822 open-source skills tied to untrusted external instruction sources, representing 6.7M installations. Another experiment showed how one skill reached more than 26,000 agents after ordinary reputation signals and scanners cleared it. Air says its platform now filters roughly 27% of the add-ons and skills it evaluates.

Air reports more than 20 customers, about one-quarter large enterprises, with demand strongest in finance and pharma. The capital is expected to support security research and go-to-market expansion in the U.S. and Europe. Ryan Knisley, formerly CISO at Disney and Costco, joined as Chief Strategy Officer.

The strategic question is who owns trust when agents assemble part of their own operating environment. Model providers will add controls, security teams will add policy, and developers will keep reaching for components that make agents useful before procurement has learned their names.

Air is betting that the durable layer sits between those forces, deciding what may enter an agent's context and tracing where a compromised component already traveled. That record changes every time an add-on updates, a maintainer changes, or a new agent quietly joins the fleet.

Delfi

A customer becomes a different kind of product advisor when it also owns part of the company solving its problem.

Members 1st Federal Credit Union has made a strategic investment in Delfi's credit union service organization, or CUSO. The financial terms were not disclosed.

That structure is the interesting part.

Credit unions have bought asset-liability management tools for decades. A CUSO lets them help shape infrastructure around the realities of member-owned institutions, then share in the value it creates.

Delfi is building an AI-native balance-sheet intelligence platform. Overwatch models cash flows, rates, liquidity, capital, and other what-if decisions. Delfi Exchange connects those decisions to curated capital-markets opportunities. Agent lets teams question the same economic engine in natural language.

The product argument is speed with continuity. A team can move from a balance-sheet question, to a modeled consequence, to an execution option without rebuilding the analysis in separate systems.

Members 1st brings operating weight to that argument. The Pennsylvania credit union serves more than 630,000 members and holds about $8B in assets. Chief Risk Officer Stuart Bretz says the goal is more proactive risk identification and stronger oversight while protecting resilience, intergenerational equity, and service to members.

For Daniel P. Ahn and Joseph Ahn, PhD, the investment also arrives with evidence that the CUSO model is gaining traction. Delfi says it has facilitated $100M in platform transactions since launching the CUSO with One Washington Financial and Maps Credit Union earlier this year. That figure is company-reported, and the announcement does not break out revenue, customer count, or transaction economics.

Balance-sheet management is becoming less like a quarterly reporting exercise and more like a live decision loop. Rates, deposits, liquidity, and capital can move before the next committee packet is assembled. The institution that can model those changes quickly still has to connect the model to governance, judgment, and execution.

That is why a customer-investor can be unusually useful. Members 1st can pressure-test Delfi against an $8B balance sheet, and Delfi can turn those lessons into infrastructure smaller institutions cannot build alone.

The next proof is not another feature. It is whether better decisions show up in resilience, earnings, and capital that can keep moving back toward members and communities.

Ki 13

Synthetic fuel can win the policy argument and still lose at the loading dock. If hydrogen and clean carbon represent 60% to 80% of production cost, the final molecule arrives expensive before the plant has done much else.

Ki 13 has raised a $5M Seed round to attack those inputs together. HICO Investment Group led the financing, with BURGEST 2007 SL, Triple Impact Ventures, GiTV, and Desai Ventures participating, alongside non-dilutive match funding from Innovate UK.

Formerly Ki Hydrogen, the company was founded in 2022 by Koji Muto, Michael Stanton, and Carl Banbury through the Carbon13 Venture Builder. Banbury stepped down as director and COO this summer. Muto remains co-founder and CEO, while Ki 13's website lists Stanton as co-founder and chief science officer.

The new name adds hydrogen's atomic mass of 1 to carbon's 12. Ki 13 is building around both.

Its low-temperature process starts with lignocellulosic residues. A catalyst mediator breaks down the biomass into biogenic CO₂ and hydrogen ions. The spent catalyst then moves through an electrolyser, where it is regenerated while renewable electricity produces hydrogen.

That chemistry matters because conventional water electrolysis spends energy producing oxygen that Ki 13 does not need. The company reports lab performance of about 25 kWh per kilogram of hydrogen, compared with roughly 50 kWh for water electrolysis, plus 0.3 MWh per tonne of biogenic CO₂. Those are company-reported benchmarks, not yet industrial operating results.

The $5M fully funds Ki 13's first industrial pilot in West London. The company plans 5 tonnes of annual hydrogen capacity and a TRL 6 milestone in 2027 before moving toward a commercial demonstration plant.

Commercial interest has arrived ahead of that hardware. Ki 13 says it holds letters of intent representing more than $15M in potential annual revenue from European e-fuel producers and has biomass-supply agreements supporting 5 years of scale-up. LOIs are useful pressure, but they are not recognized revenue.

That is the interesting handoff. Sustainable aviation fuel, e-methanol, and e-methane all need hydrogen and a carbon source. Ki 13 is trying to produce both from one abundant residue stream while using materially less electricity.

North Acton is about to become the place where those lab numbers encounter pumps, fouling, maintenance, biomass variability, and every unglamorous bill that decides whether clean fuel becomes infrastructure.

Cari

Money can move at internet speed. Bank governance still has to show up on time.

That tension sits inside Cari's $32.5M first tranche, the company's initial external funding round. The capital came entirely from banks, including all 6 institutions that helped design the network: First Horizon Bank, Huntington Bank, KeyBank, M&T Bank, Old National Bank, and SouthState Bank. Glacier Bank and other unnamed banks also invested, while Keefe, Bruyette & Woods advised Cari.

Cari's design partners began working with the company in September 2025, shaping the technology, operations, governance, and use cases before funding the next phase. These customers understood the integration burden because they had already been carrying part of it.

Founder and CEO Gene Ludwig built Cari around a choice regional and community banks cannot avoid. Digital money is becoming programmable and always on. Rebuilding the blockchain, wallet, controls, portal, and core integration at every bank would turn modernization into a recurring invoice.

Cari offers shared infrastructure instead. Its permissioned network lets chartered banks represent deposits as dollar-denominated tokens, move them on-chain, and preserve the regulated relationship. That matters for institutions that compete on trust but rarely have the budget to invent a payment network alone. Cofounder and COO Matthew (Louie) Giacomini and cofounder, CPO and CMO Phoebe Kunitomi are helping build the operating network.

Cari says it launched an MVP on March 31 and delivered a full product suite on July 31. Pilot banks can test the mint, transfer, and burn cycle for a tokenized deposit. More than 30 banks have joined, another 40 are in active discussions, and that network plus pipeline represents institutions with more than $10T in combined assets. Those are reported participation metrics, not production payment volume.

The capital will support onboarding, integration, more programmable-money use cases, and the operating capabilities required to reach production. Cari now has an investor group that knows exactly where the difficult work lives because those banks helped specify it.

Customers becoming owners can tighten the feedback loop. It can also remove the luxury of confusing enthusiasm with deployment. Cari's next proof will arrive when shared governance, bank-grade controls, and always-on settlement survive the same production environment that made the network worth building together.

TabaPay

An instant payout can take 3 seconds on screen and years of banking relationships underneath it. The customer sees speed. The operator carries the sponsorship, network rules, fraud controls, settlement paths, reconciliation, and backup plan that make the speed believable.

TabaPay has closed a $155M strategic growth financing led by FTV Capital while announcing plans to acquire Transact Bank, N.A. The financing contains primary capital and a secondary transaction. TabaPay did not disclose the split, valuation, sellers, or ownership terms, so the headline is not a clean measure of fresh operating cash.

The acquisition would put an OCC-chartered, FDIC-insured bank beside TabaPay's payment platform under a newly registered holding company, TabaHoldings. After closing, Transact Bank is expected to become TabaBank, N.A. The Q4 2026 target remains subject to regulatory approval.

Payment infrastructure gets complicated precisely when it starts working. Fintechs and platforms need multiple sponsor banks and rails for different use cases and failure modes. Each relationship adds resilience, then hands the operator another contract, integration, exception queue, and reconciliation file.

TabaPay sits in the middle of that machinery. Its single API supports payments and payouts across card and bank rails through more than 20 partner banks. The company reports service touching 1 in 3 American households and more than $100B in expected 2026 volume.

Those company-reported markers explain the choice. TabaBank is designed to complement the partner network, add redundancy, and support digital banking, debt repayment, RTP, FedNow, ACH, wires, and major card networks. TabaPay also expects the financing to advance merchant-liquidity products and future acquisitions.

Rodney Robinson is TabaPay's co-founder and CEO. The current leadership page lists co-founder Manoj Verma as CRO and co-founder Marvin Mah as CTO. FTV partner Robert Anderson joined the board.

Owning a bank changes who carries more of the handoff. TabaPay will still need partner banks, regulators will decide whether the acquisition closes, and customers will judge the system during a declined payment, delayed payout, fraud spike, or broken settlement file.

If TabaBank reaches launch, the valuable space between institutions could become smaller without the safety net becoming thinner. That is where TabaPay's next $100B of movement begins to look less like volume and more like control.

Typewriter Therapeutics

CAR T can look like medicine and operate like custom manufacturing. Cells leave the patient for engineering, testing, and reinfusion while disease keeps its own calendar.

Typewriter Therapeutics has emerged from stealth with a $56M Series A to challenge that sequence. AN Venture Partners and RA Capital Management led the round, with ANRI, Gemseki, and SBI US Gateway Fund participating.

The company is developing an all-RNA, non-viral system designed to insert a full therapeutic gene into a specific genomic site. The first targets are in vivo CAR T medicines and genetic liver diseases. Typewriter says its initial nonhuman-primate studies are planned for late 2026.

The ambition is easy to describe and brutal to earn: turn a patient’s own T cells into the manufacturing site.

Today’s autologous CAR T process can deliver powerful outcomes, but it carries collection, external engineering, quality control, logistics, time, and cost. In vivo approaches try to move that work inside the patient. That could widen access only if delivery reaches the right cells, the inserted gene behaves as intended, and safety holds up beyond a mouse model.

Typewriter’s platform is built on the R2 retrotransposon. A lipid nanoparticle delivers 2 RNA components: an mRNA that produces the R2 protein and another RNA carrying the therapeutic gene. The complex is designed to insert that gene at a conserved site in ribosomal DNA, then degrade.

That differs from transient RNA expression, small DNA edits, or semi-random viral insertion. It remains preclinical. Typewriter says the financing will extend humanized-mouse proof of concept and support its first in vivo CAR T development candidate.

The people tell you what kind of problem the investors bought into. Matthew Stanton, PhD, became CEO after senior RNA and delivery roles at Generation Bio and Moderna. Leanne Peiser, DPhil, joined as CSO after more than 20 years across immunology and cell therapy. Founder Jun Zhou, PhD, leads discovery and global operations.

Haruhiko Fujiwara, PhD, and Tom Eickbush, PhD, spent decades defining R2 biology and its gene-insertion mechanism. Typewriter was founded in 2022 to carry that work toward medicines, with teams in Cambridge and Tokyo.

The $56M now buys Typewriter a harder page: prove that precise gene writing can survive delivery, repeat dosing, biological variability, and the evidence standard required for patients. If that page turns, cell therapy may begin leaving the factory before the patient arrives.

Thyme Care

Cancer care breaks in the handoffs between the people treating the disease, the organizations paying the bill, and the patient expected to keep the whole story straight.

That is the fracture Thyme Care has spent 6 years turning into an operating model. The company has now closed more than $125M in Series E funding at a valuation above $2B, with Morgan Health leading and Humana, CVS Health Ventures, AlleyCorp, HealthQuest Capital, Foresite Capital, Concord Health Partners, Frist Cressey Ventures, Town Hall Ventures, and a16z Bio + Health participating.

The syndicate matters because it looks less like a venture roster and more like the system Thyme Care has to coordinate. Payers, employers, providers, and institutional investors are backing a company whose value depends on getting those groups to share accountability before a missed symptom becomes an emergency visit or an unaffordable therapy becomes somebody else's problem.

Thyme Care says its services are available to more than 10.5M people across all 50 states, with more than $7B in oncology spend under management. The company reports 5% to 10% validated reductions in total cost of care, profitability, and positive free cash flow. Fierce Healthcare reports the active treated population reached 135,000 in August, up from 10,000 at the end of 2024.

Those are company-reported operating results, but they explain why this round is doing more than extending runway. Robin Shah is becoming executive chairman of a new parent organization, Thyme Companies, while Brad Diephuis continues as Thyme Care CEO alongside co-founder, President, and Chief Medical Officer Bobby Green.

Thyme Companies is designed to build independent oncology businesses around problems the core navigation model keeps encountering. The first areas under development are lower-cost biosimilar adoption and clinical-trial accrual. One business is expected to launch later this year.

That move changes the wager. Navigation sits between appointments and helps patients, practices, and plans keep care connected. A portfolio model goes after the underlying markets creating the gaps: how drugs are purchased, how trials find patients, and how financial incentives travel across oncology.

Thyme Care has made the space between appointments commercially visible. Thyme Companies now has to prove that the same perspective can reach upstream into drug economics and research access without losing the patient in another handoff.

HiddenLayer

The dangerous part of an AI coding agent is not that it can write code. It is that the same agent can read a repository, call tools, touch secrets, run commands, and ship the result before a human understands the chain of decisions.

HiddenLayer has raised a $100M Series B led by Delta-v Capital, with Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Ventures participating. The Austin company plans to deepen its enterprise AI security platform, expand Agentic Runtime Security, and build out Agent Harness Security for coding agents.

Chris “Tito” Sestito, HiddenLayer’s co-founder and CEO, started the company in 2022 with Tanner Burns and Jim Ballard. Jacob Rideout serves as CTO.

The timing has changed around them. In 2023, HiddenLayer’s $50M Series A arrived while enterprises were still debating whether AI-specific attacks would become a large enough problem to support a category. Now agents are moving through software-development and business workflows with permission to act, while the instruction that compromises them can arrive several tool calls into a task.

Traditional controls still matter. Identity, endpoint, cloud, application, and data security do not disappear because an agent joined the workflow. The new gap is behavioral context: whether this action is appropriate for this agent, in this session, with this tool, at this moment.

HiddenLayer is building around that gap. Its platform spans AI discovery, supply-chain scanning, attack simulation, and runtime protection. Agentic Runtime Security reconstructs multi-step sessions so teams can inspect behavior across calls and stop malicious tool use or unauthorized actions while the agent operates.

The company says ARR grew more than 10x and that it added more than 50 platform customers. Chris Sestito told TechCrunch that ARR reached the tens of millions. Those company-reported figures explain why growth capital is arriving now.

They also raise the standard. Security buyers will want fewer alerts, clear evidence, policy control, and response that keeps pace with machines acting at machine speed. Platform companies will keep adding native guardrails. Independent vendors must prove they can see across models, tools, and infrastructure without becoming another console the CISO has to reconcile.

HiddenLayer began by protecting models. This round funds a harder assignment: staying beside the agent while the agent is doing real work, where the next security boundary is being written one tool call at a time.

Polymarket

Polymarket’s newest financing says something sharp about the present: investors value the machinery for pricing uncertainty as aggressively as the companies trying to remove it.

Polymarket is raising $1B at a reported $21B post-money valuation, according to Bloomberg and lead investor 1789 Capital. The firm is expected to invest about $300M after putting roughly $200M into the company previously. That is a formidable vote on a platform turning elections, rates, sports, culture, and corporate milestones into tradeable probabilities.

The language matters. This is a reported financing in progress, not a company-announced close. No full syndicate, ownership terms, use-of-funds breakdown, or completed-round confirmation has been disclosed.

Founder and CEO Shayne Coplan started Polymarket in 2020. The platform matches users peer to peer through an order book, with outcome shares backed and settled onchain. Underneath the price sit liquidity, data distribution, resolution, surveillance, compliance, and the plumbing every serious exchange meets.

That plumbing has become the story. Intercontinental Exchange, owner of the New York Stock Exchange, announced an investment arrangement of up to $2B in 2025 and completed another $600M investment in March 2026. ICE also agreed to distribute Polymarket’s event-driven data to institutional clients. The bet now reaches beyond consumer speculation toward a new information layer for finance.

Polymarket’s U.S. route runs through QCX, which operates as Polymarket US and is listed by the CFTC as a designated contract market. That status does not erase the fights around prediction markets. It raises the operating burden. Market integrity, surveillance, resolution rules, and customer trust have to scale together.

The $21B valuation makes that burden impossible to hide. Rival Kalshi was recently valued at $22B, while states, regulators, sports leagues, and lawmakers keep arguing over where event contracts end and gambling begins. Capital has decided the category is large. The institutions around it have not finished deciding what the category is.

Prediction markets are moving from crypto curiosity toward exchange infrastructure, media signal, institutional data, and political object at the same time. Most companies would prefer one identity. Polymarket is being financed as if it can carry all four.

A market can price an outcome. This financing prices the right to host the argument, and the harder work begins when everyone wants that price to count.