When the Noise Takes the Room

REGENT

Erebor Bank is financing half of REGENT's $240M Series B. That $120M debt commitment has a physical job: buy tooling and equipment for a 255,000-square-foot Rhode Island factory that now has to turn Seaglider engineering into repeatable production.

Mare Liberum and AE Ventures co-led the equity half. DCVC joined, while Founders Fund, Caffeinated Capital, Lockheed Martin Ventures, Japan Airlines, and Giant Step Capital returned. That lineup spans maritime strategy, aerospace, defense, aviation, venture capital, and industrial banking.

Billy Thalheimer and Michael Klinker founded REGENT around a vehicle that behaves differently at each stage of its trip. A Seaglider floats at the dock, rises onto hydrofoils, then flies in ground effect above the water. Viceroy is designed to carry 12 passengers at speeds up to 180 mph across routes up to 180 miles. Squire carries the architecture into autonomous defense missions.

The vehicle's modes are sequential. The company's obligations are arriving at once.

REGENT says its commercial order book exceeds $10B across 6 continents, multiple years of factory capacity are booked, and its U.S. Marine Corps contract has expanded to $15M. Those figures are company-reported demand and contracted work, not recognized revenue. Viceroy's first full-scale human flight and commercial certification remain ahead.

That makes the 50-50 financing structure unusually legible. Equity can absorb technical, regulatory, and market uncertainty. Debt can help finance equipment with a defined industrial purpose. Neither capital source makes a supplier hit tolerance, a quality system catch a defect, a regulator approve evidence, or a production crew repeat yesterday's result.

REGENT now has more than $340M raised across equity and debt, a facility, customers across 6 continents, and defense demand pulling on the same core technology. It also has 2 clocks running at different speeds. Customers want vehicles. Certification and manufacturing will move only as quickly as the evidence allows.

The founders' work is becoming less cinematic. Billy Thalheimer has to align customers, capital, and the production transition. Michael Klinker has to keep the technology moving through tests while manufacturing decisions harden around it. Engineering cannot be rushed because the factory is expensive.

The next useful proof will arrive as a sequence: Viceroy flies with people aboard, regulators accept the evidence, tooling becomes a production system, and finished vessels leave Rhode Island for customers. REGENT financed every handoff at once. Now each one has to hold.

TIFF Investment Management

The smallest companies can demand the most personal kind of private equity underwriting.

At the lower end of the lower middle market, a polished deck is not enough. The business may have less than $10M in EBITDA, the sponsor may be working outside a traditional fund, and a modest transaction can carry enormous weight for everyone at the table. The investor has to understand the person steering the deal and the company being acquired.

TIFF Investment Management has raised more capital for that job. PE Direct Fund V reached a final close of $300M on July 29, hitting its hard cap after an oversubscribed fundraise. The prior vintage closed at $165M, making the new vehicle roughly 82% larger.

This is TIFF's fifth fund investing alongside independent sponsors in direct transactions, a strategy the firm says it has pursued since 2014. Independent sponsors do not begin with a committed blind pool. They find a company, assemble a plan, line up capital for that transaction, and put their credibility beside the asset.

That structure can open parts of the market where large institutional buyers rarely compete. It also refuses to let underwriting become abstract. TIFF has to judge whether the sponsor can operate, recruit, allocate capital, and handle pressure, then determine whether the underlying company deserves the risk. One weak half can erase the strength of the other.

Carolyn Patton, Managing Director of Private Markets, linked the step-up from $165M to $300M to a deeper independent-sponsor landscape and broader institutional acceptance. Brendon Parry, Head of Private Markets and Deputy CIO, described TIFF's team as purpose-built to underwrite both sides.

TIFF is bringing a larger vehicle into a market that still resists mass production. Fewer institutional buyers at this deal size may mean more reasonable valuations and more room for operational work. Those are TIFF's stated expectations, not a promise about Fund V's returns.

The close shows institutional capital becoming more comfortable with independent sponsors. It does not make every sponsor institutional, every deal scalable, or every small company ready for ownership change. Fund V now has to convert a bigger capital base into the same disciplined sequence: pick the person, test the company, and keep both aligned after the wire clears.

For independent sponsors, Fund V creates a larger potential partner. For LPs, it creates exposure to a part of private equity that can still depend on judgment before machinery. TIFF's advantage will live in the handoff, when a sponsor arrives with one company and the investment team decides whether the people and the business are ready to grow together.

Socure / Fravity AI

Five years after its Series E, Socure reopened the round instead of inventing a new letter.

Reuters reports the extension totals $156M, citing Socure CMO Colton Pond. The valuation moves from $4.5B to $5.2B. Summit Partners led, joined by Goldman Sachs Alternatives, Wells Fargo, Docusign, and others.

Socure says the transaction mixes primary capital with a secondary tender that gives employees liquidity, but it has not disclosed the split. That means the headline amount is larger than the operating capital landing on the balance sheet, even as the deal creates money for expansion and a market for some employee shares.

Socure bought Fravity AI, the agentic fraud, risk, and compliance platform founded by CEO Kedar Samant and CTO Rushik Upadhyay. The purchase price was not disclosed.

Socure was co-founded in 2012 by Johnny Ayers and former CEO Sunil Madhu. Johnny Ayers now leads the company as CEO alongside CTO Arun Kumar. They are folding Fravity into RiskOS as RiskOS_Agents, connecting investigation workflows to Socure's identity graph, models, decisioning, and resolved-case data.

That is a meaningful expansion of the job. Identity verification decides whether a person or business belongs in the transaction. Fraud and compliance teams still have to investigate what triggered the alert, collect evidence, apply policy, explain the outcome, and leave a file that can survive audit or regulatory review.

Fravity is built around that operational handoff. Socure says existing deployments have reduced cost per case by 80%, accelerated resolution 5x, and cut false positives by as much as 70%. Those are company-reported results, not independent benchmarks, and scale will ask harder questions about evidence lineage, model governance, and human control.

Socure enters that test with company-reported Q2 ARR of $364M, up 63% year over year, 133% net dollar retention, 0.01% logo churn, and more than 3,000 customers. The company says the investment will support global expansion and Fravity's native integration.

The employee tender matters here, too. Five years is a long stretch between priced rounds. Liquidity can recognize the people who carried a private company through that interval without pretending every dollar has the same destination.

The valuation is the clean number. The harder work begins where the capital, employee liquidity, identity data, and investigative agents meet inside RiskOS. Banks and enterprises will decide whether that combination closes cases faster while keeping every decision explainable enough for the people who must defend it.

AngelList / Ark

A capital call can be approved in fund software, delivered through an investor portal, funded through a bank, and reconciled somewhere else. Every handoff gives the record another chance to lose the plot.

AngelList is buying the software built to remove those handoffs. On July 16, the private-markets platform acquired Ark, a Boston-founded fund-operations company serving administrators whose clients include more than 500 VC and private equity firms with over $185B in assets on platform.

Deal terms were not disclosed.

Ark gives administrators and managers fund accounting, LP reporting, investor portals, fundraising, allocations, capital calls, distributions, and a general ledger. The configurable software lets administrators keep client relationships without turning every report change into a vendor project.

AngelList brings another set of rails. Its platform already spans fund management, accounting, cap tables, investor records, and a banking and payments network with more than 150,000 accredited investors holding accounts. The company says Ark customers will gain access to that infrastructure in the coming months.

Bill Ward, Ark's co-founder and CEO, built the company around a plain complaint from fund administrators: legacy software was rigid, expensive to change, and too dependent on vendors for work users should be able to handle themselves. Ark became the configurable layer for administrators such as Standish Management, MeritageAG, and Kranz Consulting.

Avlok Kohli is now putting AngelList behind that layer. The announced roadmap connects Ark's software with banking, payments, cap tables, and AI-native workflows across LP onboarding, portfolio monitoring, and reporting.

The AI piece will get the attention. The accounting piece will decide whether it deserves it.

An agent can answer a question quickly only if capital accounts, bank movements, investor commitments, valuations, and reporting logic agree underneath the interface. Private-capital operations are full of records that look simple after somebody has spent hours making them reconcile.

Ark says the same team and service focus remain. AngelList is promising broader product investment and integrated infrastructure. Neither company has disclosed pricing changes, migration terms, or a detailed integration schedule.

That leaves the combined team with work more demanding than announcing an AI roadmap. It has to connect software, money, and trust while hundreds of firms keep calling capital, paying investors, closing books, and expecting every number to arrive with its history intact.

Maglut Heavy Industries

Two rare-earth elements can enter the same column looking nearly identical and leave at different times because one held tighter to a molecule on a resin bead. That small difference is where Maglut Heavy Industries wants to build an American industrial advantage.

Curtis Wu and Azhar Yerzhanova founded the Long Beach company around a difficult idea: borrow purification logic from biotech and apply it to rare-earth separation. Co-founder and CEO Curtis Wu worked in biologics purification at Amgen. Co-founder and COO Azhar Yerzhanova brings investment banking, materials venture capital, and commercialization experience.

Maglut Heavy Industries has emerged from stealth with $3.1M in pre-seed funding from Wave Function, Nova Threshold, and Julian Capital. The company says its ARC-1 process separated real-world feedstock into individual rare-earth oxides at 99.9% purity or higher at pilot scale.

ARC-1 turns a rare-earth compound into a water-based stream and sends it through ligand-coated resin beads. Each element binds with different strength. One exits first, another later, and a mixed feed becomes separate oxide streams.

The chemistry is elegant, but the industrial obligation is less polite. A commercial process also has to deliver recovery, throughput, repeatability, resin life, manageable waste, and cost at scale. Maglut has not publicly disclosed those figures, named customers, or provided a demonstration timetable. Its official site says the next milestone is demonstration-scale deployment.

Separation sits between the mine and finished magnets, motors, semiconductors, and defense systems. DOE found China controlled almost 90% of global rare-earth separation. DoD has been building toward a domestic mine-to-magnet supply chain capable of supporting U.S. defense requirements by 2027.

A six-person team, according to company-provided reporting, is moving a biotech-inspired method through one of heavy industry's least forgiving transitions. The $3.1M does not settle the cost curve or prove a factory. It gives the team room to make the next experiment larger, more repetitive, and harder to dismiss.

Wave Function, Nova Threshold, and Julian Capital are backing a method before the market can see the operating ledger. Curtis Wu and Azhar Yerzhanova now have to turn different exit times inside a column into reliable output outside the lab, where feedstocks vary, equipment wears, and every percentage point carries a price.

Maglut's proof will arrive run after run, until demonstration-scale data gives industrial buyers something stronger than an elegant mechanism to underwrite.

TerraBlaster

A 2.5-acre grid asks one soil sample to speak for an area large enough to hide both deficiency and waste.

The fertilizer spreader may be able to change its rate across the field. The prescription guiding it still depends on what the measurement found, where it found it, and how much variation disappeared into the average.

TerraBlaster has raised a $12.5M Seed round led by Ulu Ventures to tighten that loop. Participants include Khosla Ventures, Trailhead Capital, Builders Vision, Bidra, Ag Startup Engine, Reservoir, Sunna, Plug and Play, GrainInnovate, and the Hort Innovation Venture Fund. Total company-reported funding is now $16.5M, and Ulu partner Steve Reale is joining the board.

CEO and co-founder Jorge Heraud previously built Blue River Technology, acquired by John Deere in 2017. CTO and co-founder Matt Colgan developed precision-ag sensors and algorithms. Co-founder and board member Dr. Pablo Sobron developed LIBS instruments for space research.

Now that technology is riding behind a tractor.

TerraBlaster's Blaster 1000 opens a shallow trench, fires a laser at the soil, and turns the resulting light into nutrient maps. The system can travel at up to 6 mph, take roughly 20 readings per second, and report six 1-inch layers plus a 0-to-6-inch composite.

The commercial point is less theatrical than the Mars origin story and more useful. TerraBlaster says it can map nutrients at 0.15-acre resolution, roughly 80 feet square, while common grid sampling assigns one value to 2.5 acres. The output uses file formats agronomists already know.

That matters because application precision is wasted when measurement remains blunt. A machine can place fertilizer differently every few dozen feet only if the map gives it a defensible reason to do so.

TerraBlaster estimates up to $45 per acre in combined lab, labor, and yield benefits in certain examples. That figure is company-reported, pricing is unfinished, and named customers remain undisclosed. Those boundaries belong inside the story.

The Seed capital will support R&D, team growth in California and Iowa, and more Midwest field work. Demonstrations and presales are planned for fall 2026, with initial systems targeted for spring 2027. The first commercial product measures nutrients; an integrated measure-and-apply system remains future work.

Farmers already own equipment capable of acting on a sharper prescription. TerraBlaster now has to prove that its map earns the agronomist's trust, survives different soils, and changes enough decisions in time to pay for the pass across the field.

RQD* Clearing

Brian Donnelly built RQD* Clearing around one consequential technical choice: own the code underneath the clearing platform. That gives CEO Michael Sanocki and CTO Jon Fowler more control over how clients see trades, risk and records in real time. Every expansion promise also gives that machinery a larger operating surface to defend.

RQD* Clearing has now secured a $74M minority growth investment led by Bain Capital's Tech Opportunities team, with ABN AMRO Clearing Bank and Nyca Partners participating. The money is aimed at expansion across North America, Asia and the Middle East, plus digital assets, tokenization and custody infrastructure.

A broker-dealer, RIA or foreign financial institution can make U.S. market access feel like software. Underneath it, somebody still has to clear the trade, safeguard the assets, reconcile the books, calculate risk and stay inside regulatory boundaries. The obligation stays stubbornly institutional.

RQD* built its platform around that contradiction. Its proprietary infrastructure combines clearing, custody, execution and risk visibility through direct APIs and real-time processing. CEO Michael Sanocki and CTO Jon Fowler are selling speed without pretending speed removes responsibility.

RQD* reports more than 543M ledger transactions and approximately 515M equity transactions year to date, representing 69.5B shares, nearly $2T in notional value and roughly 2.43% of the NMS equities market. It also reports nearly 64.8M options contracts and $3.93T in options notional value. Those figures are company-reported, not independently audited, but they put operating weight behind the architecture.

DTCC's NSCC moved to 24x5 equity clearing in June, extending the operating week across time zones. The SEC is also working through how tokenized securities fit existing ownership, custody, trading and settlement obligations. More access creates more hours, markets and failure modes for infrastructure to handle.

Bain Capital is backing RQD* after the machinery has processed meaningful volume. ABN AMRO Clearing Bank and Nyca Partners are returning supporters. The capital widens both the company's geography and the surface where performance has to hold.

Digital-asset custody cannot be a slide about modernity. Global access cannot outrun risk controls. Real-time infrastructure earns trust one reconciled account, cleared trade and uneventful operating day at a time. RQD* has made the hidden layer of investing behave more like software. The next phase asks whether it can travel farther, cover more assets and stay quiet when the stakes get louder.

Mubadala Capital / Arrive Logistics

Mubadala Capital is buying control of Arrive Logistics at the moment a freight broker's 2 favorite growth levers, more people and more software, can start pulling in opposite directions.

The alternative asset manager signed a definitive agreement to acquire a majority equity interest in the Austin company. Financial terms were not disclosed, and closing is expected in Q4 2026, subject to customary conditions.

This is a change-of-control investment, not a disclosed funding round with a clean amount for the headline. ATL Partners, Lead Edge Capital, and other existing investors will retain meaningful stakes. Arrive's management team is rolling significant equity into the transaction.

Matt Pyatt and Eric Dunigan founded Arrive Logistics in 2014. Matt Pyatt remains CEO, Eric Dunigan remains President, and the company now reports more than 2,000 employees, 5,500 customers, 10,000 core carriers, 10 North American locations, a $4B run rate, and more than 1.9M projected loads in 2026.

The next chapter is deliberately heavy. Matt Pyatt told FreightWaves that Arrive plans to hire 1,000 people in 2026 while expanding cross-border and additional freight modes. Arrive also plans to invest more in risk management and cargo-theft prevention.

At the same time, Arrive is accelerating ARRIVEnow, its proprietary transportation management system. It automates work across the load lifecycle, giving employees more room for the relationship work software cannot fake.

Freight brokerage still makes a missed handoff somebody's empty dock, late production run, damaged margin, or Friday-night phone call. Arrive's August market update describes softer Q3 demand and accepted truckload volumes roughly 2% to 4% below 2025 through much of August, even as structural capacity constraints keep the market capable of tightening fast.

Mubadala Capital is not buying a clean software multiple wearing a logistics costume. It is buying a people-heavy network whose technology matters only when it helps thousands of employees, customers, and carriers make better decisions under pressure.

The continuing stakes held by management, ATL Partners, and Lead Edge Capital keep that pressure where it belongs. Liquidity is part of the transaction, but so is another round of exposure to the same question: can Arrive add modes, borders, talent, and automation without turning scale into distance from the shipper and carrier?

Arrive spent 12 years building the network. Mubadala Capital now owns the control position behind a harder piece of freight: making every new hire, workflow, and service line arrive as one company.

Levanta

After closing a $22M Series B, Levanta gave eligible employees a chance to take some cash off the table while continuing to participate in the company’s future.

That detail belongs near the top because creator commerce is an industry obsessed with attributing value after a transaction. Ian Brodie, Rob Schab, and Spencer McKenney are applying the same question inside Levanta: who helped create the value, and how should the economics reach them?

Volition Capital led the round after leading Levanta’s $20M Series A in 2024. Long Run Capital also participated, according to Axios. Levanta says the new financing takes total funding above $43M.

The product started with Amazon sellers trying to recruit creators, track sales, and pay commissions without running the operation in spreadsheets. It now connects brands, publishers, influencers, and marketplaces across Amazon, Walmart, and Shopify.

This sounds like affiliate software until the transaction leaves a clean line. A creator can introduce the product on TikTok or YouTube, a shopper can finish on a marketplace, and a seller can still be left asking which relationship produced incremental demand. Every system wants a receipt. Influence rarely arrives with one.

Levanta puts discovery, product sampling, flat-fee collaborations, commission management, payments, and cross-channel measurement into the same workflow. The company reports more than 90,000 vetted creators, millions more discoverable through AI-powered Social Discovery, and 80% year-over-year revenue growth in 2026. Those are Levanta’s numbers, not independently audited results.

The Series B will fund technology and go-to-market work while Levanta brings dozens of additional retail marketplaces onto the platform. That expansion is more difficult than adding logos to a landing page. Each marketplace carries different catalogs, identities, transactions, commission rules, returns, and payout records.

Volition is financing a category argument. Retail media monetizes the demand already inside a retailer’s properties. Levanta’s Creator Commerce Media thesis is that marketplaces can also fund and measure demand created across the open internet.

The opportunity gets larger with every seller and creator brought into the same network. So does the obligation. Attribution can reward the people who moved the sale, or quietly become the story each platform tells itself.

The next integration will test both sides: whether Levanta can make creator demand easier to buy, and whether the economics remain credible when the shopper, seller, creator, and transaction begin in different places.

Cyclic Materials

Rare-earth recycling has two industrial problems at once: recover the material, then keep discarded magnets moving into the plants that process it. Cyclic Materials is financing both sides of that handoff.

Cyclic Materials completed a $75M strategic financing led by accounts advised by T. Rowe Price Associates, with ERI joining as a new strategic investor. The company reports $237M in total equity funding.

This is the company's second $75M financing of 2026. January brought an oversubscribed Series C for wider expansion. The new round has a narrower operating job: move faster on U.S. rare-earth infrastructure in Arizona and South Carolina.

Founder and CEO Ahmad Ghahreman and CTO Alex Forstner are scaling a Hub-and-Spoke system. MagCycle mechanically separates magnet material from end-of-life products and recovers copper, aluminum and ferrous material. Refining uses hydrometallurgy and solvent extraction to turn concentrate and production scrap into rare-earth oxides for permanent-magnet supply chains.

The chemistry can work and the plant can stand ready, but neither creates feedstock by itself. ERI's July partnership brings a national electronics collection and preprocessing network into the system. ERI says its eight U.S. recycling centers process more than 1M pounds of e-waste per day. Its move from strategic partner to investor puts that upstream flow inside the financing story.

The capital is expected to support Cyclic Materials' Mesa, Arizona plant and an integrated campus planned for McBee, South Carolina. The company expects South Carolina groundbreaking in Q4 2026. Earlier plans called for annual capacity of 2,000 tonnes of magnet material and 600 tonnes of mixed rare-earth oxides, with operations targeted for 2028. Those are buildout targets, not operating results.

The timing has become political as well as commercial. A July 30 White House determination classified end-of-life rare-earth permanent magnets, manufacturing swarf and other critical-mineral-bearing scrap as resources necessary to national defense. That policy does not settle plant economics or customer qualification. It makes the waste stream Cyclic Materials is chasing part of an explicit U.S. supply-security agenda.

The next proof arrives through ordinary industrial details: qualified feedstock, commissioned equipment, stable yields, customer acceptance and material moving back into manufacturing. Cyclic Materials has more capital and a partner that can help fill the front end of the system. The work now passes from the cap table into Arizona receiving bays and a South Carolina construction site.

Quaise Energy

A drill bit becomes a consumable long before rock reaches 500°C. The deeper and hotter the well, the more conventional drilling charges for every foot in time and equipment.

That physical problem is where Quaise Energy is putting a $180M Series B.

Carlos Araque and Matt Houde founded Quaise around millimeter wave drilling developed from MIT research. A gyrotron sends high-power energy through a waveguide to ablate hard rock, opening a possible route into temperatures and depths where mechanical tools become painfully expensive.

The final close brings the full round to $180M. Prelude Ventures led the $134M first close in July, joined by JERA and Idemitsu Kosan. Nabors Industries supplied $35M at the final close and says it now owns 14% of Quaise on a fully diluted basis. Quaise reports $280M in total funding.

The accounting matters. This is not another $180M arriving in August. The final close added $46M to July's first close; Quaise has not identified who supplied the $11M beyond Nabors' portion.

The more interesting part sits outside the cap table. A Nabors PACE-B rig is drilling at Project Obsidian in Oregon. The companies' framework covers a dedicated rig, reservoir modeling, well design, technology integration, commercial development, and an exclusive drilling-services relationship.

That makes Nabors an investor, operator, equipment partner, and potential beneficiary if the platform scales. Quaise is pulling the machinery, crews, data, and incentives into the same hole.

For investors, that arrangement puts capital beside the contractor that will see where the drilling system loses time, wears hardware, and needs redesign. The feedback loop becomes part of the asset.

Project Obsidian's first phase targets 50 MW and first grid power by 2030. Quaise says later phases could add up to 1 GW, while the drilling system advances toward depths beyond 5 km.

DOE says superhot resources above 375°C can deliver several times the power density of cooler rock.

Quaise reported more than 100 meters drilled through granite in 2025 and said in July that its Texas field program was approaching 1 km. Commercial superhot wells still have to prove long-duration materials, reservoir behavior, repeatable drilling, project economics, and a schedule that survives contact with the subsurface.

The round gives Carlos Araque and Matt Houde a rare combination: patient climate capital, strategic energy investors, and a drilling contractor with steel already on the ground. Project Obsidian now has to turn that alignment into heat, flow, turbines, and electricity that arrives when the grid asks for it.

ProFound Therapeutics

A blood-pressure cuff can reveal preeclampsia after 20 weeks of pregnancy. It cannot explain which hidden protein pushed the biology there, whether it can become a warning signal, or whether medicine can reach it in time.

ProFound Therapeutics has received an investment commitment of up to $35M from the Gates Foundation, including an initial $20M investment, to search that gap. The job: find novel proteins in placental and serum samples, connect them to preeclampsia and eclampsia, and turn the strongest evidence into biomarker and therapeutic hypotheses.

The World Health Organization estimates that preeclampsia affects 3%-8% of women who give birth worldwide. It can progress to eclampsia and life-threatening seizures. Hypertensive disorders accounted for about 16% of maternal deaths globally in 2023, roughly 42,000 lives.

John Lepore, MD and CTO Tom Chittenden are bringing ProFoundry to the problem. Protein detection, high-throughput functional assays, computational methods, and the ProFoundry Atlas are meant to move a candidate from “we found it” toward “we understand what it does.”

A newly detected protein is interesting. A protein reproducibly tied to disease biology, validated across representative samples, and useful as a biomarker, protein drug, or drug target is something a health system can begin to act on. Placental biology still gets the final vote.

The investment also supports a preeclampsia- and eclampsia-specific AI tool. “Agentic AI” is the fashionable phrase. Causal grounding and experimental validation decide whether the system produces another ranked list or a research path scientists can trust.

The Gates Foundation has committed $2.5B through 2030 to women’s-health R&D, citing an analysis that found only about 1% of healthcare research and innovation goes toward female-specific conditions outside oncology. This commitment gives that promise a laboratory: samples, proteins, assays, models, and a disease neglected for generations.

ProFound has not announced a clinical candidate, validated diagnostic, trial, regulatory plan, sample partner, or timetable for the remaining $15M. Those are the handoffs between discovery and a globally accessible product.

Flagship Pioneering initially committed $75M when it unveiled ProFound in 2022. Four years later, the company must narrow tens of thousands of possible proteins into a few findings strong enough to change how preeclampsia is detected, understood, or treated.

For mothers and babies, the biology has never been abstract. ProFound now has a funded obligation to make more of it readable.

Hermetiq

The pull request arrived before lunch. The build failure stayed after dinner.

AI coding agents can produce more code than a team could reasonably ask humans to write. Every change still has to compile, link, test, and survive the build system before it becomes a working product. That handoff is where machine speed meets an infrastructure problem with a human waiting on the other side.

Jake Newfield, CEO and co-founder, and Timothy Potter, CTO and co-founder, built Hermetiq around that delay. The company raised a disclosed $2.2M early-stage round, with Jet Ventures investing $750K as the largest single investor.

Niranjan Tulpule, Jason Calacanis through LAUNCH, Gokul Rajaram, Lighthouse Seed Fund II, Chilli Ventures, and Gi21 Capital also participated. Hermetiq called the financing pre-seed on LinkedIn, while funding databases and the intake record classify it as Seed.

Hermetiq works inside Bazel and remote-execution environments where a failed build can hide across cache behavior, queue pressure, worker infrastructure, tests, cost data, and configuration drift. The platform combines those signals so engineers and coding agents can see what failed, why it failed, and where the next investigation belongs.

That is the current product. Automated repair is the direction.

Hermetiq says an unnamed top-four frontier AI lab runs more than 2M builds per day on the platform. Jet reports more than 3B cumulative build events from that customer and an Etsy pilot. Those are company and investor claims, not independently audited customer proof, but they place this young product inside a workload where “we will check the logs” stops sounding like a plan.

The $2.2M will strengthen engineering and accelerate the move from diagnosis toward repair. That progression sounds obvious until software can change a build inside somebody else's environment. A recommendation can wait for review. An automated fix has to respect permissions, data boundaries, and the possibility that a green build can still be wrong.

Hermetiq supports hosted and customer-managed deployments and describes SSO, RBAC, mTLS, machine authentication, regional data control, and SOC 2 Type I. Those controls matter as build intelligence moves closer to consequential code.

Code generation gets the spectacle. Build systems keep the receipts.

Hermetiq now has to earn enough trust for an agent to read those receipts, make the repair, rerun the affected work, and hand an engineer evidence instead of another mystery. The round buys engineering time for that handoff, while the build loop decides how much autonomy it is ready to carry.

Cloverleaf AI

Cloverleaf AI closed its Series A in January and waited 7 months to announce it. By August, the financing had already become people, product work, and a much larger map of the government buying process.

The Denver company raised $8M in a round led by S3 Ventures, with Jackson Square Ventures, Ivy Ventures, and Tawani Ventures participating. The funding sits beside a stubborn sales problem: a government RFP can arrive late in the actual competition.

Requirements get shaped before procurement opens. Budgets move through committees. Agency leaders reveal priorities in meetings, minutes, notices, and legislation. Sellers who enter when a bid appears may be acting after better-positioned competitors spent months learning the account.

CEO and Co-Founder Adam Zucker and CRO and Co-Founder Jeremy Becker built Cloverleaf AI around that earlier record. The platform monitors 70,000+ public-sector agencies and 200,000+ verified speakers, then configures the resulting intelligence around a seller's territory, accounts, and solution relevance.

The product has expanded beyond meeting video into agendas, minutes, budgets, purchase data, municipal code, and a library of 1M+ verified contacts and speakers. Native CRM and MCP integrations put those signals inside the tools sellers already use.

Government sales is not a conventional intent-data market wearing a municipal badge. An agency does not wake up, click a pricing page, and become an opportunity. Its need can emerge through a capital plan, a public-safety complaint, a committee debate, or a budget line that will not turn into an RFP for months.

Cloverleaf AI says customers such as Uber Transit have seen up to $1M in pipeline per rep per quarter. That is company-reported pipeline, not audited revenue, but it gives the Series A a commercial target: prove earlier public signals can create more timely conversations, not simply more alerts.

The capital has already supported hiring across sales, marketing, business development, and customer success, alongside continued platform investment. That 7-month operating window turns the announcement into a partial receipt. The round did not arrive yesterday. The company has had time to begin showing where it went.

Cloverleaf AI's next pressure sits inside the same promise it sells. Finding a signal earlier only matters if a seller can understand it, enter the account with context, and stay useful through the long public-sector buying cycle. The software can open the door sooner. The relationship still has to survive everything that happens before the contract.

Arga Labs

The model demo ends at the first tool call. From there, an AI agent enters permissions, webhooks, mutable records, and the part of enterprise software where one plausible action can become somebody else's operating incident.

Arga Labs raised a $10M Seed to build that missing practice ground. General Catalyst led the round, with BoxGroup, Emergence Capital, Gradient, and SV Angel participating.

Co-founder and CEO Phillip Li and co-founder and CTO Akira Tong are building real-world sandboxes for agents that use business software. Arga creates stateful twins of tools such as Slack, Salesforce, GitHub, Gmail, Stripe, and Notion, then lets teams test workflows without sending the message, changing the customer record, or moving money in production.

A tool-using agent is judged across a chain, not a single answer. Authentication has to hold. Permissions have to mean something. Webhooks arrive later. Records change during the workflow. A Salesforce lead and a HubSpot contact can describe the same company while looking different enough for an agent to act twice.

The company reports that customers ran more than 100,000 twin instances during the 16 weeks before the announcement. Arga also says it can clone a SaaS application's backend behavior in under 12 hours with 100% fidelity. Those are company-reported claims, not independent audit results, but they reveal the standard the product has chosen to sell against.

The Seed will fund R&D on a harness intended to generate twins in minutes instead of hours, expand Arga's evaluation platform, and support hiring across engineering, research, product, and GTM.

General Catalyst's Yuri Sagalov framed the investment around where agents create economic value: inside business applications. The better an agent becomes at choosing and using tools, the more expensive a realistic test environment becomes to ignore.

Arga still has hard work ahead. Third-party software changes constantly. Buyers will expect security evidence and coverage across the systems their workflows touch. HIPAA and SOC 2 readiness remain in progress, and ArgaBench research has not been released.

The round gives Arga room to turn a young technical product into infrastructure enterprise teams can trust. It also puts a deadline on the promise: generate the world around an agent quickly enough that testing keeps pace with what the agent is allowed to do.

AI agents are moving from answers to actions. Arga Labs is financing the place where those actions can break safely, leave evidence, and return to the model as something it can learn from.

Legato Audio

Glasses are one of the few medical-adjacent objects people put on their face without being asked to hide them. That ordinary habit is the commercial bet behind Legato Audio, Inc.'s emergence from stealth with $12M.

Mehul Trivedi and Steve Romine founded Legato in late 2024 after working on both sides of hearing at Bose. Trivedi helped create Bose Frames and later led smart-eyewear work at EssilorLuxottica. Romine ran Bose's hearing-aid business and later became Audicus COO. Their question was why so many adults who could benefit from hearing assistance still avoid it.

Neotribe Ventures, Listen Ventures, and Village Global financed that thesis. No lead investor, round stage, valuation, or prior-round history was disclosed. The public record supports one clean number: $12M in funding, largely used for product development and marketing ahead of launch.

The product is Legato Frames, eyewear with hearing-assistance technology inside the arms. The company says AI separates voices from background noise, while a dual-speaker system directs sound toward the wearer and limits leakage. Legato reports 4 granted patents and more than 20 pending applications. A federal device listing identifies 4 Frames models as over-the-counter wireless air-conduction hearing aids.

That listing matters, but it is not a customer review. It establishes the product category, not all-day comfort, speech separation in crowded rooms, or whether people will still choose the frames when the novelty becomes routine.

The distribution plan makes the strategy more interesting. Legato expects to sell online and through eye-care providers this fall. It says Frames will cost a fraction of traditional hearing aids and may often qualify for vision insurance, though pricing and coverage remain undisclosed.

The FDA says close to 30M U.S. adults have some degree of hearing loss, yet only about one-fifth of people who could benefit from a hearing aid seek intervention. Legato is treating that gap as an identity and product-design problem alongside the acoustic one. A device can work technically and still lose the customer when they decide what wearing it says about them.

The $12M must carry 3 jobs: perform like hearing technology, wear like eyewear, and sell like a consumer product without shrinking the medical responsibility underneath it. Each has its own buyer, proof standard, and failure mode.

The founders already understand the individual pieces. This launch will show whether those pieces can survive the same day on one person's face, then earn the ordinary second morning when that person reaches for the frames again.

Digs

Builders FirstSource is giving Digs two scarce assets at once: $25.3M and five years inside a national building-supply workflow.

The Series A makes Builders FirstSource the solo lead. The five-year commercial agreement gives Digs' AI platform a harder test than another demo: can plans, selections, approvals, warranties, conversations, and project history move through one build without losing context?

Ryan Fink, CEO and co-founder, and Ty Frackiewicz, CPO and co-founder, started Digs after seeing that a home could be the largest asset a family owns while the record explaining it still lived across paper, PDFs, email, text messages, and somebody's memory. CTO Jason Goetz now carries the technical obligation alongside them.

Digs turns those materials into a searchable source of truth for builders and a 3D digital twin for homeowners. The platform supports takeoffs, diagramming, collaboration, e-signatures, handoff, warranty, and aftercare. One detail can begin as an estimate, become a field decision, and answer a warranty question years later.

Digs raised a $7M pre-seed in 2023, a $7M Seed in 2024, and another $5M in 2025, bringing pre-Series A funding to nearly $20M. The new round brings disclosed financing to about $44.3M. The announcement does not disclose valuation or other Series A participants.

Builders FirstSource brings reach that most construction software startups spend years earning: more than 140,000 customers and approximately 565 locations across 43 states. Digs brings a system designed to structure disconnected construction documents and carry their context from pre-construction through warranty.

A strategic investor can become a distribution channel, product partner, customer gateway, and source of technical gravity at the same time. Integration depth will matter more than the logo on the cap table.

The companies say the partnership will accelerate AI chat, visual coordination, homeowner handoff, maintenance guidance, warranty support, and integration. That language is forward-looking. HousingWire reports company-provided takeoff results, but the proof will come from builder adoption, shipped integrations, cycle-time gains, and fewer mistakes across live projects.

This round finances a handoff between software and supply-chain scale. Builders FirstSource can put Digs closer to the people carrying estimates, materials, schedules, changes, and homeowner promises. Digs has to keep that record coherent as the workflow gets bigger, noisier, and less forgiving. That is the operating test behind the financing.

H.I.G. Capital / HBK

The legal seam inside HBK's new investment tells the story: H.I.G. Capital can finance the firm's expansion while licensed CPA partners keep control of the audits and reviews carrying HBK's name.

That is how a 77-year-old accounting firm takes its first institutional capital without pretending professional authority is an ordinary portfolio asset.

An H.I.G. affiliate has agreed to make a strategic growth investment in HBK, including HBK CPAs & Consultants, HBKS Wealth Advisors, and technology advisory firm Vertilocity. The amount, valuation, and ownership percentage were not disclosed. The transaction is expected to close in Q4 2026, subject to customary conditions and regulatory approvals.

Thomas M. Angelo, Managing Principal and CEO of HBK CPAs & Consultants, and Christopher M. Allegretti, CEO of HBKS Wealth Advisors, will continue leading their businesses with HBK's partners. H.I.G. Managing Director Chris Byrne is backing a platform that the company says spans 27 offices across 7 states and India.

The structure matters more than the usual partnership language. Hill, Barth & King LLC will remain owned and controlled by CPA partners and keep the attest work. HBK Advisory Group LLC will house tax, consulting, accounting, and technology services. HBK Sorce Advisory LLC will continue operating HBKS Wealth Advisors, which says it manages more than $7.5B.

That separation gives the capital somewhere to work. HBK says the resources will go toward people, technology, client-service capabilities, and scale. The firm entered 2026 with a Top 50 accounting practice, reported approximately $184M in revenue in March, and kept expanding. Its May combination with Tampa-based BB&K was already its second strategic combination of the year.

H.I.G. brings $75B of capital under management and a playbook built around middle-market growth, recapitalizations, and consolidation. The obvious opportunity is a broader national platform where accounting, tax, advisory, technology, and wealth services meet the same entrepreneur or family at different moments.

The obligation travels with that opportunity. Faster hiring, better systems, and more combinations can make HBK more useful. They also create more interfaces where independence, client continuity, data governance, and professional judgment have to survive the handoff.

Clients were promised the same teams and standard of care. The next record will be written in retention, audit quality, technology adoption, and whether scale lets those teams see more of a client's financial life without turning trust into another integration metric.

OliverAI

An AI agent never waits for a dashboard refresh. Once it gets access to enterprise data, one useful workflow can become a continuous stream of queries, tool calls, follow-up questions, and compute bills.

That machine-speed appetite is the customer Oliver AI is building around.

Co-founders Praneet Sharma, Grace Johnson, and Toby O'Brien announced pre-seed funding from Menlo Ventures and Unusual Ventures to scale OliverDB. The amount and valuation were not disclosed, leaving the technical promise to carry the story.

Most analytical infrastructure was designed for people. An analyst writes a query, reviews an answer, and eventually asks the next question. Agents compress that cycle until the pauses disappear. Performance, compute cost, permissions, and auditability become one operating problem because the same system has to carry all 4.

OliverDB approaches it as an agent-native data layer. Its columnar, cellular engine can run as an embedded binary or a service. Managed and VPC deployments give enterprises different control boundaries, while ingestion from systems including Postgres, Snowflake, ClickHouse, and Databricks lets the existing data estate stay in place.

Agent Flight Control adds the part buyers inspect after the speed demo. Each agent receives a policy that scopes the data and actions it can reach. Queries are rewritten around row and column permissions, tool calls are constrained before execution, and activity remains attributable for review.

The performance claims are substantial, and OliverAI qualifies them. Current internal benchmark notes report up to 162x CPU performance, up to 1000x GPU performance, and 50-300x lower compute use in evaluated workloads. The company says the results are directional, workload-dependent, and not independently audited.

That matters because infrastructure is purchased on somebody's data, hardware, concurrency, security rules, and failure tolerance. A benchmark can earn the meeting. A repeatable workload result earns the architecture decision.

Menlo partner Tim Tully framed agentic AI as a new infrastructure category. John Vrionis of Unusual focused on giving AI meaningful data access without surrendering control or trust. Those are investor theses today. OliverAI still has to make them customer evidence.

The capital gives this young company room to do that work. Every enterprise evaluation now tests whether OliverDB can make agents faster, cheaper, and governable without forcing a rebuild beneath them. The database may be built for machines, but the decision still belongs to people who have to defend the cost and the consequences.

Deep Cogito

A general-purpose model can know plenty and still know almost nothing about the metric that gets a security team paged at 2 a.m. Zscaler wanted its products, data, and outcomes trained into the intelligence itself, which is a more demanding request than adding a thin customization layer after the model arrives.

That customer problem now sits inside Deep Cogito's $43M Series A. TQ Ventures led the round, with Benchmark, Nexus Venture Partners, Atreides Management, South Park Commons, and Zscaler participating. The financing pushes total outside funding above $56M and gives a 2-year-old San Francisco lab more room to hire researchers, scale training infrastructure, and turn its open-weight work into enterprise systems.

Drishan Arora and Dhruv Malrana built Deep Cogito after working on Google's AI Search products. Their thesis puts the decisive work after pre-training, when a broadly capable model must learn how to reason, use tools, follow a customer's objectives, and improve against increasingly difficult tasks.

Deep Cogito calls one research direction Iterated Distillation and Amplification. The model spends more computation searching for a stronger solution, then that improvement is distilled back into its weights. The useful distinction is timing: this is a training loop that can produce a stronger model, not evidence that a deployed model quietly rewrites itself while customers use it.

The team has been building in public. Cogito releases have ranged from 3B to 600B+ parameters, culminating in Cogito v2.1 671B. Deep Cogito reports strong benchmark performance and lower token use against models of similar capability, but those figures remain company-reported and the financing announcement adds no new benchmark table.

That disclosure boundary matters because the enterprise pitch is larger than a leaderboard. A buyer is being asked to put proprietary data, evaluation criteria, and operating outcomes into a post-training process, then trust that the resulting model is more useful for its exact work. Zscaler gives Deep Cogito a real customer relationship, while its participation as a strategic investor means the only named customer in the announcement is also financially aligned with the company.

The next proof will live inside customer-specific evaluations: whether Deep Cogito can improve the tasks companies actually pay to complete, preserve control around proprietary data, and make those gains reproducible across accounts. The $43M buys more experiments, more infrastructure, and more people. Enterprise trust will be earned one specialized model and one measurable outcome at a time.

MiiHealth AI

Clinical AI is moving upstream.

Ambient scribes listen after the appointment starts. MiiHealth AI is building for the conversation before it, when a patient can explain symptoms, history, and goals before the clinician enters.

That timing is the company’s bet. MiiHealth AI closed a Seed led by Russell Glass, former CEO of Headspace and founder and General Partner of Arteria Capital. An August 5 company-sourced announcement and the Phoenix Business Journal reported $2.8M; a later Business Wire version called the amount undisclosed. The discrepancy belongs in the record.

Kelvin Summoogum, founder and CEO, is building DAINA with co-founders John Wall and Jacob Lester. The Dynamic AI Intake and Navigation Agent conducts multilingual voice intake, captures clinical history and red flags, and prepares a structured EHR note before the visit.

The product sounds conversational. The commercial work is institutional.

A useful intake agent has to ask the right specialty-specific questions, map answers into the right fields, move through HL7v2 and other standards, surface risks, preserve an audit trail, and give the provider something worth trusting. If the clinician rebuilds the note, the automation only changed the shape of the admin.

MiiHealth AI reports 300+ patients across 2 Mayo Clinic cardiology workflows. Its cardiology page says DAINA saves 8 minutes per patient on average, 93% find it easy to use, and a 16-patient day can return 2+ hours to the provider. Those are company-reported results, not an independently audited clinical study.

Still, 8 minutes can become a serious operating asset when it survives a full schedule, multiple specialties, different patient populations, and the permissions of a health system. Time matters because it can be returned to clinical judgment and patient conversation. Trust decides whether the workflow survives long enough to earn that return.

The Seed will expand engineering and AI teams, specialty protocols, clinical reasoning and safety checks, EHR integrations, and customer capacity. That list tells the story better than a victory lap could. Every new specialty and system gives DAINA another place to prove that natural conversation can become reliable clinical structure.

MiiHealth AI is entering that work with a focused promise: prepare the chart before the clinician needs it. The record now moves into the integrations, safety decisions, and patient conversations where saved time becomes valuable only when the information arriving with it deserves to be trusted.