The Experiments Nobody Remembers

Ares

The most expensive commitment in private equity may be the one a manager makes to its own fund.

LPs supply the capital, but GPs still put their own balance sheets behind new vehicles. They also finance new strategies and succession plans before exits or distributions can replenish the firm.

Ares Management raised about $4.2B for its inaugural Ares Global Structured Solutions Fund and affiliated vehicles, more than 4 times the $1B target. The October 1 close gives Ares a dedicated pool for those decisions.

AGSS is secondaries capital moving upstream. It is designed to help private-market managers increase commitments to their own funds, launch strategies through structured LP commitments, and manage ownership succession. Bloomberg reports the vehicle will primarily use preferred equity.

That customer differs from the one many people picture when they hear “secondaries.” An LP-led trade helps an investor sell a fund interest. A continuation vehicle offers liquidity while giving a sponsor more time with selected assets. AGSS can enter earlier, while the manager is deciding what it can put behind the next fund, product, or generation of leadership.

Nate Walton, Ares Partner and Head of Private Equity Secondaries, has spent more than a decade building around that pressure. Ares says its funds have deployed nearly $9B in structured solutions since 2013. Its Secondaries Group managed $44B as of June 30, 2026, inside a platform with more than $671B under management.

The market can support this specialization. Jefferies estimated global secondary volume reached a record $240B in 2025, including $115B of GP-led activity. Capital that once bought LP positions or extended ownership is becoming part of private markets' operating architecture.

The close leaves important terms undisclosed: the LPs, the split between the fund and affiliated vehicles, and the return and fee terms. It still shows demand was materially larger than Ares first targeted.

Blair Jacobson, Co-President of Ares, described AGSS as an expansion of the firm's ability to back managers' growth initiatives. Here, growth can mean a larger GP commitment, a new strategy that needs an anchor, or a succession plan that preserves the franchise without forcing the next generation to finance yesterday's ownership structure alone.

Ares is betting that the next chapter of secondaries will be written inside the management companies deciding which funds get built, who leads them, and how much of their own balance sheet follows.

KKR / Gen II Fund Services

Private markets run on ambition, leverage, and work that has to reconcile to the penny.

KKR's agreement to acquire Gen II Fund Services for a $5.1B enterprise value puts a public price on that invisible operating layer.

Gen II sits behind more than 275 investment managers representing over $2T in private-fund capital. Its people and systems handle the accounting, tax, compliance, treasury, investor communication, and data work that turns a legal structure into something managers and limited partners can trust.

That trust gets harder as managers add strategies, jurisdictions, asset classes, and investor groups. Every expansion creates more capital calls, allocations, valuations, reports, regulatory obligations, and exceptions that have to land correctly.

Steven Millner, Steven Alecia, and Norman Leben founded Gen II in 2009. General Atlantic and Hg invested in 2020. Gen II now says it supports over $2T from 13+ offices and has quadrupled revenue and EBITDA through organic growth and 4 acquisitions.

The technology matters because scale is a terrible product if the answers arrive slowly or change when somebody asks twice. Gen II's Sensr suite connects subscriptions, investor portals, analytics, data feeds, and purpose-built AI to private-market records and calculation engines. The software still depends on experienced operators, controls, and client relationships.

KKR has agreed to acquire Gen II through its Core Private Equity strategy from Hg, General Atlantic, and minority investors. Steven Millner and the leadership team will stay. KKR plans to back international expansion, broader services, proprietary technology, AI-enabled tools, and employee ownership.

That ownership plan deserves attention. Fund administration is relationship infrastructure. Clients remember who answered when a report broke or a capital event refused to fit the template. Giving employees a stake can align the people carrying those relationships with growth, although the economics have not been disclosed.

The deal is expected to close in 2027, subject to regulatory approvals and customary conditions. KKR has not disclosed the equity value, financing mix, valuation multiple, or exact close date.

Private capital keeps getting larger, more global, and more complicated. The firms that keep its records exact, its investors informed, and its obligations moving are becoming assets in their own right. Gen II built the machinery in the room nobody photographs. KKR just put $5.1B of enterprise value around it.

Autopoiesis Sciences

Months disappear inside the decision to abandon a promising research path. If the failure reaches a paper at all, the judgment behind it may be reduced to one clean paragraph.

That disappearing process is the business case behind Autopoiesis Sciences.

The San Francisco AI research company raised $11M in Seed funding led by Threshold Ventures. Informed Ventures, Alpaca VC, VitalStage Ventures, Correlation Ventures, DarkMode Ventures, and several angels also participated.

Co-founders Joseph Reth, Dr. Eike Gerhardt, and Dr. Lawrence Callahan are building around an awkward truth for scientific AI: code has tests, math has proofs, and frontier science often has no answer key. A model can sound brilliant while following the wrong question with immaculate grammar.

Autopoiesis first built Aristotle, an AI research partner for literature, hypothesis generation, analysis, and experimental design. The company says thousands of researchers across institutions including Pfizer, the FDA, Harvard, GSK, Merck, MIT, Stanford, and the NIH use it. That is company-reported usage, not proof that every institution is a customer.

The interactions also create a record traditional papers usually discard: hypotheses rejected, evidence challenged, searches abandoned, and the moment a scientist decides to change direction.

Autopoiesis used that record to develop Metacognitive Steering, an inference-time controller that shifts a model among exploration, methodical execution, and critical reassessment without changing its weights. Columbus-1 puts that control inside a long-horizon system with planning, tools, and persistent memory.

The team's September arXiv preprint reports 8 independently reproduced, attacker-reachable BlueZ vulnerabilities and work on a 10-foot rocket intended for propulsive landing. Those are company-authored preprint results, not a peer-reviewed victory lap across science.

The $11M gives the team room to hire, expand its research, and push the system beyond controlled demonstrations. It also raises the standard of proof. Scientific software earns trust by showing its work, surviving independent scrutiny, and knowing when confidence is less useful than another question.

Autopoiesis is trying to preserve the part of discovery that disappears before publication. If that judgment becomes teachable, the most valuable output may not be a faster answer. It may be a research system that remembers why the attractive answer was abandoned.

Satlyt

A ground station is a place to discover that most of a satellite’s data was never worth sending home.

That is the operating problem behind Satlyt’s $8M Seed, led by non sibi ventures. Satlyt’s software runs AI and other workloads on computers aboard satellites, letting operators filter, analyze, and diagnose data before paying to move every raw byte to Earth.

Founder and CEO Rama Afullo saw the gap after working at Tesla, Google Cloud, and SpaceX’s Starlink. Satlyt is not building spacecraft or waiting for a giant orbital data center. It is making existing hardware more programmable.

The first proof is already in orbit. In a Google DeepMind case study, Satlyt ran a quantized Gemma 3 model onboard a satellite to examine logs, software errors, and stack traces. In 2 representative tests, the model reduced diagnostic payloads by 64.4% and 64.8% while still producing a root-cause diagnosis and recommended action.

That matters because bandwidth, power, thermal headroom, and contact windows are all rationed in space. A spacecraft that can decide what is useful before transmission can shorten the distance between an anomaly and a response, while freeing downlink capacity for data customers actually want.

CTO Nelson Kigen Psenjen brings distributed-computing depth. Satlyt’s longer ambition is harder: coordinate compute across spacecraft operated by different companies, creating a shared software layer instead of another closed constellation.

The syndicate also includes TLCOM, Antler, Slauson & Co., Launch Africa Ventures, Enza Capital, Askya Investment Partners, Demos Capital, BAG Collective, Gaingels, and AXIAN Investment. Satlyt says the money will expand engineering and customer delivery, support more spacecraft, and push its virtual AI data-center network forward.

SpaceX’s Transporter-18 carried Satlyt software on a TakeMe2Space spacecraft alongside research and commercial applications. The launch succeeded. Running those applications in orbit, then coordinating real work across multiple satellites, remains the execution job.

There is a clean business proposition hiding inside all that orbital ambition: make the hardware customers already own more productive. Satellite operators get another use for idle compute. Software companies get a route to orbit without becoming spacecraft manufacturers.

If Satlyt can make that handoff routine across operators, the company will not need to own the sky. It can get paid for helping the machines already there decide what deserves to come home.

Autoheal

An engineer reviewing a pull request from an AI agent is approving more than code. They are approving the prompt, tools, model choice, context, permissions, and hidden habits that produced it.

That burden gets heavier as coding agents generate more work. The commit arrives faster. The review comments, CI failures, production incidents, security fixes, and token bills still land on humans who have to reconstruct what happened.

Autoheal has raised a $7.9M Seed round led by Innovation Endeavors, with Emergent Ventures, U&I Ventures, Darkmode Ventures, Batch Ventures, and Param Hansa Values participating. Harpinder Singh is joining the board.

Co-founders Sid Choudhury, Utkarsh Ohm, and Puneet Saraswat are building a self-improving software factory for the post-coding SDLC. The useful part is the mechanism underneath it.

Autoheal connects coding agents, repositories, CI/CD, observability, cloud environments, and issue trackers into a shared engineering context. Worker agents handle incident investigation, vulnerability remediation, release readiness, support escalation, and coding-cost analysis.

Then 2 supervisory agents inspect the inspectors. An Evaluator scores runs against review comments, failed builds, and production incidents. A Healer proposes changes to skills, prompts, tools, or model selection, checks them against historical benchmarks, and opens a pull request for an engineer to approve.

Enterprises do not need an agent that becomes more autonomous by becoming less accountable. Autoheal is betting improvement can be version-controlled, regression-tested, budgeted, audited, and kept inside the customer's security boundary.

The company names Nomura Bank and AvidXchange among its customers. Both say incident investigations can move from hours to minutes. These are company-supplied accounts, not independent benchmarks, but they point to the correct buying question: does the system return trusted engineering time?

Innovation Endeavors is backing an infrastructure layer, not another coding demo. Autoheal told VentureBeat it began selling 3 months ago with 13 engineers across Silicon Valley and Bengaluru. The round will support enterprise rollouts, go-to-market growth, platform reliability, and customer support.

The commercial opportunity sits in a handoff most AI coding stories skip. Generating code is one kind of leverage. Proving that agents can keep learning without quietly teaching themselves the wrong lesson is the work that earns production trust.

KymaThera

PI3Kα has already produced approved medicines. That success made the next design problem harder to ignore.

Block too much of the wild-type enzyme and the biology that validates the target can narrow the dose through toxicity, including hyperglycemia. Cover too little of the mutant landscape and patients with important helical-domain mutations may still sit outside the strongest response window.

KymaThera has raised an $80M Series B to move a specific answer into the clinic. Alta Partners led the financing, with Venrock, Foresite Capital, new investor J. Wood Capital and others participating. The round brings total capital raised above $100M.

The answer is K-1728, an oral, pan-mutant selective PI3Kα inhibitor designed to hit disease-driving kinase and helical domain mutations while sparing wild-type PI3Kα.

That sounds like one sentence of medicinal chemistry. It is actually the whole wager.

KymaThera says K-1728 produced tumor regressions at low, once-daily doses across both mutation classes in preclinical models, with a wide exposure window before hyperglycemia-associated levels. Those findings are preclinical. The $80M is financing the move from an elegant profile in models to evidence in patients.

Founder and CEO Rob Kania spent more than 20 years at Pfizer and later led research at Kinnate Biopharma. Founder and CSO Jason Cox came through Merck and Kinnate. Co-founders Brooke Grandinetti and Shawn Ouyang add discovery biology and medicinal chemistry experience from Vertex, Novartis, Amgen and ImClone.

K-1728 will initially pursue HR-positive/HER2-negative breast cancer as a monotherapy and in combinations, plus PI3Kα-driven vascular malformations as a monotherapy. The disease settings are different. The molecular challenge travels between them: inhibit the mutations driving disease without leaning so hard on normal PI3Kα biology that tolerability writes the dosing strategy.

Alta's Bob More said the investment began with the people. Foresite's Michael Rome pointed to a wholly owned asset. Venrock's Mariana Mihalusova highlighted what the team built in roughly 2 years.

The syndicate is underwriting experience, but the clinical program still has to translate the design. KymaThera expects Phase 1 patient dosing to begin in Q4 2026, with the financing intended to carry K-1728 through initial clinical proof-of-concept.

From there, the record will be written by dose, tolerability, mutation coverage and whether the window seen in models survives contact with human biology.

Chiplytics

A reel of 50,000 chips can look uniform enough to trust and expensive enough to inspect poorly.

That is the pressure Chiplytics is building around. When a manufacturer is one missing component away from a line-down event, the authorized channel may not have the part. Independent supply can keep production moving, but it also asks how much evidence belongs between thousands of identical packages and a critical system.

Chiplytics has secured a $4.5M package to move that evidence closer to the purchase order. GeekWire reports the total includes $2.5M in Seed financing led by Scout Ventures and Waypoint Ventures, plus a $2M Air Force Tactical Funding Increase award. QK Special Services, Industrious Ventures, Interwoven Ventures, Alliance of Angels and Cornell University's Red Bear Angels also participated.

The split matters. Private capital is funding the commercial operation, while the Air Force award advances the inspection technology behind it. Together, they support a new lab inside Quality King's 600,000-square-foot distribution center in Bellport, New York, plus hiring in testing and engineering.

Steven Dourmashkin, co-founder and CEO, leads product, software and operations. Stephen DiBartolomeo, co-founder and President, brings Navy cryptology and national-security experience to sales, procurement and capital. Connor Dempsey, PhD, leads engineering, quality and R&D as CTO, while Devon Richman, PhD, runs the lab as VP of Operations.

Chiplytics combines Sandia National Laboratories' Power Spectrum Analysis with machine vision and supplemental methods to compare electrical signatures, check package markings and preserve traceable results. Its Verified by Chiplytics platform connects sourcing, test planning, reporting and a transparent verification fee.

That combination changes the commercial handoff. The part can move through procurement, verification and delivery inside one chain of vendors, records and accountability, without asking trust to fill the gaps between them.

The capital still leaves hard work in front of the company. Chiplytics has to turn national-lab science and government validation into repeatable throughput for manufacturers whose tolerance for false confidence is much lower than their tolerance for another software demo.

The Bellport lab puts that argument on the floor, next to the inventory and the clock. Each verified reel gives procurement one more way to keep a line moving while evidence follows the component through delivery.

Vitalize

Two nurses call out before sunrise. By 7 a.m., the staffing plan has been rebuilt a dozen times while the census moves, the ED boards patients, overtime rules tighten, and a spreadsheet negotiates with a group text.

That is hospital labor in its natural habitat. The clinician may exist, but the system cannot move that person to the right unit fast enough.

Vitalize has raised a $31M Series A led by Oak HC/FT, with Norwest, .406 Ventures, Constellation Ventures, Y Combinator, Rock Health Capital, and MemorialCare Innovation Fund participating.

Co-founder and CEO Veeraj Shah and co-founder and COO Sanketh Andhavarapu built Vitalize after working inside staffing offices, command centers, and nursing units. They watched leaders reconcile demand, qualifications, union rules, payroll, and callouts across systems that never shared one live plan.

Vitalize connects those signals. Its platform forecasts demand, models staffing supply, builds and rebalances schedules, supports shift pickup and swaps, and pushes approved decisions into existing hospital systems. A forecast that cannot survive local policy, payroll rules, or a sudden acuity change is still only a dashboard.

Oak HC/FT says Vitalize serves 9 health systems across 37 hospitals and roughly 40,000 covered staff. The company reports millions of shifts managed weekly and more than 2,000 clinical-leader hours saved daily. Rochester Regional Health reported 50% less manual staffing time. St. Luke's Health reported a 54% reduction in premium labor spend after 12 weeks.

Those are customer and company reports, not an independent audit. They explain why this round has weight. The American Hospital Association says hospital workforce spending exceeded $1T in 2025. A small deployment improvement can touch overtime, agency costs, manager time, available beds, patient flow, and the workload that follows nurses home.

The money will expand Vitalize's growth and deployment teams and deepen its forecasting, recommendation, and patient-throughput models. The larger bet puts staffing and capacity in one decision loop. A bed without a qualified clinician is an accounting entry. A clinician placed after demand shifts is expensive motion.

Vitalize must make that loop work across more hospitals, service lines, and local rules without losing the speed early customers describe. The next proof arrives before sunrise, when another plan breaks and the hospital has minutes to rebuild it.

Ledgebrook

Insurance underwriters make long-term risk decisions inside an industry that loves short-term production targets.

Ledgebrook made a different choice. Fortune reports that the company uses salary and equity instead of individual bonuses or premium quotas. That changes the pressure surrounding the person deciding whether a difficult risk belongs on the books.

Now Ledgebrook has raised $200M in primary equity financing, co-led by Allianz X and Rockefeller Capital Management. Existing and new investors also participated, though the announcement did not name them. Fortune reports a $2.6B post-money valuation, citing a source close to the company.

Allianz Re separately agreed to a multi-year reinsurance relationship. The amount was not disclosed. Together, the transactions pair equity for the platform, technology, and team with reinsurance supporting the risk-moving machinery around them.

Founder and CEO Gage Caligaris, FCAS built Ledgebrook for the excess and surplus market, where wholesale brokers bring custom risks that do not fit neatly inside admitted-carrier rules. The company says it is on track to surpass $1B in cumulative written premium since it began writing in 2023.

Blackbird reads submissions, classifies risk, and calculates a technical price. CTO Nathan Hall leads the technology organization, but an experienced underwriter makes the final decision. That distinction is the product strategy.

Specialty insurance gets hard because speed and judgment have to improve together. A fast answer that misunderstands the risk is future claims activity wearing a stopwatch.

Ledgebrook now has roughly 300 people, including about 80 underwriters and 50 engineers. Its 2 rated carriers received A- financial strength ratings from AM Best in August. The company is building the balance-sheet, carrier, and technology infrastructure to keep more of the insurance relationship in-house.

The $200M will extend Blackbird and fund technology and talent. The operating obligation is to preserve underwriting discipline while adding products, people, premium, and capital relationships.

That is where the equity, reinsurance, and compensation model meet. Ledgebrook is asking software to make underwriters faster while its incentives ask them to remain patient with risk. If those systems keep agreeing as volume grows, brokers will feel the difference before the market finds another label for the company.

General Medicine

Ashwin Muralidharan changed his title while the company was working.

General Medicine had launched nationwide, signed up hundreds of thousands of customers by its own count, and built a catalog of more than 2,900 healthcare products and services. Then the founding CEO decided TJ Parker should move from venture into the CEO seat full time, while Muralidharan shifted to Chief Product Officer and kept leading daily business. Elliot Cohen remains President.

General Medicine has raised a $120M Series B led by Andreessen Horowitz, with Matrix, VXI Capital, Eli Lilly and Company, Mercy Health through Granger Management, and BoxGroup participating. The financing brings total disclosed funding to $152M.

The company's argument begins where healthcare AI gets less glamorous. Advice can arrive in seconds. Acting on it still means finding the right service, gathering records, checking insurance, understanding the cash price, scheduling a clinician, and carrying context into whatever happens next.

General Medicine is building the transaction and fulfillment layer beneath that journey. Customers can shop directly, talk with a licensed clinician, or use AI chat to explore a need. Its common product catalog reconciles the different language used by patients, clinicians, insurers, labs, and health systems so care can be priced, compared, matched, and fulfilled through one storefront.

The early numbers are promising and company-reported: an average NPS of 80, 83% of customers addressing a need they had postponed, and one-third getting care delayed for more than 6 months. The average customer record spans 11 years when it reaches a clinician as a prioritized summary.

The investor mix matches the ambition. Matrix backed the company early, while Eli Lilly and Mercy Health sit inside medicine and physical care delivery. Their participation does not establish a commercial network, but it places capital near many of the handoffs General Medicine wants to simplify.

The new funding will expand the store into more categories of care. That creates more definitions to reconcile, prices to explain, clinical pathways to coordinate, and moments where software must hand work to a real person without losing the thread.

TJ Parker, Ashwin Muralidharan, and Elliot Cohen are organizing the company around the part of healthcare that begins after somebody knows what they need. The $120M gives them room to make that handoff feel ordinary, even when the machinery underneath it is anything but.

Parallel Systems

The most expensive mile in freight may be the one a railroad decides is too short to run.

Long trains spread crews, terminals and locomotives across a lot of cargo. That efficiency can make smaller loads wait for scale. Trucks collect the business by leaving more often, even when rail track already connects the places a container needs to go.

Parallel Systems has raised a $100M Series C led by AVP | Advance Venture Partners to attack that gap. Hillspire, Agility Global and Cobalt Capital joined Anthos Capital, Congruent Ventures, Riot Capital and Collaborative Fund. Parallel says total funding now exceeds $200M.

Matt Soule, John Howard and Ben Stabler founded Parallel in 2020 after SpaceX. Soule, now CEO, spent 13 years in avionics. Autonomous rail is not software looking for a vehicle. Hardware, batteries, perception, braking, train control, railroad rules and field service must agree before a customer lets the system carry freight.

Parallel's Panther vehicles are self-propelled, battery-electric platforms built for standard intermodal containers. They can move in smaller platoons and integrate with existing railroad systems, giving operators a path to more frequent service without assembling a conventional long train.

The Series C will scale third-generation Panther production, support commercial deployment and fund international expansion. Agility joined days after announcing a collaboration with Parallel across the GCC and wider EMEA.

The Federal Railroad Administration approved a limited test program on 160 miles of Georgia Central and Heart of Georgia railroad infrastructure operated by Genesee & Wyoming. Parallel calls the work a commercial deployment. FRA's authorization remains a supervised test framework, not a national permission slip.

Parallel says major railroads are under contract and previously reported a backlog above 300 vehicles. It has not named those customers, disclosed contract values, reported revenue or published a production schedule. The $100M gives manufacturing discipline, railroad integration and operating reliability room to become the product.

That is the wager behind the round: whether smaller rail movements can win freight that defaults to the road. The next evidence will be Panthers built, routes approved, containers delivered and customers willing to plan around the service.

Parallel is trying to make existing steel useful before the shipment gets tired of waiting and takes the highway.

Healthleap

The chart can know a patient is in trouble hours before the care team has time to assemble the evidence.

Labs live in one field. Weight changes live in another. Poor appetite or trouble swallowing may hide inside a note. The patient is in the building while the people who could act are working through more information than a shift can hold.

That operating gap is where Healthleap AI built its company.

Co-founders Josiah Meyer and Jemima Meyer started with malnutrition, a condition Jemima had watched clinicians miss as a dietitian. Healthleap turned her first tool into an EHR-integrated platform that reads structured data and clinical notes, then writes a risk score into the care team's workflow.

Healthleap has announced $38M across two rounds: an $8M Seed co-led by Sequoia Capital and First Round Capital, and a $30M Series A led by Hummingbird Ventures. Including its 2022 pre-seed, disclosed funding totals $39.1M.

The financing follows a year in which Healthleap says it expanded from 3 hospital partners to more than 50 and grew revenue by more than 10x. Customers include Penn Medicine, Cedars-Sinai, Intermountain, Houston Methodist and Emory Healthcare.

Clinical AI earns trust differently than a product demo. A 2025 peer-reviewed study covered more than 166,000 admissions over 3.75 years. Healthleap reported 88% higher sensitivity than the modified Malnutrition Screening Tool and identification an average of 4 days earlier.

At one Penn Medicine hospital, Healthleap reports $23.8M in annualized financial impact: $6.3M from reimbursement and $17.5M from shorter stays. The company says Penn Medicine's Strategic Decision Support team validated the analysis.

Healthleap prioritizes patients for human review rather than diagnosing them. Hospitals are buying a way to direct scarce clinical attention toward patients most likely to need it, inside workflows people already use.

The capital will expand engineering, product, sales and customer success as Healthleap builds for more conditions. Malnutrition has the strongest published validation today. Delirium, aspiration pneumonia, pressure injuries and heart-failure readmission risk widen the opportunity and the burden of proof.

Hummingbird Ventures, Sequoia Capital and First Round Capital are funding the move from one validated clinical wedge toward a broader inpatient risk layer. Its value will arrive patient by patient, in the distance between what the chart knew and how quickly somebody acted.