July 2026 DD - Sourcing Sweep

Fort is the strength-training wearable the fitness-tracker industry never built - a screenless band from three ex-Tesla engineers that uses automotive-grade motion sensing to auto-log reps, sets, form and recovery with no phone tapping. Backed by Y Combinator (W2026), Patron and angels from OpenAI and Tesla, it's in pre-order with real early press (WIRED, Fitt Insider) ahead of a Q3 2026 ship. Whoop and Oura ignored lifters; this team is applying Cybercab-grade sensor rigor to a large, underserved category.

Shadowfax is rebuilding analytics for the AI era - an auditable, reproducible platform where business analysts orchestrate AI agents to go from raw data to shipped apps without touching Python. The founders are Palantir, Snowflake and Rubrik veterans who've scaled data products past $1B ARR, and they're backed by Khosla Ventures and the Snowflake accelerator. As enterprises struggle to trust AI-generated analysis, a transparency-first, analyst-in-the-loop platform from this pedigree is a credible wedge into a massive market.

Avala is physical-AI training-data infrastructure built by Emal Alwis, the third engineer on Tesla's Autopilot team - a genuinely elite founder-market fit for the embodied-AI data problem. On $4.7M seed it claims real scale (>1B labels/day) across global training centers, though the dataset's 150 headcount appears to count that distributed labeling workforce, not ~31 core FTEs. The category (data labeling) is crowded and margin-thin. Keep for the founder and the physical-AI positioning, but reconcile the headcount and interrogate defensibility.

MP Space applies deep ex-Tesla battery and energy engineering to satellite power systems, shipping integrated Li-ion packs and full EPS for LEO constellations. The 13-hit Tesla pedigree is real and unusually deep for pre-seed, and $2.85M from MFV/Speedinvest keeps it well under cap. The risk is a semi-commoditized hardware category and a thin balance sheet for a manufacturing ambition. Keep as a credible picks-and-shovels space play with elite battery talent.

Atomic is a Tesla-supply-chain-alumni startup (Michael Rossiter, Neal Suidan) building agentic AI planning software, born from the Model 3 spreadsheet chaos. It has a confirmed $3M seed (DVx / Madrona, under cap), a live product, and early customers citing 20-50% inventory-cost cuts. Strong founder-market fit in a crowded incumbent-heavy market. Keep - watch for the next raise Axios flagged and whether it can out-execute o9/Kinaxis.

Astrolab builds the rovers that will carry Artemis astronauts and cargo across the lunar surface - full-scale, tested FLEX/FLIP vehicles now under a NASA Lunar Terrain Vehicle contract worth up to $4.6B. Founder Jaret Matthews (ex-JPL, ex-SpaceX) has assembled a 150-person, SpaceX-heavy hardware team that has already put operational rovers on the ground. It's less a stealth seed bet than a de-risked, government-anchored moon-mobility leader - belongs on a serious space list, if slightly outside a pure early-stage cap.

Chestnut Robotics (formerly TetherIA) builds human-like robotic hands - its open-source, $314, 16-joint Aero Hand won real attention as a cheap alternative to proprietary dexterous hands. The founding team is precisely right: Tesla Optimus mechanical plus a Waymo foundation-model researcher. The open-source, ultra-low-cost strategy makes monetization the central question, and there's no disclosed funding yet. Keep for the on-point pedigree and momentum; press hard on the business model before conviction.

Argo Space is building water-propelled, refuellable orbital tugs and a lunar-water propellant supply chain for in-space logistics - an ambitious ISRU bet from a deep, verified SpaceX propulsion team. At $9.9M across 7 rounds (Crosslink-led Oct-2024 round plus SBIR) it remains just under cap, with no newer raise found - flag it to re-check, since one more round tips it over. The thesis is capital-ferocious and speculative (lunar water is unproven, and a crowded race), so it's the slowest-return profile here. Keep for the elite team and mission; underwrite the very long timeline.

Hex (Hextronics) builds autonomous drone-in-a-box docking stations with rapid battery swapping, seeded with $2M from Founders Fund - verified, well under cap. Founder Curt Lary (Georgia Tech) bootstrapped the first prototype from his family's garage in 2020 and has since enabled 100k+ flights across 25 countries, with an early SpaceX launch engineer on the team. Genuinely engineering-dense and deep-tech, but capital-intensive hardware with a thin post-seed raise history - diligence current traction and burn. Keep for elite validation plus real deep-tech signal.

Durin builds sensor-instrumented, increasingly automated drill rigs for critical-minerals exploration, raising a $3.4M pre-seed led by 8090 Industries with Andreessen Horowitz participating - a16z verified, under cap. Founder Ted Feldmann is a third-generation mining professional (Georgia Tech) marrying robotics to an industry where drilling eats ~70% of exploration capital, backed by a strong deep-tech syndicate (a16z, Lux, 8090, Bedrock) and an ex-SpaceX engineer. Very early and capital-intensive, and deep-tech rather than pure software, but the elite validation and genuine builder DNA are real. Keep.

Kerrigan is building a vendor-agnostic orchestration layer for factory robots. Real 30-person company, actively hiring, with a named pre-seed syndicate. Re-verification: the founders are senior automotive software leaders (infotainment and vehicle OS at Tesla, Rivian, VW), not factory-automation program owners, and the $5B figure is self-reported equipment value touched by their software. Credible team in a crowded lane against Intrinsic and Formant; needs a public customer or demo to justify a higher rating.

Wondercraft is a YC- and Will Ventures-backed AI audio studio from ex-Palantir engineers (Dimi Nikolaou, Youssef Rizk) plus a Spotify/Acast media operator, turning text into podcasts and audiobooks at scale. It has real traction - 30k+ users and marquee podcast customers - on a confirmed $3M seed (under cap). The main risk is heavy reliance on ElevenLabs TTS in a commoditizing category. Solid keep, funding fully verified.

Belli builds end-to-end air-cargo management software for complex airlines, founded by ex-SpaceX/ex-Teleport operators who know the cargo-IT pain firsthand. It has real paying customers (Etihad, SpaceX), a $2.3M seed led by Iterative, and a Palantir Startup Fellowship slot. The auto-scorer over-credited the Palantir fellowship as a $2.5B founder 'exit' - that's noise; the real signal is genuine traction in a sleepy, digitizable vertical. Keep as a solid seed-stage operator play.

Exlens is a quality/inspection operating system for factories built by two ex-Tesla factory-software directors - precisely the people who lived this problem at scale. The product framing (an AI reasoning engine that clusters defects to root cause and drafts the CAPA) is sharp, and the founding pedigree is director-level and verified. The catch: its ~$3M round isn't publicly confirmed and the space is crowded. Keep for the elite domain authority; confirm the round before deep diligence.

T51 is a 2025-vintage, two-person Palantir-mafia startup (Jonah Adler, MIT; Vipul Shekhawat, Harvard, ex-Palantir CTO) building forward-deployed AI decision systems on top of manufacturing/logistics ERPs. The pedigree and lean FDE ethos are exactly the profile that wins in industrial AI, but it is pre-funding, pre-traction, and in a crowded lane. Keep and watch - conviction depends on a real deployment and a named lead investor.

Motives (YC S25) replaces the $20k, two-month consumer focus group with a team of AI agents that runs and analyzes video interviews in a day or two, and it already has real logos - big US/European consumer brands including Unilever's Wild. The founding pair is credible: ex-Palantir CTO Sam Stoll and ex-Google/TrueLayer CEO Sean Conley, both Oxford. Funding is a small, honest pre-seed (Rebel Fund + YC), well under cap. The risk is a crowded AI-research category and an unproven moat, but the early commercial pull and lean team make it a solid watch.

Headroom is a lean, engineering-heavy (4 of 7) team of ex-Palantir operators building an 'AI operating system' that turns the SMB inbox into a CRM/task system. Pedigree is genuine - co-founder Jishai Evers and colleagues are verified ex-Palantir - and the team is refreshingly small and unfunded (do not confuse it with the unrelated ex-Palantir '$12M' raises at Perceptic/Conduct). The concern is a brutally crowded AI-email/SMB category with zero disclosed traction yet. Strong team, unproven wedge: a KEEP to watch for early customer signal.

Bastion is an AI-native compliance and security platform built by an ex-Palantir cybersecurity team (7 years protecting sensitive orgs), with a shipped, purchasable product (AWS Marketplace) and a clean, in-cap $2.8M seed. The forward-deployed 'embedded engineers + AI agents' wedge is a credible differentiator in an otherwise brutal compliance-automation market dominated by Vanta and Drata. Pedigree and product fit are strong; the concerns are heavy competition, a sales-weighted org, and no marquee logos yet. A solid KEEP.

Ryvn (YC) lets B2B SaaS companies deploy into customers' own clouds and on-prem/air-gapped environments with a Vercel-like push-from-GitHub flow - and its two ex-Palantir founders built exactly this infrastructure at scale (10k+ deploys/day across hundreds of environments). The team is razor-lean (4 people), YC-backed with an in-cap ~$500K pre-seed, and the pedigree-to-product fit is as tight as it gets. The risks are scope-vs-headcount and a contested BYOC-deployment market with no public logos yet. Strong founder fit; KEEP pending traction.

Webstudio is a mature open-source Webflow/Framer alternative built by a genuinely credible founder - Oleg Isonen, ex-Webflow engineer and creator of the JSS library - and backed by a notable angel roster (Naval, Masad, Balaji) plus Seedcamp/Tiny.vc on a small, in-cap ~$1.2M pre-seed. It has a real open-source user base and paid hosted tiers, but faces the classic open-core monetization challenge against deep-pocketed incumbents, with ARR undisclosed. Strong technical/community pedigree; the open question is durable revenue. A KEEP.

Teller offers a screen-scraping-free banking-data API (7,000+ institutions) positioned directly against Plaid, seeded with $4M from Founders Fund, Lightspeed and Max Levchin's SciFi VC - an elite fintech syndicate, verified and under cap. Founded 2014 by Stevie Graham, it's run remarkably lean (18 of 25 are engineers) and quiet since its 2020 seed, which reads as either disciplined bootstrapping or a stalled trajectory - diligence current revenue. Off-core-thesis regulated fintech, but genuine engineering density and elite validation. Keep.

anchr builds an AI-native back-office/logistics OS for food distributors, raising $5.8M led by a16z Speedrun with OpenAI-ecosystem angels - the elite-investor claim is verified (via Speedrun, a16z's accelerator, rather than a partner-led fund round). Founded in 2025 by childhood friends Tzar Taraporvala and Smayan Mehra; note the Tesla 'Senior PM' pedigree badge actually belongs to a senior software engineer (an ex-intern), not a founder. Very young and off-core-thesis vertical SaaS, but real engineering density and a credible AI-ecosystem syndicate. Keep, with the pedigree caveat flagged.

AL Advanced Industries is a lean, two-person German team of ex-Tesla security engineers building ruggedized autonomous surveillance (video + radar) for critical infrastructure that works where power and connectivity don't. The Tesla physical-security pedigree is a genuine fit for the problem, and the company is refreshingly unfunded rather than hiding an over-cap round. The concern is pure earliness: no disclosed customers, no capital, and a hardware-heavy category that usually demands both. A watch-and-verify candidate, not a conviction bet.

Viscera (viscerahealth.com) is a 2025 Palantir-alumni startup automating documentation, coding and billing for urology practices - a 'digital revenue operator' with a clean real-time-to-EHR wedge and fast hiring (+90% in 90 days). Co-founder David Starnino has a verified ~4-year Palantir tenure; the 'Jeremy Myers' pedigree carries a name-collision caveat (Palantir FDE vs a same-named urologist MD) worth confirming. Note the separate, unrelated $8M IBD 'Viscera Health' - don't conflate. Narrow but credible; keep and diligence founder identities and funding.

Endeavor is a Craft Ventures-backed ($7M seed, under cap) AI platform automating sales/production/supply-chain workflows for American manufacturers, founded by 22-year-old Wharton grad Sami Senapathy. The Palantir/AWS pedigree is real but college-era, not senior FDE - don't over-index on it. Confirmed funding and a live product are positives; a ~22% headcount drop and a very crowded lane are cautions. Keep, with diligence on burn and founder depth.

Overstand Labs (YC W25) turns a company's scattered customer communications - email, Slack, calls, CRM - into real-time AI insights for expansion and issue-spotting, built by an ex-Palantir BD + PM/eng founding pair with genuine customer-facing pedigree. It's a lean, in-cap YC company (~$1M incl. YC), but with only one engineer, a crowded revenue-intelligence field, some positioning drift across YC launches, and no disclosed logos, execution risk is high. Credible founders in a hard, contested category - a borderline KEEP.

Kyten is bringing aerospace-grade battery packs to the whole aerospace and defense industry, promising qualification and delivery in weeks instead of months. Founder Cooper McBride went from a Tesla intern to building and leading 30-plus person engineering teams at SpaceX (Starlink) and Shield AI, and the team has shipped thousands of packs to low earth orbit. A lean YC W26 company on roughly $500K, with elite hardware pedigree against a real supply-chain bottleneck. Strong.

Beyond Reach Labs makes deployable solar arrays that unfold to football-field scale in orbit, targeting the power shortage facing orbital data centers and stations. CTO Pele Collins ran parachute engineering across 30-plus SpaceX Dragon missions, and CEO Mitchell Fogelson is a CMU/NASA deployable-structures PhD who invented the core truss tech. It just closed a $10M seed led by Interlagos, which sits it right at the top of our thesis band, so a solid keep rather than a low-raise gem.

BloomPath AI is an operational-intelligence layer that watches engineering work in real time and flags bottlenecks and risks without status meetings. Co-founder Mohammad Kotb brings a verified staff engineering role at Waymo plus senior time at Google, and the company has a clean $1.3M pre-seed led by Raed Ventures. The pedigree and low raise fit the thesis; the dev-productivity category is the main knock. Keep.

Klavis AI sells the live, verifiable environments and agentic data that frontier labs need to post-train long-horizon agents, plus Strata, an MCP server that beats the official GitHub and Notion servers on the MCPMark benchmark. Founder Xiangkai Zeng was a senior engineer on Google Gemini at DeepMind and co-authored the Gemini paper. On roughly $500K with HSG and YC behind it, a high-pedigree, low-raise infrastructure bet squarely on thesis. Strong.

Minerva Humanoids is taking on humanoid actuators and controls, one of the genuinely hard problems in embodied AI, from a near-zero raise. Founder Sandor Felber is a former MIT CSAIL humanoid-robotics researcher who interned on Tesla's Optimus program and has the Forbes AI25 nod to match, backed by high-signal scouts including Pareto and Sequoia Scout. Elite young-founder pedigree plus a pre-seed check size. Strong.

Gibran is an ex-Google DeepMind team building foundational AI that performs in low-data regimes, borrowing from complex adaptive systems and starting in drug discovery. The $2.6M Together Fund seed and the DeepMind lineage check out, though the pedigree is spread across the team rather than one marquee role, and it sits a bit off the core US-frontier-lab thesis. A credible keep on strength of team and thesis.

Sisir Radar is an India-based deep-tech company building high-resolution L- and P-band Synthetic Aperture Radar and India's first private SAR satellite (targeted 2026) for defense, mining and environmental monitoring - a genuine heavy-industry, dual-use play. Its real founder is Tapan Misra, an ex-ISRO scientist who worked on India's first SAR mission (RISAT), an elite and correctly-load-bearing pedigree; the Apple-COO name in the raw data is misattributed noise. Funding of about $8.57M (a ~$7M Series A led by 360 ONE in Dec 2025) is verified as near-cap but under the $10M line, and the defense angle is real. The main risk is that SAR satellites are brutally capital-intensive and $8.6M will not go far. Strong fit for the mining/heavy-industry priority.

Increase is a profitable banking-API company built by Stripe's first employee, Darragh Buckley, moving hundreds of billions of dollars a year for fintechs like Ramp, Check and Pipe with direct network and bank access. The team is a dense, verified Stripe alumni bench and the business turned profitable in 2025 with tripling payment volume. The catch is that Increase is deliberately self-funded and has taken no institutional VC, so the real question is not stage or quality but whether Buckley would ever accept outside capital. Elite by every operating metric; access, not merit, is the constraint.

Apata is a Dublin 3D Secure authentication platform built by Niall Hogan and Shekinah Adewumi, the founders who previously built Touchtech Payments and sold it to Stripe in 2019, then spent three years inside Stripe before returning to their exact domain. They bootstrapped for two years and raised roughly EUR 5.5M (well under cap) from Global Founders Capital and Byld. This is a verified repeat-founder team in payments infrastructure - squarely within the Mural Pay precedent - with a strong engineering fraction. The only knock is that 3DS is a mature, standards-bound niche where the edge is execution rather than novelty.

Ramen builds autonomous network-and-compute infrastructure for the 'uncarpeted enterprise' - warehouses, plants, and industrial sites - pairing plug-and-play radios and edge-AI appliances with 15+ agents that run operations and build a live digital twin. It is founded by verified serial deep-tech operators: Sanjay Kasturia (Qualcomm WiFi lead, Airgo/Teranetics exits) and Prashant Kumar (128 Technology to Juniper, Acme Packet to Oracle). Funding is clean and under cap at an $8.4M Cota Capital-led seed, with a genuine 50-person, 66%-engineering org. This is the strongest, most correctly-reported entry in the top ten.

SiLogy is a genuine quant-crossover: co-founder Kay Li spent roughly eight years in high-frequency trading at Hudson River Trading before building an AI verification engineer for chip design. The product is live, on-prem, and YC W24-backed with a lean technical team. The HRT-alumni claim is load-bearing and checks out at founder level. Deep-tech domain plus verified elite-quant pedigree makes this a top pick in the group.

Metis (metis.nyc) is correctly disambiguated from the several unrelated companies sharing the name; it is a finance-focused enterprise-AI platform. CTO co-founder Brian Mahlstedt is a verified ex-SpaceX and ex-Citadel engineer, and a second founder came from Citadel, making this a bona fide quant-crossover. CEO Steve Geskos brings 20-plus years of institutional investing. Very early with no disclosed funding, but an elite team, 100% engineering headcount, and a top-tier finance advisory network make it a standout.

DashFuel is a real, operating petroleum-logistics SaaS with a verified Bridgewater Associates founder-CEO, Miles Moen, whose elite-finance pedigree checks out at founder level. It carries the group's strongest quant signal and a genuine 18-person, 56%-engineering team. Funding sits comfortably under cap (roughly $3.25M-$6.4M depending on source). Strong quant-crossover with real traction, the only caveat being a less deep-tech domain than peers.

Godela is a Stanford-founded, YC X25 AI physics engine with a verified ex-Apple founder (Cinnamon Sipper) and an ex-Intel co-founder. The Apple-alumni claim checks out at founder level. Real hard-science moat, a signed contract displacing Ansys, and early enterprise interest give it credible traction despite zero disclosed outside funding. Strong deep-tech pick, though it is a major-tech rather than quant crossover.

ExploreTech is a Stanford spin-out building critical-mineral exploration software that fuses probabilistic geophysical inversion with AI drill-target planning, led by two verified Stanford PhDs - Alex Miltenberger (geophysics, PhD on probabilistic inverse theory) and Tyler Hall (geology, TomKat fellow in AI subsurface planning) - with textbook founder-market fit for their own research. At about $1.9M raised plus a Stanford grant it is comfortably under cap and genuinely on the mining-tech thesis the operator cares about. The risks are simply stage and scale: a seven-person team facing long, conservative mining sales cycles. Small and early, but exactly the right kind of heavy-industry deep tech. Strong.

Dataland is a YC-backed, ex-Palantir team building forward-deployed AI agents for complex customer support and operations, reportedly reaching multi-million-dollar ARR within about eight months as just two founders while staying profitable. The founder pedigree (Howard Zuo, Palantir 2017-2020, plus a Palantir product co-founder and an FDE bench from Palantir/Applied Intuition) is verified and squarely on-thesis for lean product velocity. The main risk is a very crowded AI-support-agent market and a founder-reported ARR figure that is hard to independently confirm. Strong signal for its stage; worth a close look.

MinersAI is a genuine mineral-exploration data play with real geoscience founders, a Speedinvest-led pre-seed plus Techstars Paris backing, and a platform that standardizes exploration geodata for critical minerals. It surfaced in three independent queries, which matches its authentic category presence, and every funding reading puts it far under cap. Concerns are a very low engineering fraction (2 of 18), inconsistent tracker data on the raise, and no named paying customers. Keep as a true Durin/Realm analogue worth an operator conversation, with eng-depth as the first diligence question.

Voluna is a genuinely deep-tech mineral-exploration play: drone- and rover-deployed neutron activation analysis with a founding team spanning Blue Origin, Astranis, NASA Kennedy nuclear physics, and a fifth-generation Peruvian mining family. It has a live copper-project pilot in Peru, Breakthrough Energy Fellows and Rio Tinto accelerator backing, and Pillar VC on the cap table, all definitively under cap. It stays below STRONG because it is pre-product with a single alpha pilot and the central physics and regulatory risk of mobile neutron sources is unresolved. Beware the wrong-entity trap with Terra AI's Khosla-led round, which belongs to a different company.

Uplift Microhome is an Austin hardware startup building reusable, off-grid ~60-square-foot disaster-relief housing with onboard power and water, founded by a verified ex-SpaceX and ex-Tesla engineering team and validated by an MIT $100K competition win. It is genuinely early and under cap, with real technical pedigree confirmed in independent press. The open question is market: transitional/disaster housing is grant- and government-dependent with uncertain venture-scale economics, and traction beyond the prize is unproven. A credible ground-floor watch, not yet a conviction bet.

Cadstrom is a legitimate, well-funded deep-tech PCB-validation company with a real Sigma Engine and tier-1 seed backers (Bison, Innovation Endeavors). Funding is roughly $7.45M lifetime, comfortably under cap. However the load-bearing Stripe-alumni claim does not check out: the founders are ex-Microsoft Azure Quantum and Synapse hardware people, with no Stripe link. Solid company but it is not a quant-crossover, so it belongs in the group on merit of the product, not pedigree.

Modelcode AI is a credible, well-staffed legacy-code-modernization company led by serial entrepreneur Michael Fertik with a verified ex-Apple co-founder (Antoine Raux). The Jump Trading quant tie is real but comes through a founding-team AI member, Nir Ben Israel, rather than a titular co-founder, so the quant-crossover framing is somewhat generous. A published benchmark and 68% engineering headcount signal genuine rigor. Solid keep, though the space is crowded and the load-bearing quant claim is secondary.

NetAI is a live, technically substantive GNN-based AIOps company led by a verified ex-Google networking scientist (Deepak Kakadia, 85 patents) with real validation via NVIDIA Inception and an MWC 2026 Google Cloud demo. The founder pedigree is genuine deep-tech but not a quant crossover, so this sits correctly in the raising group. The YES-raising signal is founder-attested only and not publicly confirmed, and the 37-headcount-on-$1M ratio warrants diligence. Keep for the tech and validation, but treat the active raise as unverified.

BiaTech is a Tampa energy digital-twin startup whose two load-bearing pedigree claims both check out: founder-CEO Nathaniel Hartwig genuinely held a VP role at SLB New Energy, and Equinor VP Sarah Delille genuinely sits on the board. Both claims are mildly over-framed, since the SLB tenure was about one year and the Equinor VP is a board director, not an operator. Funding is small and substantially grant-based, and all traction metrics are vendor-reported. Keep for further diligence with the next step being confirmation of paying customers and the actual equity round versus grants.

Tetro is a live, very early AI freight-dispatch startup. Founder Ryan Houlihan has a plausible Stanford robotics and logistics background, but the specific Boston Dynamics plus Apple pedigree claim could not be independently confirmed (public sources point to Baton and Ryder instead). It is a logistics/robotics play rather than a true finance quant-crossover, so the group tag is generous. Worth keeping for the technical-founder signal but the load-bearing pedigree claim is only partially verified.

Frame is an early 'payments-plus-compliance OS' (frameOS) for high-compliance merchants, founded in 2023 by serial founder Alaxic Smith with a roughly 20-person team that includes some Stripe alumni as individual contributors. It sits within the acceptable payments-infrastructure lane (Mural Pay precedent) and is under cap, but the founder's track record is consumer/social (Communly, Flora) rather than payments, so the real payments pedigree is on the bench, not in the founder seat. The currentlyRaising = MAYBE signal comes from the Head of Brand, not a founder, so it is low confidence. Keep as a watch-item, but it is clearly weaker than Apata, whose founders are the domain experts.

Mineflow is a genuine YC S24 mining-tech bet: a CMU-AI, ex-Google-Ads-ML solo founder building deposit-shape prediction models that showed order-of-magnitude accuracy gains in early lithium trials. Funding is a confirmed $590K from YC, Decacorn, Karman, and Leonis, far under cap. The risks are concentrated: one founder, roughly two people, no geology domain expertise in-house, and only trial-stage traction in a slow-moving industry. Keep as a real if fragile Durin-style exploration-tech play; a wipeout or a fund-returner with little in between.

Wafer is a legitimate and fast-moving YC S25 GPU-inference startup with a confirmed $4M seed led by Fifty Years and a remarkable angel roster including Jeff Dean and Wojciech Zaremba. But the quant-crossover premise that surfaced it fails: co-founder Steven Arellano's Two Sigma experience was a three-month undergraduate internship, and both founders are recent grads with internship-only track records in a brutally crowded inference market. On the operator's finance-alumni thesis this is a miss; as a pure early YC infra bet it would be a defensible keep. Rated WEAK under the thesis, with a note that the entity may be a rebrand of Herdora.

atdepth is a credible MIT ocean-engineering spinout doing GPU-accelerated ocean digital twins for carbon-removal verification, with a real Deep Sky partnership and an ARPA-E grant north of $2.5M that dwarfs the $50K in our data. But the D.E. Shaw quant-crossover hook fails the founder-level bar: it is one PhD-summer internship in corporate development by the CEO, and no founder has a finance career. With grant-only funding, no equity round, a .org academic profile, and a narrow subsidy-dependent carbon-credit market, this is a science project commercializing rather than a venture-scale quant crossover. Weak, watchable if it ever raises a priced round.

Jetoptera passes both critical checks: it is not over-cap (roughly $3-5M real capital raised; the $8.79M Series A label is a mirage of crowdfunding plus government awards) and it is not a zombie, with 2024-2025 subscale flights, a Paris Air Show engine debut, and active DARPA and USAF work. The Point72 tie is founder-level but backwards: co-founder Denis Dancanet left Jetoptera to run Point72's Cubist, and the operating strength is CEO Andrei Evulet's GE9X pedigree. Eleven years without a full-scale aircraft, chronic slippage, and retail crowdfunding at a self-set $130M valuation cap it at WEAK.

Hstar Space is a four-person Antler-backed pre-seed team pursuing orbital solar reflectors (OASIS) to boost terrestrial solar farms, with a stated side-ambition in reusable heavy rockets. The SpaceX and Tesla pedigree is real but skews toward supply-chain and build roles rather than orbital-systems physics. The category is crowded and scientifically contested, with rivals like Reflect Orbital and Star Catcher already holding $20M to $65M, and a four-person team splitting focus between reflectors and rockets is a serious concentration risk. Under cap and genuinely early, but the thesis and team depth are thin.

Bridge Defense is a national-security services company founded by verified ex-Palantir enterprise leads (Jack Kilcoyne, Davis Booth) and SpecOps veterans, delivering cleared secure-software development and ruggedized tactical compute - but it operates as an acquisitive government-services roll-up, having made a strategic investment in Federated IT and acquired Syntelligent. The 'strategic' funding flag is directionally wrong: Bridge is the acquirer, not a venture-round recipient, and its moat is clearances and contract vehicles rather than scalable product IP. The founder pedigree is genuine, but the model is services and M&A, not a venture-scalable startup. Weak for a venture thesis.

Kubespaces is a solo-founder 'clusterless' Kubernetes platform letting developers self-serve namespaces without managing clusters, founded in mid-2024 with no disclosed funding and a one-person team. The founder, Alessandro Stefouli-Vozza, has a credible infrastructure-operations background (ex-Head of IT Ops at Susquehanna), but this is an ops title rather than a marquee platform-engineering pedigree. The market is savagely crowded (hyperscalers, Vercel, Fly.io, Railway) and there is no visible traction, customers or co-founder bench. Interesting operator, but far too embryonic and undifferentiated to back now.