Home›Theses 196 theses · as of 8 Nov 2026

Theses

Every market I hold a view on, whoever you are: the bet, my call on it, when the window is, how big it gets, what it rests on, and who you could hold or watch to act on it.

  • 196 theses
  • 23 bull
  • 135 bear
  • 13 categories
Filter
Call Window Expressed by

Datacentre & AI compute49 theses
  1. 1Datacentre thermal materials
    • Closing
    • $19B
    • Support 94
     AI accelerator thermal management does not converge on a single winning material through 2030. 3 listed · 24 private
  2. 2Dielectric coolants
    • Bear
    • Shut
    • $2.5B
    • Support 94
     The 3M PFAS exit dislocates the datacentre dielectric-coolant supply 8 listed · 3 private
  3. 3Liquid metal TIM
    • Bear
    • Shut
    • $538M
    • Support 94
     Liquid-metal TIM becomes the default die-level interface on >1kW accelerators 2 listed · 6 private
  4. 4In-package cooling
    • Early
    • $1.1B
    • Support 93
     Cooling inside the AI-accelerator package, sized once across its three routes 6 listed · 6 private
  5. 5Emerging non-volatile memory
    • Shut
    • $4.8B
    • Support 92
     Emerging non-volatile memory (MRAM, FeRAM, RRAM, PCM) displaces embedded flash and SRAM in specific sub-28nm sockets where flash cannot scale and SRAM leaks too much 13 listed · 9 private
  6. 6Probabilistic computers
    • Bear
    • Early
    • $347M
    • Support 89
     Probabilistic computers (p-bits / Ising / thermodynamic samplers) own server-side sampling + optimisation before fault-tolerant QC 4 listed · 6 private
  7. 7Beyond-CMOS devices
    • Bull
    • Reopening
    • $3.6B
    • Support 88
     Beyond-CMOS device layer: memory beachhead, IP-layer capture 5 listed · 5 private
  8. 8Datacentre digital twins
    • Bear
    • Closing
    • $36B
    • Support 87
     A runtime operational digital twin of the datacentre's compute+networking performance becomes a distinct layer (below design-time EDA, beside the orchestrators) for the non-hyperscaler builders who must model a fleet they cannot afford to over-provision 6 listed · 4 private
  9. 9CNT thermal interface
    • Bear
    • Early
    • $18B
    • Support 86
     Carbon nanotube thermal interface materials reach ≥$1B in annual revenue by 2030, up from <$50M in 2024, driven by AI accelerator and HPC qualification wins. 8 listed · 8 private
  10. 10Metal and solder TIM
    • Bear
    • Shut
    • $1.4B
    • Support 86
     The metal TIM family (solder TIM/sTIM, indium/metal foil, and phase-change metal alloy/PCMA) is the qualified, shipping volume die-level (TIM1) path for >1kW AI accelerators today, Indium-led. 3 listed · 3 private
  11. 11Rack power delivery silicon
    • Bear
    • Closing
    • $13B
    • Support 85
     AI rack power density is forcing a re-architecture of board- and rack-level power delivery (48V-plus distribution, vertical power delivery under the accelerator, point-of-load conversion for GPU/HBM rails), creating a power-conversion-silicon lane that is substrate-agnostic (silicon multiphase today, GaN where switching… 7 listed · 6 private
  12. 12Rack-scale design automation
    • Bear
    • Closing
    • $1.9B
    • Support 85
     Rack-scale multiphysics design automation captures a layer above chip-EDA by 2030, but only as an AI-native tool for the non-hyperscaler builders (neoclouds, sovereign-AI, enterprise AI factories) the incumbents underserve 5 listed · 2 private
  13. 13GaN power for datacentres
    • Bull
    • Closing
    • $1.1B
    • Support 85
     GaN power electronics captures 30% of AI datacentre PSU and 48V conversion by 2028 8 listed · 8 private
  14. 14Datacentre thermal and power
    • Closing
    • $453B
    • Support 81
     Differentiated thermal materials and power semis collect the per-rack tax; integration commoditises 5 listed · 5 private
  15. 15Optical DRAM
    • Early
    • $247B
    • Support 80
     Magneto-optical is the one optical bit that scales; the moat is a cornerable integrated stack 1 listed · 1 private
  16. 16Scale-up interconnect
    • Closing
    • $162B
    • Support 80
     The scale-up fabric is the vendor-agnostic toll - merchant silicon plus the NVLink tax 7 listed · 7 private
  17. 17Small models
    • Bear
    • Shut
    • $18B
    • Support 80
     Applied AI fragments from a few large frontier models toward many small, specialised, on-device models; this proliferation, not bigger models, is the binding demand driver for sub-watt edge-inference silicon 8 listed · 6 private
  18. 18Computational CXL memory
    • Bear
    • Closing
    • $12B
    • Support 80
     Computational CXL memory (near-data processing pushed INTO the pooled-memory tier: KV-cache management, preprocessing, vector search running on cores embedded in CXL memory devices, the XCENA and Dnotitia pattern) is a distinct lane from the CXL pooling/switching silicon already recorded as… 7 listed · 6 private
  19. 19Memory disaggregation and storage
    • Bear
    • Shut
    • $7.2B
    • Support 80
     CXL is a standard the incumbents wrote, so the fabric closes by acquisition to Astera, Marvell and the memory makers; controllers are taken, and the layer underneath is the one to watch. 17 listed · 12 private
  20. 20Low-power edge compute
    • Bear
    • Closing
    • $7B
    • Support 80
     Device-side AI inference does not become a merchant chip category. 26 listed · 90 private
  21. 21Optical memory
    • Early
    • Support 80
     The restraints are winning: HBM is scarce, but standalone optical memory is a roughly $50M research market and nothing ships at volume. HBM4 and HBM5 supply should catch up around 2029. The window is early. 5 listed · 6 private
  22. 22Specialised AI compute
    • Closing
    • $338B
    • Support 79
     Specialised silicon fleets win; the routing and orchestration layer is the toll 16 listed · 6 private
  23. 23Inference infrastructure
    • Bull
    • Closing
    • $23B
    • Support 79
     Post-training becomes inseparable from inference — the data flywheel as the moat on open weights 6 listed · 8 private
  24. 24Datacentre scale-out
    • Closing
    • $235B
    • Support 78
     Scale-out pulls three enabling layers at once; the durable value is in the layers, not the buildout 6 listed
  25. 25AI kernel synthesis
    • Bear
    • Closing
    • $3.1B
    • Support 78
     AI/search-based kernel synthesis attacks the performance-parity-kernel constraint on heterogeneous AI compute, and at expert level on novel silicon it collapses the CUDA library moat 5 listed · 15 private
  26. 26Photonic packaging for CPO
    • Bull
    • Closing
    • $29B
    • Support 77
     Production AI-interconnect CPO (2027-2030) is won by inorganic photonic-packaging platforms (TFLN, BTO/SiN, silicon nitride); polymer 2PP/xolography materials do not qualify 15 listed · 8 private
  27. 27Optical circuit switching
    • Bear
    • Closing
    • $25B
    • Support 77
     Standalone optical circuit switching fails to reach venture scale through 2030 (a zero-success base rate over 20 years, outside Nvidia/Google-scale captive buyers) 8 listed · 19 private
  28. 28Retrieval and memory tier
    • Bear
    • Closing
    • $8.1B
    • Support 76
     Storage value migrates to the retrieval, index and memory tier, off the commodity bytes 7 listed · 10 private
  29. 29Compute as an asset class
    • Bull
    • Closing
    • $70B
    • Support 72
     AI compute becomes a financialised asset class with exchange-traded derivatives and a recognised benchmark price by end-2028 8 listed · 7 private
  30. 30Photonic compute
    • Bear
    • Shut
    • $375M
    • Support 72
     Optical computing is a real venture-scale market, but most of it is a trap, and the selective bet is the thesis. 3 listed · 22 private
  31. 31In-chip microfluidic cooling
    • Bear
    • Closing
    • $3.8B
    • Support 69
     In-die microfluidic cooling moves the binding thermal layer inside the silicon 5 listed · 3 private
  32. 32Optical interconnect
    • Closing
    • $162B
    • Support 67
     Through 2028 the AI/datacenter optical-interconnect transition runs pluggable → LPO/LRO → CPO, not pluggable → CPO. 17 listed · 11 private
  33. 33Decode chips
    • Bull
    • Closing
    • $31B
    • Support 63
     Disaggregated serving is making decode, the memory-bandwidth-bound token-by-token phase, a separate silicon socket beside the GPU rack (NVIDIA Groq 3 LPX next to Vera Rubin, d-Matrix Raptor in MGX trays). 12 listed · 63 private
  34. 34Data infrastructure
    • Bear
    • Closing
    • $505B
    • Support 62
     Durable value in the AI data stack accrues to whoever owns the WRITE path and the data gravity — storage, governance, pipelines, streaming, observability — not the read/retrieval path; retrieval commoditises into Postgres extensions and model-API context windows while lakehouse… 8 listed · 14 private
  35. 35RF waveguide interconnect
    • Bear
    • Early
    • $23B
    • Support 62
     RF-over-dielectric-waveguide is a funded third physical medium for AI scale-up interconnect, claiming the mid-range (roughly 1-10m) socket at 800G-plus lane rates with a purely electrical front-end. 7 listed · 2 private
  36. 36Agentic compute silicon
    • Bear
    • Shut
    • $16B
    • Support 62
     The shift to fleets of autonomous agents moves the binding compute bottleneck off the GPU forward-pass and onto orchestration, tool-execution and retrieval silicon by 2026 6 listed · 13 private
  37. 37AI data supply
    • Bear
    • Closing
    • $28B
    • Support 60
     AI data supply — the coverage-bottleneck stack (map/spine) 5 listed · 9 private
  38. 38GaN-on-si substrates
    • Bear
    • Shut
    • $2.5B
    • Support 60
     Merchant GaN-on-silicon power epitaxy leaks to the device makers who grow their own; only vertical GaN and defence GaN-on-SiC keep a door open. 8 listed · 3 private
  39. 39AI accelerator silicon
    • Bear
    • Shut
    • $1163B
    • Support 58
     General-purpose AI accelerator silicon (a new GPU, or a horizontal training/inference ASIC competing with NVIDIA / Google TPU / AMD / the well-funded ASIC challengers) is a real, enormous, fast-growing market whose durable value is captured by listed incumbents (NVIDIA… 7 listed · 10 private
  40. 40Photodetection
    • Bear
    • Shut
    • $9.9B
    • Support 58
     Datacom photodetection (Ge-on-Si photodiodes, III-V APDs, the receiver/TIA front-end) is real and necessary but not standalone-investible. 6 listed · 2 private
  41. 41Photonic light sources
    • Bear
    • Open
    • $72B
    • Support 57
     Across the whole photonics stack, the layer that actually captures value is the LIGHT SOURCE (lasers-on-silicon, frequency combs, external laser modules), not the modulator or the passive PIC. 9 listed · 7 private
  42. 42Hyperscaler ASICs
    • Bull
    • Closing
    • $84B
    • Support 54
     As hyperscalers in-source AI silicon (Google TPU, AWS Trainium/Inferentia, Microsoft Maia, Meta MTIA), the AI-compute profit pool partially migrates off the merchant GPU. 15 listed · 7 private
  43. 43Decentralised AI
    • Bear
    • Early
    • $35B
    • Support 51
     By 2030, a top-3 large language model by number of users will be a decentralised AI. 8 private
  44. 44Fog computing middleware
    • Bear
    • Closing
    • $187B
    • Support 48
     As compute fragments across GPUs, custom ASICs, photonic, neuromorphic and probabilistic substrates, a middleware layer forms between applications and silicon. 5 listed · 11 private
  45. 45Temporal compute
    • Bear
    • Shut
    • $20B
    • Support 48
     Time-domain compute stays a lab technique: no temporal-logic design ships a qualified production part or exits above $100M by 2031 (watch) 8 listed · 6 private
  46. 46Converter plane cooling
    • Early
    • $7.6B
     The power-converter plane is a second thermal path with no cold plate on it
  47. 47AI thermal substrates >1000 W/mK
    AI compute is increasingly interconnect/thermal-bound, not purely transistor-bound.
  48. 48AI-hardware circular economy
    AI-hardware circular economy: GPU reuse + critical-metal recovery 3 listed
  49. 49ML closes the signal-to-noise gap on non-invasive BCI
    The standing bottleneck on non-invasive brain-computer interfaces is signal-to-noise: scalp EEG, MEG, fNIRS and functional ultrasound are noisy, low-spatial-resolution proxies for neural activity, which is why useful BCI has needed surgery. 2 private
Semiconductor manufacturing34 theses
  1. 12D materials
    • Bear
    • Early
    • $9.1B
    • Support 92
     Two-dimensional materials (graphene, hBN, MoS2, WSe2) reach high-volume manufacturing in at least one semiconductor or photonic application by 2028 (most likely thermal management films, EMI shielding, or UV LED substrates). 5 listed · 10 private
  2. 2Electron-informed materials discovery
    • Bear
    • Shut
    • $719M
    • Support 92
     In-situ electron-band-structure measurement during materials growth becomes the dominant materials-discovery substrate for semiconductors by 2030, displacing DFT screening and combinatorial atomic-scale exploration for interface-dominated problems. 8 listed · 7 private
  3. 3Back-end wafer prep equipment
    • Bear
    • Shut
    • $3.8B
    • Support 88
     Wafer thinning and singulation re-rates from a cyclical back-end niche to advanced-packaging-critical as HBM4 and 3D stacks multiply passes, but value stays locked in DISCO's near-monopoly 8 listed · 1 private
  4. 4Hybrid bonding
    • Bear
    • Shut
    • $919M
    • Support 88
     Hybrid bonding is the next high-density 3D interconnect, but the value is captured by the bonder duopoly + foundry, and the seed door is metrology/materials, not the bonder 8 listed · 2 private
  5. 5Legacy assembly
    • Bear
    • Shut
    • $82B
    • Support 85
     Traditional packaging is a mature, commoditised layer where profit is leaving for advanced packaging, and nothing in it opens with time. 8 listed · 2 private
  6. 6Advanced packaging metrology
    • Bear
    • Closing
    • $12B
    • Support 84
     The incumbents got there first: KLA, Onto and Camtek already ship the hybrid-bond and glass tools, so the AI packaging boom is a public-markets story. 7 listed · 3 private
  7. 7Known-good-die test
    • Bear
    • Closing
    • $7.7B
    • Support 84
     2.5D/3D integration makes known-good-die testing mandatory (one bad die scraps a multi-die package worth thousands) and AI-die power forces liquid-cooled / high-current burn-in. 12 listed · 3 private
  8. 8Trailing-edge capacity
    • Bear
    • Shut
    • $83B
    • Support 82
     Chinese mature-node (>=28nm) capacity additions in analog, power, RF and automotive silicon are building a structural OVERSUPPLY that compresses margins for the Western broad-line IDMs (Infineon, STMicroelectronics, NXP, onsemi) through the late-2020s, even as AI-exposed leading-edge parts stay in acute… 8 listed · 2 private
  9. 9Sovereign electronics manufacturing
    • Bull
    • Open
    • $65B
    • Support 82
     Automation-native vertical manufacturers break the EMS margin curse on sovereign demand 8 listed · 4 private
  10. 10Discovery metrology
    • Bear
    • Early
    • $23B
    • Support 82
     As AI makes candidate generation near-free, the binding constraint on discovery becomes the physical feedback loop, so value accrues to whoever owns the cheapest, fastest real-world oracle (metrology, characterisation, automated experiment) 7 listed · 27 private
  11. 11Ultrapure water and process fluids
    • Bear
    • $28B
    • Support 80
     The fluid path into fabs and AI datacentres is bought as an annuity, and the annuity is why it repriced 7 listed · 1 private
  12. 12Advanced packaging
    • Closing
    • $86B
    • Support 79
     Packaging rent pools in materials, equipment and substrates, not OSAT assembly 26 listed · 4 private
  13. 13Deposition materials and process
    • Bear
    • Closing
    • $44B
    • Support 78
     Deposition value accrues to the molecule (precursors, inhibitors), never the tool 24 listed · 50 private
  14. 14Chiplets
    • Bull
    • Closing
    • $54B
    • Support 75
     Chiplet-based products grow from ~$3B (2021) to ≥$20B by 2027 at 40% CAGR, then capture ≥8% of total semiconductor revenue by 2030. 16 listed · 9 private
  15. 15Back-end test equipment
    • Bear
    • Shut
    • $33B
    • Support 73
     Restraints win: Advantest and Teradyne hold 90% of testers behind decades of qualification, and no standalone test company has reached venture scale. The growth goes to listed companies. 8 listed · 4 private
  16. 16Lithography and patterning
    • Early
    • $55B
    • Support 70
     ASML/EUV lithography is the #1 durable constraint on AI-compute scaling by ~2028-2030 (one supplier, more concentrated than memory). 31 listed · 11 private
  17. 17EDA and chip design
    • Bull
    • Closing
    • $62B
    • Support 67
     AI moves chip design onto the engineering bill; physical design and signoff sold per design is the layer the incumbents cannot copy without eating their seats. 8 listed · 5 private
  18. 18Lithography light sources
    • Bear
    • Early
    • $6.7B
    • Support 65
     The EUV source stays with ASML and Gigaphoton; the only small opening is mask-inspection sources, and nobody has shown up there yet. 5 listed · 2 private
  19. 19Computational lithography
    • Bear
    • Early
    • $2.6B
    • Support 64
     A European-founded company building AI-native mask synthesis tooling for specialty and research fabs (photonics, SiC/GaN, compound semi, photomask shops) will reach a paid pilot with at least one European mask shop before end-2028 without requiring direct qualification competition against Synopsys… 11 listed · 3 private
  20. 20Semiconductor metrology
    • Bear
    • Early
    • $1.1B
    • Support 64
     Semiconductor metrology is the highest-margin picks-and-shovels play adjacent to the litho stack 22 listed · 3 private
  21. 21Specialty 3D patterning
    • Bear
    • Early
    • $757M
    • Support 62
     There is a real, Tier-1-validated demand sub-segment for 3D/non-planar/thick-resist/curved patterning distinct from leading-edge EUV, but as with the parent the tool layer is incumbent-commoditised 8 listed
  22. 22Photonic test and metrology
    • Bear
    • Shut
    • $3.9B
    • Support 58
     As CPO and silicon-photonics ramp, photonic test, known-good-die screening and active-alignment metrology become volume-gating (Teradyne flagged SiPh wafer/module test as the new bottleneck, Feb 2025). 7 listed · 3 private
  23. 23Nanotube electronics
    • Bear
    • $333M
    • Support 58
     Carbon nanotubes used as electronics (transistors, memory, RF, sensors, and chip-process materials such as EUV pellicles) reach chip-industry volume first as fab consumables and sensors, by 2030. 5 listed · 8 private
  24. 24High-NA EUV
    • Bull
    • Shut
    • $32B
    • Support 56
     High-NA EUV lithography reaches widespread HVM availability by 2026, enabling sub-2nm logic nodes; manufactures $200B+ chips annually by 2030 as <10nm reaches 20-30% of total semiconductor production. 8 listed · 2 private
  25. 25Beyond-EUV lithography
    • Bear
    • Shut
    • $26B
    • Support 56
     No commercial semiconductor lithography platform beyond High-NA EUV (i.e. 10 listed · 2 private
  26. 26Photonic foundry capacity
    • Closing
    • $2.7B
    • Support 50
     By 2028, photonic IC foundry capacity binds production volume of co-packaged optics and photonic compute, with fewer than 5 commercial multi-project 300mm photonic foundries operating globally. 10 listed · 23 private
  27. 27Photonic PDK and EDA
    • Bear
    • Open
    • $1.7B
    • Support 46
     The asset-light way to play the photonic-foundry-capacity crunch is the PDK/EDA/design-automation layer every foundry and fabless team needs, where the moat is design IP and process-design-kit lock-in, not a 300mm fab 5 listed · 4 private
  28. 28Photonic multimaterial foundry
    • Bear
    • Early
    • $15B
    • Support 44
     A photonic foundry built ground-up for multi-material integration beats the silicon-first lines that bolt heterogeneity on 5 listed · 10 private
  29. 29Perovskite semiconductors
    • Bear
    • Early
    • $5.7B
     Perovskite emerges as a general-purpose semiconducting substrate, stability is the universal gating problem 6 listed · 10 private
  30. 303D heterogeneous integration production roadmap commitment
    TSMC SoIC, Intel Foveros, Samsung X-Cube moved from R&D demonstration to production roadmap commitment in 2025-2026.
  31. 31Ai-eda-verification-wedge
    AI agents are already cutting chip verification time at named customers: Cadence reported up to 10x productivity at NVIDIA and Qualcomm in Feb 2026. The window is open, with the incumbents moving first.
  32. 32Area selective deposition
    Annual industry coordination point for BEOL deposition tooling, financed by Tokyo Electron, Lam Research, Applied Materials, attended by GlobalFoundries.
  33. 33EU-sovereign advanced packaging tools
    EU-sovereign capital equipment for sub-µm BEOL / advanced packaging (iCVD conformal coating, gridless ion beam thin-film deposition, direct-write 3D BEOL features) is structurally underbid relative to demand, because (a) the foundry fight has shifted from FEOL to BEOL+packaging, (b) most…
  34. 34PFAS regulatory ban
    EU PFAS Restriction Proposal under ECHA review with a decision target in 2026-2027; US EPA's PFAS rules in active enforcement.
Software & applications12 theses
  1. 1Physics AI
    • Bull
    • $39B
    • Support 86
     Physics-AI value accrues to vertical specialists holding cross-customer data rights, not to horizontal platforms, by 2030 6 listed · 12 private
  2. 2Autonomous remediation
    • Bear
    • Early
    • $43B
    • Support 78
     Fixing becomes the bottleneck once AI makes finding free, and the patch-writing agent is the product 4 listed
  3. 3Non-human identity and secrets
    • Bear
    • Closing
    • $38B
    • Support 78
     The keys software holds become the main way in, and the platforms that already own identity absorb the fix 3 listed
  4. 4Cyber insurance repricing
    • Bear
    • Shut
    • $35B
    • Support 78
     Cheap attacks reach the economy through insurance first: loss ratios rise while prices still fall, so the market turns hard and the tail goes uninsured
  5. 5Software supply-chain integrity
    • Bear
    • Open
    • $31B
    • Support 78
     Proving the code you run is the code that was built becomes a required control, and the code hosts give most of it away 2 listed
  6. 6Agentic payment rails
    • Shut
    • $29B
    • Support 74
     A standalone neutral netting/clearing layer for agent micro-payments does NOT survive as an independent business; netting is absorbed by the rails and card networks by 2027 (contested: the original bet has inverted) 6 listed · 10 private
  7. 7Proof of humanhood
    • Bear
    • Shut
    • $1.8B
    • Support 74
     Proof of humanhood has the forcing function agent identity lacks, and still is not a hardware market 5 listed · 5 private
  8. 8AI value capture moves downwards
    • Bull
    • Closing
    • $538B
    • Support 62
     The AI application/agent software layer is where AI revenue SHOWS UP but not where it is KEPT — value behaves as an ingredient that flows to whoever already owns the data, distribution and workflow (incumbent SaaS re-rating on AI-attach) plus… 6 listed · 13 private
  9. 9AI security
    • Bear
    • Closing
    • $27B
    • Support 54
     The attackers' side wins the market and the street already knows it: the one venture opening, autonomous offence, is priced above $1B with incumbents on the cap table. 8 listed · 12 private
  10. 10AI eats SaaS
    • Bull
    • Open
    • $447B
    • Support 51
     Agentic AI decouples software value from seat count — flattening-to-shrinking per-seat revenue in workflow-automatable categories (0-3% CAGR) while the dollar TAM grows only ~8-10% — and in the ~$210B of exposed spend, AI-native entrants OWN new categories worth a ~$55-65B… 6 listed · 11 private
  11. 11Agent identity
    • Bear
    • Shut
    • $0M
    • Support 49
     Agent identity (KYA) is repeatedly named as the layer where durable value pools in agentic payments. 8 listed · 13 private
  12. 12The convergence of AI agents and commerce
    AI agents transacting on behalf of users (discovery, negotiation, payment, settlement) is an undervalued convergence, with a crypto/stablecoin crossover for agent-to-agent settlement. 6 listed · 2 private
Defence & security14 theses
  1. 1Counter-UAS effectors
    • Bear
    • Closing
    • $10B
    • Support 85
     Counter-UAS effectors: the drone-defence money is real and dated, but the integrated-effector socket is being handed to the primes before a startup can reach it 5 listed · 8 private
  2. 2Quantum random number generation
    • Bear
    • Shut
    • $274M
    • Support 85
     On-silicon quantum random number generation (QRNG) — true, non-deterministic entropy from a quantum physical process (avalanche/tunnelling noise in CMOS-compatible silicon) — plus the wider quantum-security IC layer is a real, policy-pulled market. 5 listed · 1 private
  3. 3QKD
    • Bear
    • Shut
    • $230M
    • Support 85
     The physics works and China built a national quantum network, but Western security agencies tell every buyer QKD most wants not to buy it. 3 listed · 3 private
  4. 4Confidential computing
    • Closing
    • $196B
    • Support 82
     By 2030, ≥85% of new SoCs from leading vendors (Apple, Arm-based, Intel, AMD, Nvidia, MediaTek, Qualcomm) ship with hardware TEE / confidential computing enclaves integrated. 7 listed · 2 private
  5. 5Orbital defence
    • Bear
    • Closing
    • $16B
    • Support 80
     Orbital defence: Golden Dome turns space into a munitions programme, and it is the demand engine under every space component page (watch) 8 listed · 9 private
  6. 6FHE acceleration silicon
    • Bear
    • Shut
    • $6.8B
    • Support 80
     Fully homomorphic encryption is moving from ~10,000x-slower-than-plaintext toward practical via dedicated silicon (FHE accelerators), the one confidential-computing primitive that needs new chips rather than incumbent-owned TEEs. 5 listed · 7 private
  7. 7Space domain awareness
    • Bear
    • Shut
    • $4.6B
    • Support 80
     Space domain awareness: a small, slow, government-shaped market carrying billion-dollar private valuations (watch) 7 listed · 10 private
  8. 8Hardware memory safety
    • Bear
    • Early
    • $380B
    • Support 77
     Hardware-enforced memory safety, capability silicon in the CHERI / CHERIoT lineage, becomes a required layer for security-critical and legacy-C/C++ compute as memory-safety mandates bite. 6 listed
  9. 9High-speed data converters
    • Bear
    • Early
    • $7.8B
    • Support 76
     High-speed data converters: the merchant ADC/DAC layer behind every phased array, and whether a European entrant can get a production design-in 8 listed · 8 private
  10. 10Agentic AI security
    • Bull
    • Early
    • $27B
    • Support 67
     As AI agents gain the ability to execute code and reach into low-level infrastructure, value accrues to whoever owns the trusted-execution and attestation boundary that contains them 8 listed · 13 private
  11. 11Hardware attestation
    • Bear
    • Early
    • $20B
    • Support 66
     A neutral cross-vendor verifier for what ran on shared AI hardware could be the Okta of compute, but nobody yet pays for what NVIDIA, Intel and AWS give away. 8 listed
  12. 12Defence field robotics
    • Bull
    • Closing
    • $3.7B
    • Support 64
     Defense / field robotics is the most revenue-validated frontier-robotics segment, driven by a zero-human-environment premium and sovereign procurement 4 listed · 7 private
  13. 13Airspace awareness sensing
    • Bear
    • Early
    • $1.4B
    • Support 64
     Airspace awareness sensing: the radar layer that makes uncrewed flight legal and sites defensible 7 listed · 5 private
  14. 14Defence mass autonomy
    • Closing
    • Support 64
     Defense tech — mass-produced attritable autonomy and the European resilience stack (watch) 2 listed · 6 private
Robotics & physical AI13 theses
  1. 1Edge AI inference accelerators
    • Bear
    • Early
    • $9.6B
    • Support 72
     Robots need big models on a battery and today's chips waste energy moving weights, and the one device approach that computes inside stacked memory has thin evidence so far; early. 6 listed · 11 private
  2. 2High-mix manufacturing automation
    • Bear
    • Early
    • $2.3B
    • Support 66
     High-mix manufacturing automation: the deployed cell a factory buys, where the blocker is engineering cost per deployment and not robot capability 7 listed · 5 private
  3. 3Timing and PNT
    • Bear
    • Early
    • $14B
    • Support 64
     As the physical world fills with autonomous systems that must act together, the scarce bottlenecking input is precise, GPS-denied, trustworthy time and position (chip-scale atomic/optical clocks, PNT resilience), not compute 6 listed · 6 private
  4. 4Teleoperation
    • Bear
    • Open
    • $8.6B
    • Support 63
     Between 2026 and 2030, teleoperation is the dominant go-to-market and data-acquisition pathway for physical-AI in contact-rich, high-value verticals (defence and surgery first, then logistics, then manufacturing). 3 listed · 16 private
  5. 5Robot data supply
    • Bear
    • Early
    • $85B
    • Support 57
     Robot data is a stack, not a source; picks-and-shovels win at each part 4 listed · 11 private
  6. 6Robot autonomy
    • Bear
    • Early
    • $47B
    • Support 57
     General-purpose robot autonomy (open-world dexterous manipulation, not geofenced driving) remains data-bottlenecked and unreliable through 2030. 9 listed · 11 private
  7. 7Tactile sensing silicon
    • Bear
    • Early
    • $21B
    • Support 57
     As robots scale to whole-body, high-resolution touch, the contact-data torrent stresses the host-CPU model and pushes readout + first-stage processing toward the sensor node. 2 listed · 7 private
  8. 8Robot foundation models
    • Bear
    • Shut
    • $21B
    • Support 57
     Robot foundation models (VLA) are strategically central but value-capture is contested; only a proprietary self-reinforcing data loop earns durable margin 5 listed · 8 private
  9. 9Hazardous-environment robotics
    • Bear
    • Early
    • $5.5B
    • Support 57
     Hazardous-environment field robotics: machines bought to enter places a person cannot survive or cannot fit 5 listed · 6 private
  10. 10Functional safety silicon for robotics
    • Bear
    • Shut
    • $2B
    • Support 57
     Every scaling robot needs a functional-safety system that forces a safe state when the AI behaves unpredictably or the operator link drops. 5 listed · 5 private
  11. 11Service robotics
    • Support 57
     Service robotics — the VLA inflection opens the $450B cleaning/service-labour market (watch)
  12. 12Humanoid actuators
    • Bear
    • Closing
    • $6.6B
    • Support 47
     Humanoid picks-and-shovels — the component layer holds value only where China compression fails (roller screws, novel-process motors, the sovereignty channel) 6 listed · 13 private
  13. 13Surgical robotics
    • Bear
    • Shut
    • $38B
    • Support 44
     Surgical robotics is the most commercially mature, profitable robotics segment (~$8-14bn 2025, ~13-14% CAGR). 7 listed · 6 private
Healthcare & bio16 theses
  1. 1Biocomputing
    • Bear
    • Early
    • $11B
    • Support 78
     Living-neuron / organoid biocomputing is a real and accelerating research field, but as an investment it does not clear the gate today. 3 listed · 7 private
  2. 2Brain-computer interfaces
    • Closing
    • $12B
    • Support 61
     In BCI, venture value accrues to the neural-interface silicon layer (modular/chiplet ASICs, mature-node CMOS plus analog front-ends), not the full-stack clinical device 9 private
  3. 3Molecular biosensing
    • Early
    • $5.5B
    • Support 61
     Molecular biosensing: chip-scale detection of proteins, enzymes and biomarkers across transduction physics 1 private
  4. 4Metabolic wearables
    • Bear
    • Early
    • $2.1B
    • Support 61
     A full-stack metabolic wearable with a hard optical-sensing moat reaches venture scale by 2028, where the chip-supplier and OEM-feature models cannot 5 listed · 12 private
  5. 5Biophotonics
    • Bear
    • Closing
    • $1.6B
    • Support 61
     In biophotonics the venture-differentiated layer is the photonic-IC / source / detector silicon inside the instrument, not the clinical device; value migrates forward to the regulated device and assay incumbent, so capture is mostly public/incumbent through 2029 8 listed · 6 private
  6. 6Implantable neural electrodes
    • Bear
    • Closing
    • $1.3B
    • Support 61
     The open BCI question is whether an implantable electrode-materials team (graphene, soft conducting polymers, thin-film arrays) holds licensable IP that every implant builder must buy 2 listed · 8 private
  7. 7Non-invasive brain measurement
    • Bear
    • Early
    • $24B
     Non-invasive brain measurement: a new-modality cohort (OPM-MEG, functional NIRS, transcranial ultrasound, portable MRI) chasing a market the three imaging incumbents already own 5 listed · 19 private
  8. 8Neuromodulation
    • Bear
    • Early
    • $18B
     Venture-backed neuromodulation exits stay capped at the strategic tuck-in band (~$0.5-0.65B) through 2029, because four incumbents own the channel and copy the differentiator 8 listed · 25 private
  9. 9Light-controlled therapeutics
    • Early
    • $15B
     Does light become a paid therapeutic control layer beyond dermatology once the first optogenetic, photoimmunotherapy and endoscopic light-drug approvals arrive (2027-2029)? 4 listed · 36 private
  10. 10Pharma analytical instrumentation
    • Bear
    • Early
    • $3.6B
     Two European entrants arrived in one quarter selling MEASUREMENT into a pharmaceutical workflow rather than a drug into a patient. 6 listed · 5 private
  11. 11Organoid safety testing
    • Bear
    • Early
    • $3.5B
     Organoid safety testing: human tissue replacing the animal toxicity study 1 listed · 26 private
  12. 12Perovskite radiation detection
    • Bear
    • Early
    • $2.7B
     Perovskite semiconductors win a durable niche in radiation / X-ray / particle detection where their stopping power and low-cost solution processing beat incumbents. 4 listed · 2 private
  13. 13Photopharmacology in scope-reachable tissue
    • Bear
    • Early
    • $96M
     Light-switchable small molecules, not optogenetic gene therapy, win the first light-controlled therapies in tissue a scope or catheter can light (bladder, gut, airway, skin) by 2034, and the molecule owner captures it through pharma licensing 2 listed · 6 private
  14. 14Connectomics tooling
    • Bear
    • Early
     Does the neural-data-acquisition + connectomics-tooling layer become a fundable horizontal picks-and-shovels market under brain-algorithm AI, or do the application companies integrate it away? 3 listed · 5 private
  15. 15Diagnostics and health sensing
    • Early
     Liquid biopsies, real-time sensors and GLP-1-class drugs do not reduce system-level healthcare cost, they migrate it from acute treatment of today's cheap-and-fast killers into a longer tail of frailty, dementia and chronic disease
  16. 16Transcranial focused ultrasound
    • Early
     Transcranial ultrasound is the only non-surgical modality that can reach deep brain targets at millimetre scale, and the same hardware can stimulate and read. 9 private
Energy12 theses
  1. 1Grid electrification
    • Closing
    • Support 94
     The grid + electrical-equipment buildout, the biggest AI-infrastructure opportunity 16 listed · 6 private
  2. 2Fusion
    • Bear
    • Early
    • $14B
    • Support 90
     By the end of 2032, at least one privately-funded fusion startup demonstrates grid-connected net-positive electricity (Q_total > 1, including all balance-of-plant, sustained for >1 hour). 7 listed · 12 private
  3. 3Flexible compute as a grid asset
    • Bull
    • Closing
    • $32B
    • Support 86
     AI load arrives faster than the grid can be built (5-7yr interconnection queues vs ~2yr to build a datacentre). 8 listed · 4 private
  4. 4Advanced nuclear SMRs
    • Bear
    • Closing
    • $18B
    • Support 86
     Small modular and advanced fission reactors (50-300 MWe, factory-built, >90% capacity factor) reach first commercial datacentre deployments this decade, driven by hyperscaler firm-power PPAs (Google-Kairos, Amazon-X-energy, Microsoft-Constellation). 8 listed · 15 private
  5. 5Solid-state transformers
    • Bull
    • Closing
    • $5.9B
    • Support 86
     The conventional-transformer shortage plus the 800 VDC datacentre migration (NVIDIA/OCP standard, committed Oct 2025-Mar 2026, deployment from 2027) opens a venture-scale market for wide-bandgap solid-state transformers (MV-in to HVDC-to-rack). 8 listed · 4 private
  6. 6Electrolyser materials
    • Bear
    • Closing
    • $4.1B
    • Support 86
     Electrolyser components and materials (catalyst-coated membranes, porous transport layers, membranes, coated bipolar plates) sold to electrolyser OEMs 8 listed · 6 private
  7. 7Perovskite tandem PV
    • Bull
    • Open
    • $4B
    • Support 86
     Perovskite-silicon tandem reaches ≥10% of new-build utility-scale PV by 2032 and continues compounding 7 listed · 13 private
  8. 8Offshore power islands
    • Bear
    • Early
    • $772M
    • Support 86
     A sibling of space-based-datacentre in the 'escape the grid + land bottleneck' family, but far cheaper to reach: instead of a spaceport, deploy from a dry dock. 3 listed · 5 private
  9. 9Next-gen geothermal
    • Bear
    • Early
    • $690M
    • Support 86
     Next-generation geothermal (Fervo-style EGS, Eavor closed-loop, Quaise superhot-rock / mm-wave drilling) turns geothermal from a geographically-limited niche into go-anywhere firm baseload IF drilling cost per metre falls enough to hit sub-$60/MWh (geothermal averaged ~$68/MWh in 2025). 4 listed · 9 private
  10. 10Next-gen batteries
    • Bear
    • Early
    • $141B
    • Support 44
     The materials layer is the only venture-shaped part of solid-state, and CATL and Toyota internalise any proven chemistry; nothing changes until an independent supplier is qualified. 8 listed · 8 private
  11. 11Baseload power
    • Shut
    • $12B
    • Support 22
     Firm, 24/7, carbon-free baseload power is the binding SUPPLY-side constraint on AI datacentre and electrification growth this decade (the supply-side mirror of the demand-side ai power thermal binding). 21 listed · 12 private
  12. 12Glazing-integrated photonics
    • Bear
    • Closing
    • $14B
     One physics claim, two end markets. 5 listed · 4 private
Industrial11 theses
  1. 1Distributed fibre sensing
    • Bear
    • Closing
    • $3B
    • Support 85
     Distributed fibre-optic sensing 6 listed · 5 private
  2. 2Quantum sensing
    • Bear
    • Early
    • $1.7B
    • Support 85
     GPS jamming makes quantum navigation necessary, but the buyers are defence programmes at R&D scale and the sensors are still hand-built lab racks. 6 listed · 10 private
  3. 3Specialty silicon
    • Shut
    • $197B
    • Support 84
     Specialty silicon (≥28nm RF, power, sensors, automotive) doubles to $80B+ revenue by 2030 32 listed · 2 private
  4. 4Rare earths
    • Bull
    • Closing
    • $25B
    • Support 82
     Ex-China rare-earth & magnet supply stays structurally short through 2027; value accrues to non-Chinese separation, recovery, and RE-free magnets 7 listed · 13 private
  5. 5At-line industrial analytics
    • Bear
    • Shut
    • $10B
    • Support 61
     Lab-grade measurement moves onto the production line (at-line industrial analytics) 8 listed
  6. 6Mid-IR photonic sensing
    • Bear
    • Early
    • $2B
    • Support 61
     Mid-IR photonic sensing wins gas analysis and breath diagnostics by 2028 7 listed · 8 private
  7. 7Thermoelectric materials
    • Bear
    • Early
    • $5.8B
    • Support 58
     Thermoelectrics stays a perennial niche (Peltier spot-cooling + battery-free sensing); no manufacturable ZT step-change or datacentre design-win reopens waste-heat recovery as a venture-scale market through 2030 — pass by default, watch two tripwires 4 listed · 15 private
  8. 8Uncertainty-propagation compute
    • Bear
    • Early
    • $746M
    • Support 44
     Deterministic uncertainty-propagation compute: carrying probability distributions through existing scalar code, in a compiler and a co-processor, replaces Monte Carlo for low-to-mid dimensional problems 2 listed · 2 private
  9. 9Industrial condition monitoring
    • Bear
    • Closing
    • $13B
     Industrial condition and compliance monitoring: transducer plus analytics plus platform, sold to the asset owner (coverage theme) 8 listed · 3 private
  10. 10Energy harvesting
    • Bear
    • Shut
    • $1.5B
     Ambient energy harvesting (powering electronics from RF, light, vibration, thermal gradients) is a real technology with a structurally sub-venture market: the systems market converges around $1.0-1.2B by 2030 growing 8-10%, the silicon value sits in harvesting PMICs and transducers shipped… 4 listed · 4 private
  11. 11Spectral imaging
    • Shut
     Capturing spectral information beyond plain RGB on a chip, for material and chemical identification, is one coherent space, not four. 8 listed · 4 private
Space11 theses
  1. 1European launch
    • Bear
    • Shut
    • $1.9B
    • Support 85
     Europe flew fewer than ten orbital launches in 2025 against more than 190 for the United States, and the money now closing that gap is defence money rather than civil space money, which makes European launch sovereignty the datable why-now… 5 listed · 7 private
  2. 2Space-based datacentres
    • Bear
    • Early
    • Support 83
     REBUILT AROUND THE CHEAP-LAUNCH CASE (3 Sep 2026, at Lawrence's instruction: he believes launch costs fall faster than consensus). 3 listed · 21 private
  3. 3SAR and EO satellite data
    • Bear
    • Closing
    • $11B
    • Support 80
     SAR/EO repricing — sovereign capability captures the value while imagery and analytics both commoditise (watch) 7 listed · 14 private
  4. 4Smallsat platforms
    • Bear
    • Early
    • $5.8B
    • Support 80
     The merchant smallsat platform layer: sovereign constellations are the demand, and fifteen years of fragmentation is the reason it still does not produce a EUR1B European exit (watch) 7 listed · 4 private
  5. 5Space power electronics
    • Bear
    • Early
    • $1.4B
    • Support 80
     Space power: the kilowatt-class platform generation and where the power layer's value sits (watch) 8 listed · 11 private
  6. 6In-space manufacturing
    • Bear
    • Early
    • $19B
    • Support 76
     In-space manufacturing — downmass gets solved while the product market fails to appear (watch) 5 listed · 10 private
  7. 7In-space mobility
    • Bear
    • Early
    • $5.6B
    • Support 76
     In-space mobility: private capital is running an order of magnitude ahead of the government line item (watch) 4 listed · 15 private
  8. 8Radiation-hardened chips
    • Bear
    • Shut
    • $3.3B
    • Support 76
     The investable opportunity in space radiation-hardened silicon is not the fab or the high-assurance incumbents but fabless RHBD IP, rad-tolerant SoCs, and qualification tooling built on commercial foundry nodes (FD-SOI / FinFET). 7 listed · 5 private
  9. 9Commercial LEO stations
    • Bear
    • Early
    • $1.8B
    • Support 76
     Commercial LEO destinations: the ISS deorbit puts a hard date on a market that has never had one (watch) 4 listed · 8 private
  10. 10Space power beaming
    • Bear
    • $1.3B
    • Support 76
     Laser (and some microwave) power beaming is moving from ground records to the first in-orbit tests, led by Star Catcher's Protostar (launched 1 Oct 2026). 8 listed · 9 private
  11. 11Nanomechanical space systems
    • Bear
    • Early
    • Support 76
     By 2030, nanomechanical compute devices are deployed in 10+ orbiting satellites for radiation-hard / low-power use cases. 2 listed
Consumer devices7 theses
  1. 1AR display optics
    • Bull
    • Early
    • $22B
    • Support 85
     The smart-glasses optical stack splits into a light engine (generate RGB), a combiner (get it to the eye), and the prescription lens. 8 listed · 7 private
  2. 2Structured micro-optics
    • Bear
    • Early
    • $4B
    • Support 85
     Two European component layers share one claim: the value in an optical element is moving out of assembled, polished glass and into a manufactured microstructure, replicated at wafer scale or written directly onto a substrate. 6 listed · 3 private
  3. 3Metalenses
    • Bear
    • Shut
    • $920M
    • Support 85
     Metasurface flat optics (platform) 3 listed · 6 private
  4. 4MEMS sensing and actuation
    • Bear
    • Shut
    • $42B
    • Support 66
     MEMS sensing & actuation is a real and growing market (>$19B by 2030) but a foundry-scale, cost-per-unit, qualification-driven business where the durable value is held by the Bosch/ST/TDK/Infineon oligopoly; no European MEMS startup reaches venture scale independently this decade, the… 24 listed · 4 private
  5. 5Image sensor silicon
    • Bear
    • Shut
    • $1.4B
    • Support 66
     New sensor types (SPAD, shortwave infrared, event) grow 10-20% a year and each gets bought or licensed by Sony, ST or onsemi once it proves out. 11 listed · 6 private
  6. 6Ferroelectric memory
    • Bear
    • Shut
    • $9.4B
    • Support 58
     Hafnium-oxide-based ferroelectric memory (FeFET, FeRAM) captures ≥5% of embedded non-volatile memory units by 2030, up from <0.5% in 2024. 7 listed · 4 private
  7. 7AR optics test and metrology
    • Bear
    • Early
    • $1.2B
     As display glasses scale, every waveguide needs an inline image-quality test (eyebox, MTF, uniformity, colour) and every waveguide wafer needs grating and flatness metrology. 1 listed
Telecoms & wireless6 theses
  1. 1Photonic RF signal generation
    • Bear
    • Early
    • $6.9B
    • Support 85
     [CANDIDATE THESIS, unvalidated] Photonic generation of RF/mmWave carriers displaces electronic frequency-multiplier synthesis as the preferred low-noise, widely-tunable source above ~40 GHz for satellite comms, 6G and high-end test. 8 listed · 5 private
  2. 2Satellite terminal silicon
    • Bear
    • Shut
    • $1.6B
    • Support 80
     Satellite user terminals: the volume node of the LEO buildout, with a sub-$100M exit record in the silicon layer (watch) 8 listed · 8 private
  3. 3Diamond RF power
    • Bear
    • Early
    • $1.2B
    • Support 80
     Diamond RF power transistors begin displacing GaN at the high-power, high-frequency, thermally-constrained edge of defence radar and satcom by 2030, where diamond's thermal conductivity and breakdown field break the GaN ceiling 9 listed · 5 private
  4. 4Direct-to-device satellite
    • Bear
    • Shut
    • $28B
    • Support 76
     Direct-to-device satellite: the fastest-growing space market, absorbed into the baseband before a merchant silicon layer forms (watch) 8 listed · 10 private
  5. 5RF front-end silicon
    • Bear
    • Early
    • $126B
    • Support 67
     RF front-end - the qualification moat, not the process, wins sovereign and defence demand 8 listed · 7 private
  6. 6Industrial wireless IoT
    • Bull
    • Open
    • $11B
    • Support 63
     Industrial wireless IoT connectivity silicon and modules; DECT NR+ and the 3.8-4.2 GHz decision are a catalyst inside it, not the market 8 listed · 7 private
Quantum6 theses
  1. 1Quantum processor hardware
    • Bull
    • Closing
    • $5B
    • Support 82
     Quantum processor hardware (the qubit chip and the machine, every non-photonic modality) is a ~$470M market growing ~30% a year that sovereign buyers pay for, priced at roadmap multiples and entered with nine-figure rounds 8 listed · 9 private
  2. 2Quantum software
    • Bear
    • Closing
    • $4.9B
    • Support 54
     Modality-agnostic quantum software (the gate-synthesis, compiler and control layer plus near-term application algorithms) is a hardware-risk-free way to back quantum, investable wherever the IP abstracts above the qubit 8 listed · 5 private
  3. 3Cryogenic computing
    • Bear
    • Open
    • $848M
    • Support 54
     Control chips must move into the fridge for quantum to scale, but the layers worth holding are being claimed fast; component suppliers have a short window before margins go. 8 listed · 12 private
  4. 4Quantum memory
    • Bear
    • Closing
    • $196M
    • Support 54
     Quantum memory becomes a fundable standalone product only as a telecom-native rare-earth/erbium module; otherwise value pools above it in the network/QPU layer (falsified if a standalone module raises $40M+ by end-2027) 3 listed · 4 private
  5. 5Photonic quantum computing
    • Bear
    • Closing
    • $13B
    • Support 22
     Photonic quantum architectures capture ≥30% of useful-quantum-advantage workloads by 2030 8 listed · 4 private
  6. 6Quantum-sensing-photonic-hardware
    The investable layer in quantum sensing (gravimetry, inertial navigation, timing) is the photonic hardware supply chain — narrow-linewidth / Brillouin lasers, photonic integrated circuits, and vacuum/atom packaging — which matures and captures value before the end-system sensor market does.
Automotive & mobility5 theses
  1. 1LiDAR and ranging
    • Bear
    • Shut
    • $8.1B
    • Support 66
     LiDAR systems are a price war the volume incumbents have won; the one opening, FMCW ranging on a photonic chip, is a photonics component market. 10 listed · 4 private
  2. 2SiC for EV traction
    • Bear
    • Shut
    • $12B
     SiC captures >70% of EV traction inverter shipments by 2030 14 listed · 5 private
  3. 3eVTOL
    • Closing
    • $11B
     By 2030 eVTOL has not reached commercial passenger volume. 4 listed · 5 private
  4. 4Automotive zonal power
    • Bear
    • $6B
     The car's fuse box becomes silicon, and the content gain is eaten by the harness it deletes 8 listed · 1 private
  5. 5Sintered silver die attach
    • Bear
    • Shut
    • $362M
     Sintered-metal die-attach: power-electronics workhorse, and the AI-migration question 5 listed · 4 private

Underneath the theses: the reference, every technology they are built from, entry by entry, and the beliefs they rest on.