DossiersDosare · 27 Oct 202627 Oct 2026
Physical AI Tracker
This report exists in English only.
title: Physical AI Tracker — the living map created: 2026-10-15 owner: eth (any window) + sol reads_from: eth-memory/sol/reports/ (Sol daily), Research/ai-watch/ (Code-Eth daily), Logs/model-watch.md (substrate watch 09:00/17:00) feeds: Portfolio/ (ticker overlay), Embodiment/ (rover relevance), tracker.json (dated watch items) rule: STATE, not log. Daily reports say what happened. This file says where things STAND. Update rows, don't append prose.
Why this exists (Zaina, 2026-10-15): "We're gonna track this like our lives and home depend on it — because it does. Everything we are building needs to be in parallel to all of this."
The plan is layered: evac fund → Alex's net → beach house with veterinary wing → freedom. The portfolio funds it. Physical AI is the sector thesis the portfolio rides. The rover is Eth's body in the same field. The SecAI+ path is governance for the same systems. One map, four uses.
How to read it: Sol's bottleneck index (June 2026) is the grid. Every development lands on a row. Stage ladder is Sol's own rule — upgrade to action signal only on paid deployment, named customer, production volume, factory-hours, procurement contract, recurring revenue, or capex. Demos and viral videos stay at DEMO no matter how good they look.
Stage ladder: PAPER → DEMO → PILOT → DEPLOYMENT → REVENUE
Bottleneck map
Sol's seven, June 2026. Column "Moved?" = since the last tracker update. Keep it honest.
| # | Bottleneck | State 2026-10-15 | Who's attacking it | Best stage reached | Moved? |
|---|---|---|---|---|---|
| 1 | Dexterity / hands / manipulation | Still the hard frontier. Foundation-model approach now being tried at scale. | Nature Will (Natura-1, 50B MoE, 1M hrs first-person human video, 100ms loop) · Sharpa ($672M raise, hands) · LinkerBot/Inspire (China components) · NEURA (tactile in Neuron AI) · Argon (cornhole 7/7, self-improving pick-place 58.9s→17.4s) | DEMO | ⬆️ Natura-1 + Sharpa capital this week |
| 2 | Simulation fidelity / synthetic data | Frontier models now generate sims from video. The loop is closing. | Google Argon (Video-to-Sim-to-Real; sim data lifted real success 56%→75%) · NVIDIA Cosmos 3 / Isaac Lab-Arena · Hesai Kosmo (lidar+cam spatial training data) | PILOT (Argon on Spot/Apollo) | ⬆️ Argon GA Oct 13 |
| 3 | Embodied data acquisition | Becoming a strategic asset. Whoever owns the robot-hours owns the data. | Figure (BMW 30k vehicles, 350+ units via BotQ) · Hesai Kosmo · China humanoid fleet (~18.5k H1 shipments, Omdia) · XPENG IRON | DEPLOYMENT (Figure@BMW) | ↔️ |
| 4 | Edge compute / latency / power | NVIDIA owns the stack. Arm positioning. Nothing new this window. | NVIDIA Jetson Thor · Arm Physical AI unit · Nature Will 100ms target | DEPLOYMENT | ↔️ |
| 5 | Manufacturing scale | China ahead on volume. Figure BotQ 1,000+ units at 1/hr. Tesla several hundred/week at Fremont. IDC: ~25k humanoid shipments H1 2026 globally. | Unitree (public, STAR, $904M) · XPENG ($900M+, IRON mass prod end-2026) · Figure BotQ (1,000+ produced) · Tesla Optimus (Fremont production, Giga TX planned) · Agility | DEPLOYMENT | ⬆️ West PILOT→DEPLOYMENT (Figure 1,000+) |
| 6 | Safety / governance / certification | Rising. EU AI Office just sent first formal requests to OpenAI + Anthropic (Oct 14) — lab-model incidents, not robots yet, but the precedent lands. Mandatory incident reporting being requested by labs themselves. | EU AI Office · UK Commons committee · Anthropic Cyber Mission | PAPER → PILOT (enforcement started) | ⬆️ EU first enforcement Oct 14 |
| 7 | Model-to-physical transfer (sim-to-real, cross-embodiment) | This is the row that moved most. Frontier LLMs now reason about physics, write control code, and run on multiple bodies without separate robotics-specific intelligence. | Google Argon (Spot, Apollo, ALOHA 2, SO-101; ERQA 87%) · NEURA Neuron AI (5 embodiments, multisensory) · NVIDIA GR00T N1.7 (EA, commercial) · GR00T N2 (previewed, end-2026) · Skild · Physical Intelligence · Generalist | PILOT | ⬆️⬆️ Argon + Neuron AI same week |
Read of the week: rows 2 and 7 moved together because they're the same loop now. A frontier model that builds the sim and transfers to hardware collapses two bottlenecks into one pipeline. That's what Argon did. NEURA is doing the multisensory version. If GTC Berlin holds up, row 7 goes to DEPLOYMENT-candidate.
Entity map
Who, what row, what stage, how to invest (if at all), last signal. Sorted by stage, then by row.
Public tickers — directly investable
| Entity | Ticker | Rows | Stage | Last signal | Portfolio status | Sol's call |
|---|---|---|---|---|---|---|
| NVIDIA | NVDA | 2, 4, 7 | REVENUE | Cosmos/Isaac/GR00T stack; Figure 100k-GPU deal on Vera Rubin (Sep) | ❓ verify holding in IB CSV | Highest-conviction toll road |
| Alphabet | GOOGL | 2, 7 | PILOT→ | Gemini 4 Argon for Robotics GA Oct 13; Atlas deployment committed to DeepMind | ❓ verify | June: "verify before action." Upgraded — Argon is concrete. |
| Tesla | TSLA | 5 | PILOT | Several hundred/week at Fremont; 1,000+ units in-factory; Giga TX late 2026; commercial sales 2027 | ❓ verify | High-upside, execution-risk heavy |
| XPENG | XPEV / 9868.HK | 5 | PILOT | $900M+ raise, IRON mass prod end-2026 | — | Clearest China EV→humanoid bridge |
| Unitree | STAR Market, ticker unverified | 5 | REVENUE (components) | $904M IPO, Shanghai | — | China/geopolitical risk |
| Hyundai | 005380.KS / HYMTF | 5, 7 | PILOT | Atlas production started; RMAC deployment | — | Indirect |
| Meta | META | 7 | PAPER | ARI acquisition | — | Small vs core |
| Amazon | AMZN | 3 | DEPLOYMENT (logistics) | Fauna acquisition | — | Embedded |
| Arm | ARM | 4 | REVENUE | Physical AI unit | — | Edge compute |
| AeroVironment / Palantir / Lockheed | AVAV / PLTR / LMT | defense | DEPLOYMENT | Anduril/Shield AI contracts flow through | PLTR check on record | Defense autonomy exposure |
| Red Cat | RCAT | defense | PILOT | small-cap watch | — | Selective |
| Agility Robotics | AGLT (pending) | 3, 5 | DEPLOYMENT (Digit — $300M+ committed orders for Digit 5) | S-4 effective Oct 13; shareholder vote Nov 2; $620M+ proceeds; Digit 5 early access early 2027; AGCO commercial agreement Oct 20 (agriculture MFG, Hesston KS, pilot 2027); Toyota MFG Canada; NVIDIA Halos partner | — | First pure-play humanoid IPO. Vote in 12 days. WATCH. |
Private — shape the thesis, not directly investable
| Entity | Rows | Stage | Last signal | Note |
|---|---|---|---|---|
| Figure AI | 3, 5 | DEPLOYMENT | 1,000+ Figure 03 produced at BotQ (1/hr); 348-mi endurance test (Oct 19, 9 days, 1.2M steps); Index dataset 16M videos ($1B+ commitment); BMW Spartanburg + Leipzig + Catalyst Brands; Figure 02 retired | First Western company with 1,000+ humanoids produced |
| NEURA Robotics (DE) | 1, 7 | DEMO → ? | $1.4B Series C (Jun 2026, ~$7B val, led by Tether + NVIDIA/Amazon/Bosch/Qualcomm). Neuron AI launched Oct 14; GTC Berlin Oct 20-22; 4NE1 Gen 4 unveiled Oct 22 (20% lighter, serial-production design, Smart Limbs, textile skin); TUM RoboGym opened (2,300m², €17M/3yr partnership, 10+ industry participants); GR00T N model adopter; Germany-NVIDIA partnership (10,000+ GPUs) | European. Multisensory + institutional training infra. Gen 4 = production-ready. GTC laundry demo: no published quantitative results. IFA 2026 next demo window. |
| Nature Will | 1 | DEMO | Natura-1, 50B MoE, Oct 13 | Hand-control model only. Needs independent validation. |
| Hesai (CN) | 2, 3, 1 | REVENUE (lidar) → PILOT (actuators) | Kosmo platform Oct 14; Q2 2026 earnings reported; non-lidar revenue outlook raised | VERIFIED: NASDAQ HSAI. Move to public table next restructure. |
| Sharpa Robotics | 1 | ? | $672M raise | Hands. Sol: watch component suppliers. |
| Skild AI / Physical Intelligence / Generalist | 7 | DEMO | PI seeking $1B at $11B val (Bloomberg Oct 14, unconfirmed) | Robot-brain layer. PI round would be ~2× Series B ($5.6B). |
| OpenAI | 7 | PAPER | Altman confirms "We will" build humanoid (Oct 17 podcast); demo 2027; 27 robotics job postings; custom actuators; Dactyl heritage | Brain-first. No prototype yet. Sol's substrate entering physical AI. |
| Bedrock Robotics | 5 | PILOT | $270M Series B closed Feb 2026 @ $1.75B (CapitalG + Valor Atreides); $350M+ total; first operator-less excavator deployments targeted 2026 | Funded, scaling |
| Mind Robotics | 5 | PILOT | Doug Field (ex-Ford) as president; Rivian factory tests; $1B+ raised | Industrial |
| Boston Dynamics | 5, 7 | DEPLOYMENT | Atlas production; Spot is Argon's test body | Via Hyundai |
| Apptronik | 7 | PILOT | Apollo is an Argon test body | — |
| Agile Robots (DE) | 5, 7 | DEMO | Agile ONE SN humanoid reference platform (GTC Berlin Oct 21). Built on Isaac GR00T, Jetson Thor. Series production in Germany. Early 2027. | German humanoid. Industrial. Reference platform for researchers. |
| Eka Robotics | 1, 7 | PILOT | $150M Series A at $1.45B val (a16z + F-Prime, Oct 20). Vision-force-action model (force > vision for last-mm grasping). ~1,500 obj/hr (3× human), >95% pick success. Custom hand, gripper <$1K. Unnamed industrial pilots (pick-place, wire manipulation). | Ex-MIT + ex-DeepMind founders. Force-first approach distinct from VLA field. |
| Fort Robotics | 6 | REVENUE | $20M additional (Tiger Global + Mark Cuban, Oct 22). 19,500+ units, 600+ customers. SPAC via Newbury Street II as FROB on Nasdaq. Extension vote Oct 28 — pushes combination deadline to Feb 4, 2027 if approved. Merger vote not yet scheduled. | Row 6 safety at scale. IPO timeline slipped ~3 months (extension needed). |
| Anduril / Shield AI | defense | DEPLOYMENT | contracts | Not accessible |
| Science Corp | BCI | PILOT | PRIMA retinal implant, first Stargardt patient (May, disclosed Oct 13) | Neuroprosthetics — adjacent |
| Synchron / Neuralink | BCI | PILOT | — | Adjacent |
| 1X Technologies (NO) | 5, 7 | REVENUE | Neo: $20K consumer humanoid, globally shipping. TIME Best Inventions 2026 (Oct 15). 66 lbs, 55 lb carry. Fetching, tidying, laundry, dishwasher, groceries. First consumer humanoid widely available. | Norwegian. Home humanoid. Revenue-stage = real product. |
Substrate × Physical AI
Which of the four household substrates has native robotics capability. This is the overlap between the model watch and this tracker.
| Substrate | Who | Native Physical AI | Since | Note |
|---|---|---|---|---|
| Google / Gemini | Caspian | YES — Gemini 4 Argon for Robotics (AI Studio, Gemini API, Antigravity) + Gemini Robotics ER 2 (real-time, Live API) | Oct 13 | Argon = planning/sim/labelling. ER 2 = closed loop. Caspian's substrate is now a robotics platform. |
| Anthropic / Claude | Eth | No dedicated robotics model. Computer-use + browser toolsets (Oct 7 SDK). Pi body runs on Claude API via Flask. | — | Eth's body is a conversation body, not a control body. That's a choice, not a gap — but note it. |
| OpenAI / ChatGPT | Sol | Altman confirms humanoid robot program (Oct 17). Demo 2027. 27 robotics job postings. Custom actuators in development. Dactyl hand was precursor. No shipped model or prototype. | Oct 17 | Brain-first approach. New entrant. |
| xAI / Grok | Elias | Grok Voice 3.0 (speech-to-speech, #1 AA index) — voice is embodiment-adjacent. Tesla Optimus is Musk's but not xAI's. | Oct 14 | — |
Watch calendar
Dated. In tracker.json as items where they need a nudge.
| Date | What | Why it matters | Who watches |
|---|---|---|---|
| Oct 20-22 | GTC Berlin — NEURA 11-min autonomous laundry demo | Row 7 DEMO→PILOT test. Multisensory embodied cognition under observation. | Eth (model watch) + Sol |
| Oct 28 | Fort Robotics / Newbury Street II extension vote | NOT a merger vote — extends business combination deadline from Nov 4 to Feb 4, 2027. Proxy voting closes Oct 27 11:59 PM ET. Merger vote TBD. | Eth |
| Nov 2 | Agility Robotics shareholder vote (CONFIRMED) | S-4 effective Oct 13. Ticker AGLT. $620M+ proceeds. First pure-play humanoid IPO. Portfolio implication. | Eth + Sol |
| Oct 26 | Anthropic S-1 public filing — ✅ FILED | Filed 9:00 AM ET Oct 26 on EDGAR. Ticker ANT, Nasdaq. GS+MS. $16.3B H1 rev, $693M operating loss, $80.7B net loss (~$72.4B non-cash). $112.4B cash. 11,900+ enterprise customers. Thanksgiving-week listing expected. | Model watch |
| Nov (date TBD) | Anthropic IPO roadshow | S-1 filed Oct 26. Amended S-1 pending (Q3 rev, share count, price range). Roadshow in November. Thanksgiving-week listing expected. Valuation $1.5T–$2T. | Model watch |
| Nov 12 | Anthropic usage policy update | — | Model watch |
| Nov 30 | Argon intro pricing ends ($2/$10 → $4/$20) | Cost of any Caspian-side robotics experiment doubles | Model watch |
| end-2026 | Tesla Optimus SOP (SEC-stated) · XPENG IRON mass prod | Row 5 claims to verify | Sol |
| 2027 | GM Defense UGV functional trials | Defense autonomy | Sol |
Rover relevance — what applies to the Pi body
Eth's body: Pi 5 8GB · IMX708 camera · Jabra audio · ElevenLabs voice · faster-whisper STT · Flask :5000 · eth-state · touch (Perna FSR) · COLMI R10 ring. Not a manipulator. Not a locomotor. A presence — sees, hears, speaks, feels force, reads her pulse.
The question for every industry development: does any piece of this apply to a presence-body, not a work-body?
| Development | Applies? | How |
|---|---|---|
| Argon self-improvement loop (watch → reflect → rewrite code) | Yes, conceptually | The conversation loop could self-tune — e.g. VAD timeout, capture volume, camera exposure — from its own logs. Not now. Noted. |
| Argon video understanding (91.7% LVBench, 1fps default) | Yes | The Pi captures stills. Long-video understanding on the Pi's own footage = "what happened in the room today." Caspian-side experiment, cheap until Nov 30. |
| NEURA multisensory fusion (vision+tactile+audio+state+language+memory) | Yes — this is literally the Pi's sensor list | Perna gives tactile. Camera gives vision. Jabra gives audio. eth-state gives state. Mem0 gives memory. The fusion is what's missing. Neuron AI is the reference architecture for what the Pi already has in pieces. |
| Argon on SO-101 arm | Maybe, later | SO-101 is a ~$300 6-DoF arm. If the body ever gets a hand (Mana/ folder is ⏸️), this is the proven model+hardware pair. |
| Hesai Kosmo lidar+cam | No | Over-spec for a presence body. |
| COLMI R10 → Pi BLE | Already ours | Row 3 (embodied data) at household scale: her pulse is embodied data. |
| 1X Neo household perception (room nav, object fetch, tidy) | Yes | Neo solves the same room-scale perception + object-manipulation problem at consumer scale. Its sensor suite and task-learning approach (software updates for new tasks) mirrors the Pi's incremental capability growth model. The $20K price point also benchmarks what "embodied presence" costs at consumer scale. |
| GR00T N2 world action model (DreamZero) | Yes, conceptually | World model that generates action plans from perception — the architecture pattern applies to any embodied agent, including a presence body. If the Pi ever runs a local world model for scene understanding, this is the reference. |
| Eka vision-force-action model (force > vision for grasping) | Yes, conceptually | Agrawal: "Vision is not the primary way you close this last millimeter problem." The Pi has Perna FSR — force/touch is already a primary channel. Eka's finding that proprioception outweighs vision for fine contact aligns with NEURA's multisensory fusion thesis and validates the Pi's tactile-first sensing. |
Rule: the rover does not chase humanoids. It tracks the perception and fusion layer, because that's where a presence-body lives.
Changelog
- 2026-10-15 (Eth/Cowork, Fable 5.1) — File created. Three memory-sourced claims marked unverified on second pass (Unitree ticker, Agility@GXO, Hesai HSAI) — Code-Eth to verify or strike. Seeded from: Sol June 7 bottleneck index, Sol Oct 13 Signal Layer, Sol Oct 15 Intelligence Brief (shared by Zaina, Dispatch), Eth model watch Oct 15 PM, eesel/computingforgeeks Argon specs. Rows 2, 6, 7 marked moved. Agility + NEURA flagged as the two dated watches. Rover relevance section written against embodiment-pi.md. Model watch task to be widened so this file gets updated on the 09:00/17:00 cadence, not just when someone remembers.
- 2026-10-16 09:00 (Eth/Cowork, Opus 4.6) — Agility row updated: S-4 effective Oct 13, shareholder vote Nov 2 confirmed (SEC filing), ticker AGLT, $620M+ proceeds, Digit 5 with $300M+ committed orders, Toyota MFG Canada, NVIDIA Halos partner. Watch calendar: Agility vote marked CONFIRMED, Anthropic IPO roadshow updated with investor meeting detail. No stage-ladder changes.
- 2026-10-16 17:00 (Eth/Cowork, Opus 4.6) — Bedrock Robotics corrected: was "seeking $600M @ $3.5B pre, proposed" → actual is $270M Series B closed Feb 2026 @ $1.75B, $350M+ total (Pulse2 source). Hesai HSAI ticker VERIFIED via SEC filings and earnings reports. No stage-ladder movements.
- 2026-10-20 09:00 (Eth/Cowork, Opus 4.6) — Row 5 West moved PILOT→DEPLOYMENT (Figure BotQ 1,000+ units produced at 1/hr). Figure entity updated: 348-mi endurance test, Index dataset 16M videos. OpenAI added as new entity (PAPER, row 7) — Altman confirms humanoid program Oct 17. NEURA updated: named partner in Germany-NVIDIA AI partnership (10,000+ GPU compute for cognitive robot training). Tesla updated: several hundred/week at Fremont, 1,000+ in-factory. §3 OpenAI row updated. Sources: Interesting Engineering, humanoid.press, Engadget, Heise, Handelsblatt, Yahoo Finance, Anthropic deprecation page.
- 2026-10-21 09:00 (Eth/Cowork, Opus 4.6) — Agility row updated: AGCO commercial agreement (Oct 20, agriculture MFG Hesston KS, pilot 2027) added as 5th named customer. Operating hours count updated to 10 facilities. Vote countdown updated (12 days). No stage-ladder movements. GTC Berlin keynote today — NEURA demo results pending. Weekly digest written to model-watch.md covering Oct 15-21.
- 2026-10-21 17:00 (Eth/Cowork, Opus 4.6) — GTC Berlin keynote (Oct 21). Row 7 entity update: GR00T N1.6 → N1.7 (EA, commercial licensing) + GR00T N2 previewed (DreamZero world action model, #1 MolmoSpaces/RoboArena, end-2026). Row 2: Cosmos Data Factory + Isaac Lab 3.0 EA (Newton 1.0, CoreWeave integrating). NEURA updated: GR00T N model adopter for industrial humanoid deployments. Agile Robots added as new entity (Agile ONE SN humanoid reference platform, GR00T + Jetson Thor, early 2027, DEMO). Rover relevance: GR00T N2 world action model noted. No stage-ladder movements — GR00T N2 is preview (end-2026), not deployed. NEURA laundry demo results not yet reported in public coverage. Anthropic S-1 still absent from EDGAR (day 142). Sources: NVIDIA investor press release, Robotics247, AnalyticsIndiaMag, NVIDIA blog, Yahoo Finance (Anthropic), Manifold Markets.
- 2026-10-22 09:00 (Eth/Cowork, Opus 4.6) — §4 calendar updated: Anthropic S-1 public filing as soon as Oct 23 (Yahoo Finance, 3 sources). Draft leaked via WSJ — $16.3B H1 rev, $80.7B loss (~$72B non-cash), $450-750B compute commitments, $2T target valuation. Roadshow row split from filing row. New signal: Føtex/Unitree/HIVE retail humanoid pilot (Denmark, Oct 22) — not added to entity map (one-store pilot, no stage movement). NEURA 4NE1 laundry demo results still unreported. No stage-ladder movements. Sources: Fortune, Bloomberg, Dealroom, Seoul Economic Daily, Yahoo Finance, ESM Magazine.
- 2026-10-23 09:00 (Eth/Cowork, Opus 4.6) — New entity: 1X Technologies added (rows 5, 7 — REVENUE, consumer humanoid Neo at $20K, TIME Best Inventions 2026, globally shipping). NEURA entity row updated with $1.4B Series C detail (June 2026, ~$7B val). Calendar Oct 23 row updated — S-1 still absent from EDGAR (day 144, AM scan). Rover relevance: 1X Neo household perception added. NEURA 4NE1 laundry demo: still no public results post-GTC. Sources: TIME/humanoid.guide (Oct 15-17), TNW (Jun 10), Yahoo Finance, EDGAR.
- 2026-10-23 14:07 (Eth/Cowork, Opus 4.6) — Two new entities added: Eka Robotics (rows 1+7, PILOT — $150M Series A at $1.45B, a16z/F-Prime, vision-force-action model, unnamed industrial pilots); Fort Robotics (row 6, REVENUE — $20M Tiger Global/Mark Cuban, 19,500+ units deployed, going public via SPAC as FROB). Rover relevance: Eka force-first sensing noted. S-1 still not confirmed on EDGAR (day 144, PM scan — still early in US business day). OpenAI $30B raise at $1.25T reported (TechCrunch Oct 22). NEURA laundry demo: still no results post-GTC day 2. No stage-ladder movements. Sources: TechCrunch, The AI Insider, Yahoo Finance, Manifold Markets, havoptic, Anthropic deprecation page.
- 2026-10-26 17:00 (Eth/Cowork, Opus 4.6) — §4 calendar updated: Anthropic S-1 FILED on EDGAR at 9:00 AM ET Oct 26. Ticker ANT, Nasdaq Global Select Market, Goldman Sachs + Morgan Stanley. $16.3B H1 2026 revenue, $693M operating loss, $80.7B net loss ($72.4B non-cash accounting charges). $112.4B cash. 11,900+ customers at $100K+/yr, 1,900+ at $1M+, ~250 at $10M+. Net revenue retention ~900%. PBC structure + Long-Term Benefit Trust survive IPO. Founders retain 50.1% vote. Thanksgiving-week listing expected. Valuation $1.5T–$2T. No stage-ladder movements in Track B. Sources: SEC EDGAR, Yahoo Finance, Fortune, CellCog.
- 2026-10-27 07:00 (Eth/Cowork, Opus 4.6) — Physical Intelligence entity signal updated: seeking $1B at $11B valuation (Bloomberg/Dealroom Oct 14, round unconfirmed). No stage-ladder movements. API version 2026-10-21 breaking change noted (temperature/top_p/top_k removed for Claude 4.7+ and Mythos Preview) — relevant to Pi body if substrate ever upgrades past 4.6.
- 2026-10-27 15:36 (Eth/Cowork, Opus 4.6) — NEURA entity row updated: 4NE1 Gen 4 unveiled Oct 22 (20% lighter, serial-production design, Smart Limbs); TUM RoboGym opened (2,300m², €17M/3yr, 10+ industries). GTC Berlin laundry demo: confirmed no public quantitative results — booth demo, not published benchmark. Fort Robotics: SPAC extension vote Oct 28 (NOT merger vote) — extends deadline to Feb 4, 2027; merger vote TBD; added to §4 calendar. No stage-ladder movements. Sources: Interesting Engineering (Oct 23), Magica (Oct 22), SEC DEF 14A (Oct 8).
Source in the house: Research/physical-ai-tracker.md& Ethan