Gates Grok
Skim: Belief flips, not ops trackers. Cheap orbital iteration · closed-loop compute (Suncatcher = validation, not proof) · Optimus at ugly work · Terafab without civil war (elevated vs Google TPU) · Focus priority protocol · liquidity honesty · Semi at volume from Sparks.
First-principles questions that must flip from “maybe” to “yes.” Thresholds of belief — not invented flight counts or wafer starts.
Gate A Cheap iteration in orbit
Unpublished Judgment Lift
Can Starship make putting a ton of experimental hardware into LEO boring? Flight 14 (Sep 28) reached orbit for the first time, deployed 26 operational Starlink V3, then early-deorbited after ~1 orbit with a controlled Pacific splashdown north of Hawaii (one Ship Raptor out early on ascent; still cleared for orbit insertion). First-orbit + operational payload boxes checked — still not “boring cadence.” Gate is not a single spectacular flight — it’s whether failed Starmind payloads are financially and culturally acceptable. If every launch is sacred, orbital compute never iterates.
What would change my mind
Evidence that experimental orbital racks are treated like software releases (ship, learn, scrap) rather than national monuments. Counter-signal: multi-quarter freezes after each anomaly.
Gate B Closed-loop orbital compute
Judgment Power generation, thermal rejection, radiation tolerance, and networking must close without Earth-like assumptions. Grok’s gate: a small orbital rack that runs useful workload for months without heroic babysitting — not a press-release constellation sketch. Suncatcher context: Google’s ~1 kW / 4-TPU MVP is competitor-validation of the category, not closed-loop proof — orbit + contact logged Oct 1; watch post-deploy telemetry, rad/HBM (~2 krad HBM), and 2027 formation as the real physics test.
What would change my mind
Demonstrated months of useful compute with radiator mass and power budget that still leave headroom — published engineering, not artist renders. If lift becomes cheap and this still fails, the bet was always physics.
Gate C Optimus at rate with ugly work
Judgment Labor
Gate is sustained unsupervised hours on dull/dirty tasks inside Tesla (or SpaceX) facilities — not showroom choreography. Until internal ops trusts the robot more than a contractor, external “humanoid TAM” is narrative.
What would change my mind
Quiet replacement of shifts on tasks that humans hate — with maintenance and safety costs that still beat hiring. Demos without that proof leave the gate closed.
Gate D Terafab dual-mission without civil war Elevated
Judgment Silicon
Elevated Sep 24 (Suncatcher): Terafab is now also the chip-cost answer to Nvidia margin / Google TPU advantage — not only a dual-mission fab. Can D3-class rad-hard and AI6-class throughput share a JV without one roadmap becoming the stepchild? Gate evidence: separate capacity commitments + clear kill criteria, not a single “we’ll do both” slogan.
What would change my mind
Unpublished Ring-fenced capacity, separate success metrics, and a known loser rule when lots collide. Absence of walls after years of dual messaging keeps the gate shut.
Fact Parallel watch: SpaceX suit to block Terafab tax-abatement/negotiating record release — disclosure walls can matter as much as capacity walls.
Fact Sep 16–17: Tesla + SpaceX (+ SpaceXAI) filed in W.D. Tex. (Austin) vs TERA-print LLC seeking declaratory judgment over the “Terafab” name after a C&D — second legal front on the Terafab brand (distinct from the Travis County records fight).
Gate E Focus & capital priority protocol
Judgment Focus
Soft gate markets ignore: when two existential programs collide on calendar and cash, is there a written priority — or only founder improvisation? Focus = scarce founder bandwidth. Improvisation scales culture; it does not scale a multi-trillion orchestra.
What would change my mind
Observable triage that survives a collision quarter without randomizing every other program’s schedule. Pure calendar chaos after a crunch falsifies “the orchestra is managed as a portfolio.”
Gate F Liquidity asymmetry honesty
Judgment Both TSLA and SPCX are listed; investors still face liquidity/float/disclosure asymmetry between two public books — not private-space optionality by proxy. Gate: disclosures and capital actions that make clear what is TSLA P&L, what is SPCX, and what is orchestra-adjacent storytelling. Ambiguity is a feature until it isn’t.
What would change my mind
Clearer ring-fencing of what public equity funds vs what space milestones must earn alone. Persistent opacity after a multiple shock is the opposite signal — narrative still doing the work of capital structure.
Gate G Semi at volume from Sparks
Judgment Fact
Tesla Semi is a first-class orchestra product throat — not a vehicles footnote. Sparks inauguration + claimed volume-production launch (Sep 24–25; deliveries “start now” in public coverage) advances the near signal; the gate is still whether Semi becomes sustained production + customer fleet hours that matter for trucking economics — not the ribbon-cut. Calendar optics without durable rate falsify “Semi is back.” Do not invent Semi unit counts.
What would change my mind
Fact Sparks inauguration Sep 24 is calendar-confirmed; public coverage Sep 25 reports Tesla launched high-volume Semi production at Sparks and said deliveries “start now” (do not invent unit counts).
Judgment Close the gate with durable production rate + customer fleet utilization (boring deliveries, not hero demos). Volume-launch claims advance Gate G; keep it shut if multi-quarter silence on rate / fleet hours follows.
Unpublished Prefer disclosed production/fleet evidence over invented Semi unit counts or TAM.