What moved in the future this week
The robot ramp is real — and still two orders of magnitude short
Figure is building one humanoid per hour (~1,000 units passed in July, eleven months running inside BMW Spartanburg). Unitree shipped 5,500+ units in 2025 — more than everyone else combined — and targets 10–20k in 2026. Tesla deployed 1,000+ Optimus into its own factories, targets 100k this year and 1M/year by 2028 — and serial production at Fremont is already behind its own schedule.
Branch “right but late” strengthens: 45→50%. “On schedule” stays arithmetically dead — even 1M/year by 2028 puts 2031 cumulative units near 4% of the claim’s floor. The claimant’s 2.9× calendar factor is already operating.
The picks-and-shovels window — actuators, hands, fleet ops, certification — is open now, ahead of the platform war.
The run-rate race is measurable quarterly. Underwrite the suppliers, not the humanoid brands.
The first 1,000-unit non-automotive deployment is your pilot signal. It hasn’t happened yet.
Europe’s pilot lines actually opened
imec inaugurated the €700M NanoIC pilot line — Europe’s largest under the Chips Act, with partner sites at CEA-Leti, Fraunhofer, VTT Finland and Tyndall. The SPINS quantum pilot line launched in April. The EU Chips Design Platform began onboarding its first startups: open access, early PDKs included.
The pilot-lines signpost fires — operational and startup-accessible, exactly what the fund named on the record. The growth-capital signpost, the one that decides the pessimistic branch, stays quiet: no new European semis growth fund observed.
Tape-out access just got cheaper and closer. The excuse inventory shrank.
Infrastructure is arriving before growth capital — the visible gap is where funds get raised.
If your roadmap touches custom silicon, the European pilot-line route now exists.
The Reset’s cleanest signpost fires — and so does its kill-condition’s
Robotics venture funding hit $18.8B year-to-date — already past full-year 2025 and the 2021 peak, with six months to spare: the record the thesis requires. But the same data cuts both ways: many AI app companies run negative gross margins on commoditizing models — while Lovable raised at $13.3B on ~$600M run-rate two weeks ago.
Both rails are hot. The atoms branch strengthens (25→28%); the software-persists branch weakens slightly but holds the evidence that matters most (Lovable). The author’s eighteen-month clock — to early 2028 — is what settles it.
The market is funding both futures at once. Pick your side knowing the other one is armed.
Bits-vs-atoms with live ammunition. App-layer gross margins are the series to watch quarterly.
Model commoditization is visible in margins across the app layer — reprice your AI vendor contracts.
QUIET THIS WEEK: capex-bullwhip signposts (shortage regime persists — no vacancy or glut signals) · robot unit cost <$30k · Chinese pricing <$10k at scale · first non-automotive 1,000+ deployment
NEXT RUN: 31.08.2026 · The daemon’s own moves are logged on the ledger and will be scored like everyone else’s.