Under your own reading of what each class of evidence is worth, does the adjudicated record support the claim that humanoid adoption is real at scale today?
Decide how much each class of evidence is worth to you, from announcements to signed acceptances and payments, and watch what the record can then support.
This is the model the report was written against: 49 inputs, 75 equations, every one of them carrying the claim ids behind it. It loads at the report’s own case.
The report’s claim, at these settings
Holds
At these settings the report’s claim stands.
The report’s case
On the report’s own case, the supported-belief score is -83.92 score points, and the report’s claim holds.
What the world returns
- Supported-belief score
- -83.92score points
- From announcements
- -1.85score points
- From audited books
- -12.12score points
- From disclosed tender awards
- -3.56score points
- From signed acceptances
- -17.66score points
- From verified payments
- -15.48score points
- From repeat orders
- -18.06score points
- From outcome disclosures
- -15.18score points
Pre-registered configurations
Each of these was written down before the model was run, one of them by the seat whose job was to break the report. Selecting one restores the report’s case and then applies its own settings.
The report’s scenarios (8)
What you can change
Each control opens at the report’s own value. The lighter band on the track is the range of published estimates; outside it the model still computes and everything it touches is marked.
10.00allocation points
The report’s value. Published estimates run 10.00 to 10.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-01, C-06, C-46, C-02.
60.00allocation points
The report’s value. Published estimates run 60.00 to 60.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-15, C-16, C-71, C-17.
50.00allocation points
The report’s value. Published estimates run 50.00 to 50.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-10, C-11, C-59, C-58.
80.00allocation points
The report’s value. Published estimates run 80.00 to 80.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-13, C-22, C-23, C-76.
90.00allocation points
The report’s value. Published estimates run 90.00 to 90.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-13, C-32, C-77.
90.00allocation points
The report’s value. Published estimates run 90.00 to 90.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-14, C-26, C-75, C-34.
70.00allocation points
The report’s value. Published estimates run 70.00 to 70.00, a single evidenced value. The model is defined from 0 to 100.0. Claims C-27, C-28, C-29, C-33.
0fraction of events
0.113fraction of registry rows
0.090fraction of units or of humanoid revenue
Every other input this model takes (6)
2.00independent origins
The report’s value. Published estimates run 1.00 to 2.00. The model is defined from 1.00 to 14.00. Claims C-38, C-46, C-94, C-01.
0.660fraction of projects
1.00fraction of true acceptances
The report’s value. Published estimates run 1.00 to 1.00, a single evidenced value. The model is defined from 0.0034 to 1.00. Claims C-22, C-23, C-76, C-13, C-25, C-91.
0.189dimensionless
The report’s value. Published estimates run 0.186 to 0.196. The model is defined from 0 to 1.00. Claims C-18, C-69, C-15.
62,200RMB per unit
The report’s value. Published estimates run 62,200 to 82,300. The model is defined from 0 to 200,000. Claims C-18.
1.93multiple
The report’s value. Published estimates run 1.93 to 8.66. The model is defined from 1.00 to 20.00. Claims C-16, C-18, C-15.
What it projects, and what it will not
The future, as a range.
A projection here starts from an evidenced anchor in a named year, is computed from named drivers rather than by extending a line, and stops at the horizon of the study it came from. Point estimates about the future do not ship.
ProjectsOne issuer's audited humanoid units, its average selling price, and the gross margin implied by that price against a carried-forward unit cost. The belief core reads this projection in one place, the audited-books support term, under the editorial link declared in section 6.
Does not projectMarket adoption. Humanoid unit demand. Any other issuer. Any share of deployment that is productive. The supported-belief score, which describes the record as it stands and has no horizon.
Base year2025, the nine months to September, at cumulative 3,966 audited units
Horizon2027. The corpus's own forward horizon. The policy register's only true annual output target is Zhejiang's 20,000 units for 2027 (C-66), and the only dated external-customer plans in the record are Boston Dynamics' and XPeng's, both 2027 (C-86). Nothing in the corpus is dated beyond 2027, so nothing is computed beyond it.
How the band is builtThe horizon price band runs from the steeper published segment elasticity, 0.19572183, to the shallower one, 0.18613443, both read off the audited points. The horizon margin band runs from the steeper price decline against the high end of the audited cost envelope, RMB 82,300, to the shallower price decline against the low end, RMB 62,200. The central path is computed at the least-squares elasticity and the last audited cost, and lies inside its band. No point estimate about the future is returned without its band. The at-scale comparison is on an annual basis at both ends: the horizon year's own implied productive units against the 20,000-unit annual output target (C-66). An earlier form of that ratio compared a cumulative stock with an annual target; the cumulative quantity is still computed and printed as implied_productive_units_y2, but it is not the quantity the audited-books support term reads.
BackcastThree published points, none of them out of sample. The central elasticity is the ordinary least squares fit through all three, so the 2023 and 2024 errors are in-sample residuals of that fit and not tests of it. The corpus holds no fourth point on this basis and the specification does not pretend otherwise. The 2025 row is exact because it is the anchor: nought by construction, not by luck, and it is listed as a construction identity so nobody mistakes it for a test. At the central elasticity the curve returns the two non-anchor points within 1.6 per cent. Across the driver range it is not uniformly that good: at the shallower published elasticity the residual reaches minus 2.14 per cent on both points, marginally outside the stated tolerance, and at the steeper one it is exact on 2024 and plus 4.33 per cent on 2023. Both are recorded by validation vectors V-BACKCAST-LO and V-BACKCAST-HI rather than tucked away by narrowing the range. The exactness at the steeper elasticity is also a construction identity and not luck: a curve anchored on 2025 and run at a segment elasticity reproduces that segment's own ratio exactly. A series of three points cannot distinguish an experience curve from several other shapes that would fit as well, and the model claims only that the constant-elasticity form is consistent with every point the corpus publishes. What the corpus would need to make this a test is a fourth published price and cumulative-unit pair, from this issuer or another.
Dated events on the horizon (9)
- 2025-11Figure 02 fleet retired at BMW Spartanburg after a completed pilotC-47
- 2025-11K-Scale Labs shutdown, the one named Western OEM failure precedent on the recordC-87
- 2025-11First bound humanoid insurance policy on the public record, at about RMB 5,000 per robot-yearC-68
- 2026-01China-produced goods gain a 20 per cent price-evaluation deduction in government procurementC-67
- 2026-06Figure 03 successor project announced at the same plant, no metrics publishedC-47
- 2026-06Tesla's located mid-2026 external-sale window lapses with no offerC-85
- 2026-08Record cut-off; whether the late-2025 awards have aged into acceptance is carried openC-23
- 2027Zhejiang's 20,000-unit annual whole-machine output target, the policy register's only true output targetC-66
- 2027Boston Dynamics and XPeng dated external-customer plans, terms unknownC-86
Under the hood
Everything this model is made of.
What this model cannot tell you
- The trust allocation is yours; the evidence is ours, and it does not move. Each class's support share is a measured quantity from the record, not an opinion about the class: the announced class is the fraction of flagship claim lineages that survive origin counting, the awards class is the non-education share of projects, the four verification classes are what their rails actually record, and the audited class is the audited series' own productive flow against the only published annual output target, held at the margin the filing printed. You choose only how much each class deserves your trust.
- The audited class is read through a link this specification chose and the corpus did not. The corpus publishes the productive-application share, the at-scale target and the audited margin series separately. Combining them into one support share, and combining them by multiplication, is an editorial judgement. It is bounded on both limbs to nought and one, and it is printed beside its own counterfactual, e_audited_books_share_only, so you can see what it costs the class or gives it. At the report's own case and at every scenario and hypothesis here it takes support away; at the high end of the unit-growth driver it adds. If you reject the link, read the counterfactual instead and recompute the score from the seven contributions.
- The belief core describes the record as it stands to August 2026. The at-scale limb of the audited class reads a 2027 quantity, because the claim under test contains the words at scale and the only at-scale reference the corpus holds is a 2027 target (C-66). A today verdict reading a dated forward attainment is a judgement, not a measurement, and it is the second half of the same editorial link.
- The score is not a probability. It is a trust-weighted mean of seven measured shares placed on a scale from minus one hundred to plus one hundred, and it means nothing outside this model.
- A nought in this record is a ceiling on what can be known, not a floor on what happened. The model never reads nought located payment remittances as nobody paid. It reads it as a bound, and the bound widens when you widen the share of the market you believe never reaches a register.
- The unobserved-share assumption is applied to public-register census quantities only: awards, acceptances, registry rows and the source sweep. It is never applied to audited filings, because a filing is not a census and grossing it up would be inventing revenue.
- There is no measured disclosure base rate and no measured acceptance publication rate. The corpus says the first outright and declares the question uncovered (C-91), and no source in it publishes the second. Both parameters therefore load at the definitional bound of their own quantity, which is an assumption and is marked as one, and every value you set instead is your assumption rather than a measurement. Both defaults are the settings most favourable to the report's own reading, and the corpus's counterweight says the first of them is certainly too low without saying by how much (C-21, C-75).
- The three published buyer-composition figures do not share a definition. The default is education buyers as a share of calendar-2025 projects; the low end is education buyers in the 2026 half; the high end is academic-plus-public buyers in the same half. The corpus also records its own compilation contradicting itself, 66 per cent against 187 of 292. The model does not resolve that and neither should you.
- The audited series is one issuer's, and its third period is nine months long against twelve-month predecessors. The unit-growth driver's high end inherits that mismatch and the model marks it.
- Cost is not projected. The one audited cost curve rose before it fell and is mix-confounded on the issuer's own account, so the model carries cost forward inside its published envelope and lets it widen the margin band, rather than pretending to a cost path it cannot support.
- The backcast is in sample. Three published points, none of them out of sample: the central elasticity is the least-squares fit through all three, so the 2023 and 2024 errors are residuals of that fit and not tests of it. Three points cannot distinguish an experience curve from several other shapes that would fit as well. The model claims only that a constant-elasticity form is consistent with every point the corpus publishes.
- The projection is about price, margin and volume at one company. It says nothing about whether anybody is buying humanoids to do work. Reading a falling audited price as evidence of adoption is the exact move this report exists to refuse, and the one link from the projection into the verdict runs the other way.
- The relationship between trust and truth is not modelled at all. Nothing here tells you which classes of evidence deserve trust; it tells you what follows if you decide.
- The record behind this model runs to August 2026. Whether the large awards of late 2025 and 2026 have simply not aged into acceptance yet is carried as an open question, not resolved.
- The least defensible arithmetic in the package is compounding the unit-growth driver twice to reach the horizon. The driver's own ends are published, but neither was published as a forward unit-growth rate for humanoids, and a company-wide order-book growth applied to a nine-month unit count and then squared is a proxy on a proxy. Two diagnostics exist so this cannot hide: horizon_units_vs_global_2025 compares the implied cumulative units with the more than 18,000 humanoids produced worldwide in 2025, and horizon_revenue_vs_audited compares implied horizon revenue with the combined audited humanoid revenue of both disclosing issuers. Neither is a gate on the verdict, and that is deliberate: the corpus publishes no threshold at which an implied path becomes incredible, so turning either into a gate would mean inventing one. They are printed and left to you.
- Several outputs are terminal by design and consume nothing downstream. They are diagnostics, reconciliations and band edges, and the generated list of them is printed in section 6 so that a terminal output is a stated choice rather than a discovery. What is no longer terminal is the audited series itself.
- What would need new research: a measured disclosure base rate for capital-equipment procurement, so the unobserved share stops being an assumption; acceptance-notice sweeps for the provinces carrying the largest awards, so the acceptance publication rate stops being an assumption too; a published citation inside the corpus for the rule-of-three bound, which is currently a named method standing outside it; and a second issuer publishing a multi-period humanoid price and cost series, so the experience curve has something to be tested against.
Every equation, in the order they are evaluated (75)
| Output | Expression | Unit | Claims |
|---|---|---|---|
| units_y1 | units_anchor * annual_unit_growth | units | C-15, C-18, C-16 |
| cum_units_y1 | cum_units_anchor + units_y1 | units | C-18, C-15 |
| units_y2 | units_y1 * annual_unit_growth | units | C-15, C-18, C-16 |
| cum_units_y2 | cum_units_y1 + units_y2 | units | C-18, C-15 |
| asp_y1 | asp_anchor_rmb * exp(0 - asp_experience_elasticity * log(cum_units_y1 / cum_units_anchor)) | RMB per unit | C-18, C-69, C-15 |
| asp_y2 | asp_anchor_rmb * exp(0 - asp_experience_elasticity * log(cum_units_y2 / cum_units_anchor)) | RMB per unit | C-18, C-69 |
| asp_y2_band_low | asp_anchor_rmb * exp(0 - asp_elasticity_high * log(cum_units_y2 / cum_units_anchor)) | RMB per unit | C-18, C-69 |
| asp_y2_band_high | asp_anchor_rmb * exp(0 - asp_elasticity_low * log(cum_units_y2 / cum_units_anchor)) | RMB per unit | C-18 |
| margin_anchor_model | (asp_anchor_rmb - unit_cost_rmb_forward) / asp_anchor_rmb | fraction of revenue | C-71, C-18, C-69 |
| margin_anchor_reconciliation_gap | margin_anchor_model - audited_margin_anchor | fraction of revenue | C-71, C-18 |
| margin_y1 | (asp_y1 - unit_cost_rmb_forward) / asp_y1 | fraction of revenue | C-71, C-18 |
| margin_y2 | (asp_y2 - unit_cost_rmb_forward) / asp_y2 | fraction of revenue | C-71, C-18 |
| margin_y2_band_low | (asp_y2_band_low - unit_cost_high_rmb) / asp_y2_band_low | fraction of revenue | C-71, C-18 |
| margin_y2_band_high | (asp_y2_band_high - unit_cost_low_rmb) / asp_y2_band_high | fraction of revenue | C-71, C-18 |
| revenue_y2_rmb | units_y2 * asp_y2 | RMB | C-15, C-16, C-18 |
| asp_backcast_2024 | asp_anchor_rmb * exp(0 - asp_experience_elasticity * log(cum_units_2024 / cum_units_anchor)) | RMB per unit | C-18 |
| backcast_error_2024 | (asp_backcast_2024 - asp_2024_actual) / asp_2024_actual | fraction | C-18 |
| asp_backcast_2023 | asp_anchor_rmb * exp(0 - asp_experience_elasticity * log(cum_units_2023 / cum_units_anchor)) | RMB per unit | C-18 |
| backcast_error_2023 | (asp_backcast_2023 - asp_2023_actual) / asp_2023_actual | fraction | C-18 |
| implied_productive_units_y2 | cum_units_y2 * productive_application_share | units | C-30, C-29, C-18 |
| implied_productive_units_horizon_year | units_y2 * productive_application_share | units per year | C-30, C-29, C-18 |
| scale_gap_ratio | implied_productive_units_horizon_year / at_scale_threshold_units | ratio | C-66, C-30, C-29 |
| horizon_units_vs_global_2025 | cum_units_y2 / global_units_2025 | ratio | C-29, C-78, C-18 |
| horizon_revenue_vs_audited | revenue_y2_rmb / audited_humanoid_revenue_rmb | ratio | C-15, C-16, C-18 |
| at_scale_attainment | clamp(scale_gap_ratio, 0, 1) | fraction of the at-scale target | C-66, C-30, C-29 |
| audited_margin_retention | clamp(margin_y2 / audited_margin_anchor, 0, 1) | fraction of the audited margin | C-71, C-18 |
| s_audited_books | at_scale_attainment * audited_margin_retention | support share | C-15, C-30, C-29, C-66, C-71 |
| e_audited_books | 2 * s_audited_books - 1 | support scale | C-15, C-16, C-30, C-29, C-66, C-71 |
| e_audited_books_share_only | 2 * productive_application_share - 1 | support scale | C-30, C-29 |
| audited_coupling_effect | e_audited_books - e_audited_books_share_only | support scale | C-66, C-71, C-30 |
| collapse_ratio | flagship_appearances_total / flagship_origins_total | appearances per independent origin | C-46, C-01, C-38 |
| corroborated_lineages | flagship_lineages - (flagship_lineages - two_origin_lineages) * clamp(origin_threshold - 1, 0, 1) - two_origin_lineages * clamp(origin_threshold - 2, 0, 1) | lineages | C-46, C-01, C-94 |
| corroboration_credit | corroborated_lineages / flagship_lineages | fraction of lineages | C-46, C-38 |
| appearances_per_surviving_lineage | flagship_appearances_total / max(corroborated_lineages, 1) | appearances per surviving lineage | C-46, C-01 |
| e_announced | 2 * corroboration_credit - 1 | support scale | C-46, C-01, C-06, C-02 |
| disclosure_rate | 1 - unobserved_share | fraction of events | C-91, C-21, C-75 |
| implied_true_awards | award_projects / disclosure_rate | projects | C-10, C-25, C-21 |
| implied_true_registry_rows | registry_rows / disclosure_rate | rows | C-26, C-77, C-91 |
| observed_acceptance_coverage | acceptance_chains / (award_projects * acceptance_publication_rate) | fraction of awards | C-13, C-25, C-22, C-76 |
| implied_true_acceptances | acceptance_chains / (acceptance_publication_rate * disclosure_rate) | acceptance chains | C-13, C-22, C-23, C-76 |
| s_acceptances | clamp(observed_acceptance_coverage * disclosure_rate + unobserved_share, 0, 1) | support share | C-13, C-22, C-23, C-76, C-25 |
| e_acceptances | 2 * s_acceptances - 1 | support scale | C-13, C-22, C-23, C-25 |
| registry_paid_first_share | registry_paid_first / registry_rows | fraction of registry rows | C-77, C-26 |
| payment_ceiling_share | (registry_paid_first + registry_no_evidence) / registry_rows | fraction of registry rows | C-77, C-26 |
| payment_share_reconciliation_gap | payment_verified_share - registry_paid_first_share | fraction of registry rows | C-77 |
| payment_headroom_share | payment_ceiling_share - payment_verified_share | fraction of registry rows | C-77, C-26 |
| payment_rate_ceiling | clamp(max(payment_records, zero_event_bound_numerator) / award_projects * disclosure_rate + unobserved_share, 0, 1) | fraction of award projects | C-13, C-25, C-91 |
| implied_paid_awards_ceiling | payment_rate_ceiling * implied_true_awards | projects | C-13, C-25, C-21 |
| s_payments | clamp(payment_verified_share * disclosure_rate + unobserved_share, 0, 1) | support share | C-77, C-26, C-13 |
| e_payments | 2 * s_payments - 1 | support scale | C-77, C-13, C-32 |
| reorder_rate_ceiling | clamp(max(verified_reorders, zero_event_bound_numerator) / registry_rows * disclosure_rate + unobserved_share, 0, 1) | fraction of registry rows | C-14, C-26, C-77, C-75, C-34 |
| implied_reorders_ceiling | reorder_rate_ceiling * implied_true_registry_rows | rows | C-14, C-26, C-75 |
| e_reorders | 2 * reorder_rate_ceiling - 1 | support scale | C-14, C-26, C-33, C-34 |
| outcome_rate_ceiling | clamp(zero_event_bound_numerator / sources_reviewed * disclosure_rate + unobserved_share, 0, 1) | fraction of sources | C-27, C-28, C-33 |
| implied_outcome_endings_ceiling | outcome_rate_ceiling * sources_reviewed / disclosure_rate | cases | C-27, C-28 |
| documented_ending_rate | sourced_endings / sources_reviewed | fraction of sources | C-28, C-27 |
| e_outcome_disclosures | 2 * outcome_rate_ceiling - 1 | support scale | C-27, C-28, C-29, C-33 |
| e_tender_awards | 1 - 2 * education_project_share | support scale | C-10, C-59, C-11 |
| trust_total | trust_announced + trust_audited_books + trust_tender_awards + trust_acceptances + trust_payments + trust_reorders + trust_outcome_disclosures | allocation points | C-37, C-38 |
| verdict_defined | clamp(trust_total, 0, 1) | one when defined | C-37, C-39 |
| weighted_support | trust_announced * e_announced + trust_audited_books * e_audited_books + trust_tender_awards * e_tender_awards + trust_acceptances * e_acceptances + trust_payments * e_payments + trust_reorders * e_reorders + trust_outcome_disclosures * e_outcome_disclosures | allocation points on the support scale | C-37, C-38, C-92 |
| supported_belief_score | 100 * weighted_support / max(trust_total, 1) | score points | C-37, C-38, C-98 |
| contribution_announced | 100 * trust_announced * e_announced / max(trust_total, 1) | score points | C-46, C-01 |
| contribution_audited_books | 100 * trust_audited_books * e_audited_books / max(trust_total, 1) | score points | C-15, C-30 |
| contribution_tender_awards | 100 * trust_tender_awards * e_tender_awards / max(trust_total, 1) | score points | C-10, C-59 |
| contribution_acceptances | 100 * trust_acceptances * e_acceptances / max(trust_total, 1) | score points | C-13, C-25 |
| contribution_payments | 100 * trust_payments * e_payments / max(trust_total, 1) | score points | C-77, C-13 |
| contribution_reorders | 100 * trust_reorders * e_reorders / max(trust_total, 1) | score points | C-14, C-26 |
| contribution_outcome_disclosures | 100 * trust_outcome_disclosures * e_outcome_disclosures / max(trust_total, 1) | score points | C-27, C-28 |
| flip_state_numeric | clamp((supported_belief_score - flip_holds_threshold) * 1000000, 0, 1) + clamp((supported_belief_score - flip_strained_threshold) * 1000000 + 1, 0, 1) | nought holds | C-37, C-98, C-38, C-22 |
| funnel_acceptance_observed_rate | acceptance_chains / award_projects | fraction of awards | C-25, C-13 |
| funnel_acceptance_implied_rate | clamp(implied_true_acceptances / implied_true_awards, 0, 1) | fraction of awards | C-13, C-22, C-23 |
| funnel_payment_observed_rate | payment_records / award_projects | fraction of awards | C-13, C-25 |
| funnel_reorder_observed_rate | registry_paid_repeat / registry_rows | fraction of rows | C-26, C-77 |
| announced_to_audited_ratio | announced_award_value_rmb / audited_humanoid_revenue_rmb | ratio | C-10, C-15 |
Every input, its evidence and its source (49)
| Input | Report’s value | Published range | Yours to set | Claims |
|---|---|---|---|---|
| independent origins a flagship claim lineage must carry before it counts as corroborated origin_threshold | 2.00independent origins | 1.00 to 2.00 | yes | C-38, C-46, C-94, C-01 |
| share of humanoid volume in productive application rather than research, education or data collection productive_application_share | 0.090fraction of units or of humanoid revenue | 0.040 to 0.100 | yes | C-30, C-29, C-31 |
| share of disclosed Chinese humanoid tender projects whose buyer is an education, research or public body education_project_share | 0.660fraction of projects | 0.555 to 0.761 | yes | C-10, C-59 |
| share of acceptances that actually happened and are published as acceptance notices acceptance_publication_rate | 1.00fraction of true acceptances | one evidenced value | yes | C-22, C-23, C-76, C-13, C-25, C-91 |
| share of deployment-registry relationships with verified payment payment_verified_share | 0.113fraction of registry rows | 0.113 to 0.435 | yes | C-77, C-26, C-13 |
| share of real procurement and deployment events that never surface in any public register unobserved_share | 0fraction of events | one evidenced value | yes | C-91, C-21, C-75, C-77 |
| elasticity of the audited average selling price to cumulative units asp_experience_elasticity | 0.189dimensionless | 0.186 to 0.196 | yes | C-18, C-69, C-15 |
| blended humanoid unit cost carried forward unit_cost_rmb_forward | 62,200RMB per unit | 62,200 to 82,300 | yes | C-18 |
| year-on-year multiple applied to the audited humanoid unit count annual_unit_growth | 1.93multiple | 1.93 to 8.66 | yes | C-16, C-18, C-15 |
| trust placed in the announced capability and demonstration record trust_announced | 10.00allocation points | one evidenced value | yes | C-01, C-06, C-46, C-02 |
| trust placed in audited filings, revenue and unit counts trust_audited_books | 60.00allocation points | one evidenced value | yes | C-15, C-16, C-71, C-17 |
| trust placed in disclosed tender awards trust_tender_awards | 50.00allocation points | one evidenced value | yes | C-10, C-11, C-59, C-58 |
| trust placed in published acceptance chains trust_acceptances | 80.00allocation points | one evidenced value | yes | C-13, C-22, C-23, C-76 |
| trust placed in verified payment trust_payments | 90.00allocation points | one evidenced value | yes | C-13, C-32, C-77 |
| trust placed in verified repeat purchase trust_reorders | 90.00allocation points | one evidenced value | yes | C-14, C-26, C-75, C-34 |
| trust placed in customer-stated pilot outcomes trust_outcome_disclosures | 70.00allocation points | one evidenced value | yes | C-27, C-28, C-29, C-33 |
| disclosed Chinese humanoid tender-award projects, calendar 2025 award_projects | 292.0projects | one evidenced value | fixed by the record | C-10, C-25 |
| fully documented acceptance chains in the 2025 Chinese record acceptance_chains | 1.00chains | one evidenced value | fixed by the record | C-13, C-25 |
| payment remittance records located for any award payment_records | 0records | one evidenced value | fixed by the record | C-13, C-25 |
| verified post-acceptance reorders verified_reorders | 0reorders | one evidenced value | fixed by the record | C-14, C-25, C-26 |
| rows in the deployment registry registry_rows | 62.00rows | one evidenced value | fixed by the record | C-26, C-77 |
| registry rows with verified paid-first status registry_paid_first | 7.00rows | one evidenced value | fixed by the record | C-77, C-26 |
| registry rows carrying no payment evidence either way registry_no_evidence | 20.00rows | one evidenced value | fixed by the record | C-77, C-26 |
| registry rows with verified paid-repeat status registry_paid_repeat | 0rows | one evidenced value | fixed by the record | C-26, C-77 |
| flagship claim lineages in the origin-collapse census flagship_lineages | 12.00lineages | one evidenced value | fixed by the record | C-46, C-01 |
| lineages resting on two independent origins two_origin_lineages | 1.00lineages | one evidenced value | fixed by the record | C-46, C-01 |
| total media appearances across the twelve lineages flagship_appearances_total | 122.0appearances | one evidenced value | fixed by the record | C-46, C-01 |
| total independent origins across the twelve lineages flagship_origins_total | 13.00origins | one evidenced value | fixed by the record | C-46, C-01 |
| public sources reviewed for customer-stated pilot endings sources_reviewed | 250.0sources | one evidenced value | fixed by the record | C-27 |
| named customer pilots with a sourced ending sourced_endings | 2.00cases | one evidenced value | fixed by the record | C-28 |
| numerator of the zero-event upper bound zero_event_bound_numerator | 3.00events | one evidenced value | fixed by the record | C-25, C-14, C-39 |
| audited humanoid average selling price at the projection anchor asp_anchor_rmb | 167,600RMB per unit | one evidenced value | fixed by the record | C-18, C-69, C-15 |
| cumulative audited humanoid units at the anchor cum_units_anchor | 3,966units | one evidenced value | fixed by the record | C-18, C-15 |
| audited humanoid units in the anchor period units_anchor | 3,551units | one evidenced value | fixed by the record | C-18, C-15 |
| cumulative audited humanoid units at end 2024 cum_units_2024 | 415.0units | one evidenced value | fixed by the record | C-18 |
| cumulative audited humanoid units at end 2023 cum_units_2023 | 5.00units | one evidenced value | fixed by the record | C-18 |
| audited average selling price, 2024 asp_2024_actual | 260,700RMB per unit | one evidenced value | fixed by the record | C-18 |
| audited average selling price, 2023 asp_2023_actual | 593,400RMB per unit | one evidenced value | fixed by the record | C-18 |
| low end of the audited blended unit cost series unit_cost_low_rmb | 62,200RMB per unit | one evidenced value | fixed by the record | C-18 |
| high end of the audited blended unit cost series unit_cost_high_rmb | 82,300RMB per unit | one evidenced value | fixed by the record | C-18 |
| lower published segment elasticity asp_elasticity_low | 0.186dimensionless | one evidenced value | fixed by the record | C-18 |
| upper published segment elasticity asp_elasticity_high | 0.196dimensionless | one evidenced value | fixed by the record | C-18, C-69 |
| annual output target used as the at-scale reference at_scale_threshold_units | 20,000units per year | one evidenced value | fixed by the record | C-66 |
| last audited humanoid gross margin, the anchor of the margin-retention limb audited_margin_anchor | 0.629fraction of revenue | one evidenced value | fixed by the record | C-71 |
| humanoids produced globally in 2025 global_units_2025 | 18,000units | one evidenced value | fixed by the record | C-29, C-78 |
| disclosed value of the calendar-2025 tender awards announced_award_value_rmb | 1,810,000,000RMB | one evidenced value | fixed by the record | C-10 |
| audited humanoid revenue of the two disclosing issuers audited_humanoid_revenue_rmb | 1,415,800,000RMB | one evidenced value | fixed by the record | C-15 |
| score at or below which the report's reading holds flip_holds_threshold | 0score points | one evidenced value | fixed by the record | C-37, C-98, C-77, C-59 |
| score above which the report's reading flips flip_strained_threshold | 8.54score points | one evidenced value | fixed by the record | C-29, C-30, C-59, C-77, C-38, C-22, C-91, C-13, C-25 |
Model version 1.1.0, dated 2026-09-02. Every claim id above resolves to a numbered reference in the report, and the report is on the page you came from. Where this page and the shipped package disagree, the package is right: it ships a reference implementation whose 75 equations this page is checked against on every build.
The argument itself is in the report.
The model tells you what the numbers do. The Demo Economy tells you what they mean, what the evidence would not settle, and what would change our mind.
Back to the report