DATED BASELINE · 26 OCTOBER 2026
The difference between a design objective, a locally measured result and an independently demonstrated complete system.
CONSOLIDATED BTG FOUNDRY · OCTOBER 26 ENGINEERING RE-RUN
The latest E-drive Foundry full-suite run reports 209 tests run, 208 passed, one optional Windows-to-WSL test skipped, with the cross-kernel test explicitly enabled. This includes the original bounded Studio/DataChip/FarmNode integration, original owner-signed consent and device revocation, source-integrity checks, negative release-evidence tests and authenticated FarmNode job dispatch.
Windows → Linux demonstration: A Windows scheduler opened an actual authenticated private-IP TCP/mutual-TLS connection to a Debian WSL2 agent. The agent executed an original ADAM/BTDU Circuit through its FarmNode worker using Linux CPU-time and address-space limits, returned a kernel-measured result, and rejected replayed, modified and owner-revoked requests. Both kernels operated on one physical computer.
Qualification boundary: No independent off-device physical hosts were attested (0 of 3), no production ENTITY owner identity was used, and no financial instruments or payments were created. The 25-seat Council and signed native HuntAR device qualification remain separate acceptance gates. The release gate still forbids production promotion.
| Evidence | What it demonstrates | What it does not prove |
|---|---|---|
| 18 of 18 integration / source-handoff tests passed | Original Studio importer and Graph Doctor connected to bounded original ADAM/BTDU logical FarmNode execution | Repair of a previously unseen production application |
| 2 of 2 bounded numeric DataChip examples accepted | Two typed compute transformations checked against separate decimal reference calculations | General software reconstruction or autonomous invention |
| 9,047 of 9,047 copied source files verified by SHA-256 | Integrity of consolidated selected files on drive E at that snapshot | 9,047 functional components or completion of asset migration |
| 157 HuntAR files indexed and hashed | Non-executing source baseline for future repair qualification | Native AR overlay, GPS, 2D map and legal checks passing on devices |
| 14 accepted Council responses at last observed check | 10 creators and 4 reviewers with accepted responses towards 25 seats | 25 independent, fully qualified model seats |
Use the actual current iOS source, reproduce defects, freeze acceptance tests, record Council proposals, build a signed app and demonstrate native camera, overlays, GPS, selectable species, legal restrictions and 2D map on real iPhone/iPad.
Repeat the process with a different codebase selected in advance. Measure errors, manual intervention, generalization and regression coverage.
Require 25 distinct responding models, identity and model provenance, independent creator/reviewer outputs and archived evaluation.
Real host enrollment, authenticated cross-machine tasks, enforced CPU/RAM/GPU boundaries, recovery, measured costs and signed receipts.
Show a scoped ENTITY grant, approved Bridge consent, actual FarmNode execution, verifiable technology usage and correct Bit Trade Global records, with financial transfers accounted separately.
Report developer time, build effort, runtime throughput, repair accuracy, false positives, network latency and operating cost using reproducible controls.
For each claim, publish the immutable input manifest, tool and model versions, frozen tests, source hashes, independent oracle, release/build conditions, errors, human changes, measured results and repeat instructions. Unsuccessful cases remain part of the record.
This snapshot describes evidence visible on October 26, 2026, not a real-time statement about concurrent Council qualification. See the historical engineering evidence library for dated reports.
BTG Farms consist of reusable DataChip and Circuit computing units. A compact Farm Family can address millions of logical Farms; each individual operation is executed by the original ADAM/BTDU FarmNode runtime on finite owner-authorized physical computing resources. The October 26 local benchmark demonstrated 1 million addressable Farms, 4 million logical slots, 4 actual Circuit executions and 8 Host receipts. This is addressing and sampled execution, not 4 million concurrent processors or market transactions.
Read the canonical logical Farm design and measured results · Full evidence register
