Legg Nature paper flags Python array-index error in Microsoft TGP code and omitted transport datasets

Legg re-examined the raw differential-conductance matrices Microsoft withheld from its 2025 peer-review package. The TGP plotting routine compared bias values by array position instead of calibrated voltage, producing incorrect particle-hole symmetrization. When corrected, multiple additional regions satisfied the protocol threshold but showed gap-closing signatures inconsistent with topological superconductivity. Raw line-cuts outside the single published region exhibit disorder-induced sub-gap states exceeding 50 µeV. Microsoft’s February 2025 announcement and June 2026 Majorana 2 reveal both relied on the same TGP output. Earlier retractions of Microsoft’s 2018 and 2021 nanowire papers followed similar disputes over gap identification. Legg’s finding aligns with independent analyses of InAs-Al devices at Delft and Copenhagen, where zero-bias peaks routinely arise from disorder rather than Majorana modes. The episode repeats a documented pattern: code-level implementation details and full datasets remain unavailable until post-publication scrutiny forces release. No independent group has reproduced a quantized 2e²/h conductance plateau under the magnetic-field and gate conditions Microsoft reported. Operational roadmaps predicated on topological protection therefore rest on an unverified signal-extraction pipeline. Microsoft has not released corrected TGP source or the withheld transport matrices. External replication attempts using open-source nanowire simulators are scheduled for 2027; failure rates above 70 % would falsify the topological-gap claim within the published parameter space.