SARS-CoV-2 carries a furin cleavage site at the spike S1/S2 junction that is absent from RaTG13, the BANAL viruses, and other known close sarbecoviruses. Under a pure-natural model this feature must have arisen by unsampled natural recombination/mutation. The DEFUSE proposal shows the specific idea “insert a (human-adapted) furin cleavage site into a SARS-related bat coronavirus spike” was concretely conceived, at the WIV/EcoHealth nexus, ~1.5 years before SARS-CoV-2 emerged. As a likelihood argument: P(observing a virus with this otherwise-unexplained feature | lab-origin at these institutions) is raised by documented intent+capability to build exactly it, so the match increases the posterior for H-5. Two limits cap the update: (1) DARPA rejected the proposal and no leaked document shows the work was performed via another funder, so intent does not equal execution; (2) intent+capability is not construction evidence, and the furin site could still be natural. The proposal is a strong motive-and-means match, not a demonstration that the experiment occurred.
Step 6 verdict — approved, checked
Reconstruction. P1: SARS-CoV-2 has a furin cleavage site at S1/S2 absent from RaTG13, the BANAL viruses, and other known close sarbecoviruses. P2: a 2018 proposal at the WIV/EcoHealth nexus set out to insert furin cleavage sites into SARS-related bat coronavirus spikes. P3 (hidden, load-bearing): documented specific intent plus capability to build feature F raises P(F | that actor produced the virus) without correspondingly raising P(F | no such actor involved). C: the match raises the posterior on H-5.
Traced load-bearing step. P3 is the likelihood-ratio move and it goes through: the proposal is evidence about the lab branch only, so the numerator rises while the natural-origin denominator is unmoved, giving LR > 1 and a posterior increase. Note the correct direction of conditioning — the update runs through P(feature | lab origin), not through the proposal’s own probability of success.
Defeater probe / why not corrected. The two decisive undercutters — DARPA rejected the proposal, so intent is not execution; and the furin site could be natural via unsampled recombination — are both already stated in the argument as explicit bounds on the update, and the statement’s own clause (“the proposal’s rejection and the lack of evidence it was executed bound how much”) carries them. The conclusion asserted is only “raises the probability”, which is precisely the claim immune to those defeaters; it never asserts the experiment occurred. So this is not a proves/refutes overreach needing weakening, and the qualitative direction is all step 7 needs from it — the magnitude is step 7’s Fermi estimate on HC-1’s prior.