# Evidence Beats Assertion Once Assertion Becomes Free To Generate ## Claim When a credential is cheap to claim, the only credible signal left is a verifiable artifact someone else can inspect. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds where the evaluator has more applicants than attention and where artifacts are inspectable at low cost. - Breaks where the artifact itself is cheap to fabricate (a static portfolio page is nearly as fakeable as a résumé line now). - Breaks where evaluation is gated by credential filters before any artifact is seen (ATS keyword screens never reach the evidence). - Breaks where the evaluator lacks the domain knowledge to tell a good artifact from a plausible-looking one. ## Evidence Christopher Penn's argument: once every applicant's materials run through the same handful of models, polish stops carrying information ("every candidate is now a hotel room painting"). His proposed and demonstrated alternative — a public, interactive artifact solving a named problem a specific employer's own disclosures admit having, narrated on video — is expensive to fake precisely because it requires the underlying research a generic model can't supply on its own. The demonstration itself, however, is a single worked example with no hiring outcome attached, which bounds how much weight the claim can currently carry. ## Links - [[3. Resources/Permanent Notes/Raw Provenance and Distilled Guidance Serve Different Memory Jobs]] — because contrasts: that note is about persisting corrections across sessions; this one is about persisting proof across an application — both are responses to the same underlying shift, that generated output is now cheap and verifiable process is not - [[3. Resources/MOCs/Product & Startup Instincts MOC]] — because applies-in: specificity-as-signal is the same logic behind product positioning under commodity AI features - [[3. Resources/MOCs/Evidence & Claims Literacy MOC]] — applies-in: Joins claims literacy on evidence versus assertion when assertion costs nothing. ## Tests / falsifiers - Compare screen-to-interview conversion rates for applications accompanied by a public, company-specific interactive artifact vs. equivalently qualified generic applications. - Find a labor market segment where a public artifact demonstrably did *not* outperform a polished but generic application — would bound the claim to specific hiring contexts (knowledge work, evaluator has bandwidth) rather than hiring generally. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Literature Notes/How to Demonstrate Your AI Chops in Job Hunting (Almost Timely News)]] — the hotel-room-painting diagnosis and the Wealthfront worked example.