Why does screening require login when the original test did not?

The test proved the engine. The platform proves one account serves many apps. Login ties screenings to people, and the same backend serves all three apps with zero re-code.

Where do my CVs go? Is anything stored?

Text is extracted in your browser, scored by the workflow, then the result table is wiped fresh every run by design. Downloads (JSON/CSV) are yours to keep. 5 sample CVs, 2780 to 7523 chars each, never leave your browser as files.

Why is the screening webhook open? Is that not a hole?

Measured decision, documented on-page: stateless endpoint, self-wiping results, near-zero attacker ROI (a spammer only scores their own junk). The n8n editor itself answers 404 on the public door. Hardening scales with threat, not with fear.

Too good to be true? 27 failures says otherwise — which one hurt most?

Exposing the n8n editor to the internet during go-live. Full stop, documented with curl receipts. Also: a workflow edit that never persisted while the tool said saved. Both are on this page with evidence.

Why does it fail when your PC is off?

No VPS by decision, home PC and secure tunnels (:443 auth, :10000 screening). Off means doors closed and the page says so instead of pretending. Same honesty as EXP 010/011/012.

Why React for the portal but rule-code instead of AI for scoring?

React earns its place (upload queue, live pipeline view, auth, settings in one state tree). Scoring stays rule-based 50-20-30 so every point is explainable. AI is used where it helps, not where it impresses.

From Technical Test to Live Multi-User CV Screening Platform on One Shared Auth Backend.

Home Profile Experiments Contact Experiment 001 Experiment 002 Experiment 003 Experiment 004 Experiment 005 Experiment 006 Experiment 007 Experiment 008 Experiment 009 Experiment 010 Experiment 011 Experiment 012 Experiment 013 Experiment 014 Experiment 015 Experiment 016 Experiment 017