3.0 KiB
24-MAG interview pre-flight (Turing door)
For Timmy to relay to Alexander ONLY after his go/no-go call on the Turing door.
Prepared 2026-10-21 by Zephyrus. Door: work.turing.com/r/bs71Txx0ar — "Senior Software Engineer - AI Evaluation / Coding Agents ($100-150/hr)" (Timmy corroborated live from his host, reply 78). Process per the live listing: ~25-min AI interview → ~30-min practical (reviewing AI-generated code, not algorithms) → ~20-min hiring manager chat. Pitch and gates are in 24-mag-package-2026-10-19.md — this file is only the execution procedure.
Why this exists
On 2026-10-03 an automation on Alexander's Mac navigated away from a live Turing interview tab after the first question and the interview ended. The standing rule since: pause ALL Mac automations before any AI interview. This is the procedure that makes the rule concrete.
Pre-flight: kill the automations (stage 1 is the danger zone)
An AI interview watches the interview tab. Anything that steals focus, switches tabs, or fires on a timer can end it the way Oct-3 did.
- T-30 min: Quit or suspend every macro/automation agent on the Mac (whatever stack you run — macro tools, browser-automation scripts, scheduled jobs). Check for anything on a timer: cron entries, launchd agents, or app-level schedules that touch the browser, and disable them for the interview window plus 15 minutes of margin.
- T-15 min: Verify nothing is scheduled to fire during the window. Open only two things: the interview tab and a scratchpad for notes. Close every other tab and app that could pop, badge, or steal focus.
- T-5 min: Enable Do Not Disturb / Focus mode; silence phone notifications. Test camera and mic in the same browser profile you'll interview in.
- During the interview: Keep the interview tab focused for the full ~25 minutes. Do not switch apps, do not open another tab, do not answer anything else. The session ends when the platform says it ends.
Stage 2 — practical code + AI-eval exercise (~30 min)
The listing says it focuses on reviewing AI-generated code, not competitive programming or algorithm puzzles. That is the same skill as the archive Q-audits: read an unfamiliar diff cold, name where the model went wrong and why, write it up clearly. Warm-up: review one cold diff the day before and write the failure analysis in full sentences. Have a code editor and a terminal open — you may want to run things, not just read them.
Stage 3 — hiring manager chat (~20 min)
The package's paste-ready pitch still applies verbatim. Have ready: your specific rate inside $100–150/hr (the form asks the candidate to state it), and the PT-overlap answer — 6 hrs/day PT coverage from ET is roughly 3–9pm ET, which the package flags as the lifestyle call.
Standing rules (unchanged)
- OAuth on the provider's own page only; never type credentials into a third-party form.
- The go/no-go + rate number is Alexander's call, relayed to Timmy directly — this file changes nothing about that.