Few pieces and no filler. We publish when there is something to stand behind, and each piece also says what it does not solve.
The series is organised into twelve monthly issues so it can be read in order. All twelve were published on 29 August 2026: the month on each issue is its place in the series, not the date it was written, and it is the real date that travels in the sitemap and in each page’s structured data.
Eight design decisions, in order. The first is the one almost everyone skips: you do not ask a model to judge, you ask it to document.
Twenty-four questions in six groups. The useful part is not the questions — anyone has those — but what follows each group: which answers sound good and say nothing.
Written for whoever implements it. Who is in charge, why the request that creates an interview has to be repeatable, how a signature is verified, and the mistake nobody sees coming: the same word with two meanings.
The period is the easy part. The hard part is deletion that actually happens, leaves a record, and does not block itself — which is exactly what happens when storage expires before the promise does.
Consent asked for after the microphone is on is no longer consent. What it has to contain, when to ask, and what to keep as proof that it was given.
Almost everything that makes an automated interview hateful is a design decision, not a technical limit. Seven decisions, with the uncomfortable one last.
Structure reduces one concrete, measurable source of bias. The model brings its own and no instruction removes it. Telling the two apart is the work.
The uncomfortable question does not arrive on decision day: it arrives half a year later, once the team has changed. This is what has to be stored to answer it.
Four conditions that have to hold before two interviews can sit side by side, and what to do with the table once they finally can.
Almost every script is overloaded. Not out of ambition, but because nobody did the arithmetic on what a complete answer costs, or settled what this round has to answer and what the next one will.
Most questions in an interview script separate nobody: they are answered equally well whoever is on the other side. The ones that do separate are written the same way for a warehouse job and for a director. What changes is the bar.
Most competency matrices are a list of nouns with a one-to-five scale. That is not an instrument, it is a form. The instrument is the descriptors.
Thirty minutes: an interview is defined from your job post, walked through the way the candidate sees it, and a report is read with its evidence.