Track F — Behavioral at Staff Altitude

Reported: technical leadership, architecture decisions spanning teams, driving consensus under pressure, concrete tradeoffs rather than soft-skills answers (../../research/source-report.md rows 33–36), plus a recruiter-screen question about where AI is headed (row 4).

And the finding that should change how much time you give this: reported sources name the values/culture round as the leading failure point at a peer lab. For companies with technical bars this high, that is a remarkable claim — and senior engineers systematically under-prepare here because it does not feel like real work.

→ Study guide: WARMUP.md — all twelve story categories with worked model answers at staff density, the probe playbook, and the forward-looking answers written out in full.

Table of Contents

• Why This Is Graded at Staff, Not Senior
• DTAO, Not STAR
• The Story Bank
• Required Story Categories
• Probe Lists
• The Forward-Looking Questions
• Drill Set
• Failure Modes
• Self-Assessment Rubric
• References

Why This Is Graded at Staff, Not Senior

AI-lab levelling is compressed. Reported consensus: an OpenAI level maps roughly one level higher than the same number at Google or Meta, and L5 — titled "Senior" — carries Staff-equivalent scope (../../research/findings.md).

The practical consequence for every story you tell:

The reliable test: if nobody in the story disagreed with you, it is not a Staff story. Consequential decisions attract disagreement. A story with no opposition is either not consequential or you have edited the opposition out — and interviewers probe for exactly this.

DTAO, Not STAR

STAR front-loads Situation — the least interesting part — and buries the decision. It was designed for interviews that wanted to know whether you were a good teammate. This round wants to know how you think.

Context goes in a clause, not a paragraph: "On the multilingual ranking pipeline, we decided X." That is enough. If they need more they will ask — and them asking is a good sign, because it means they are engaged rather than waiting for you to finish.

The Story Bank

Twelve to fifteen stories from your actual history. I will not invent them, embellish them, or let a Senior-scope story be presented as Staff-scope.

The raw material I need from you, per story — bullets are fine, prose is not required:

1. What was the decision? (one sentence)
2. What constraint made it hard?
3. What alternatives did you seriously consider?
4. Who disagreed, and what did they want instead?
5. What did you do to get alignment?
6. What was the measured outcome?
7. What did you get wrong?

Your history — multilingual search and recommendation, media streaming, networking, enterprise infra, cloud — is unusually rich for this. Ranking pipeline redesigns, index-serving migrations, embedding infrastructure decisions, and cross-org platform migrations are all naturally Staff-altitude if you write the tradeoff rather than the tour.

Each story lands in stories/NN-slug.md with: the DTAO write-up, a 90-second spoken version, a 2-minute version, the tags it covers, and its probe list.

Required Story Categories

Every one must be filled. A gap here is a gap the interviewer will find.

Category 2 deserves its own note

Four ways it fails, all of them visible from across the room:

What works: state your position as strongly as you actually held it, state theirs fairly enough that they would recognize it, say what decided it and whether the process was sound even if the outcome was not, and say how you behaved after losing — committed or sandbagged.

Then give your honest current read. All three of these are strong:

• "They were right, and here is what I had not weighted properly."
• "I still think I was right, and here is the evidence that has accumulated since."
• "We were both solving the wrong problem."

Only performed humility is weak.

Probe Lists

Every story needs three follow-ups written out — the questions an interviewer asks to test whether the story is real. Generic examples; each story gets its own specific set:

1. "What was their strongest argument?" — the single most discriminating probe in the round. If you cannot produce a strong version of the opposing case, you never engaged with it, and the whole story becomes suspect.
2. "What would have had to be true for the other option to win?" — tests whether you modelled the decision or pattern-matched it.
3. "Who else was affected that you did not mention?" — tests scope honesty.
4. "What did that cost the other team?" — cross-team decisions always cost someone.
5. "How long did it take, and how much of that was the disagreement?" — tests whether the consensus story is real.
6. "What did you measure, and how did you know it was not a coincidence?" — tests rigour.
7. "What would you do differently?" — the answer must be specific, not "communicate more."

The Forward-Looking Questions

Written answers, rehearsed weekly, kept in forward/. Company-specific material lives in ../../research/company-brief.md.

The falsifier is the move almost nobody makes. Ending "where is AI headed" with "here is what would change my mind" converts an opinion into a position, and it is the clearest available signal that you actually think about this rather than reciting.

Drill Set

Failure Modes

Self-Assessment Rubric

Hire-bar translation

References

• ../../research/company-brief.md — mission digest, talking points, the three questions you ask them
• ../../diagnostics/d4-behavioral.md — the diagnostic prompts
• ../../diagnostics/ANSWER-KEY.md — the worked weak/strong contrast
• ../../mocks/README.md — the weekly scored mock
• Larson, W. Staff Engineer: Leadership Beyond the Management Track. — the archetypes and what scope means
• Reilly, T. The Staff Engineer's Path. O'Reilly, 2022.
• Fournier, C. The Manager's Path. — useful for the influence-without-authority chapters
• OpenAI. Charter. https://openai.com/charter/
• Anthropic. Core Views on AI Safety. https://www.anthropic.com/news/core-views-on-ai-safety