Behavioral & negotiation

For Sr Staff, behavioral is > 50% of the signal. Coding is the gate, but the offer (and the level) is decided in design + behavioral rounds.

What Sr Staff behavioral actually probes

• Scope and impact. Did you move a $-meaningful metric? Org-wide or team-only?
• Technical leadership. Have you set technical direction for > 1 team, or only executed within one?
• Influence without authority. Convincing peers / partners / orgs to change course.
• Judgment under ambiguity. "We didn't know what to build" stories.
• Hiring and growth. Did you hire / mentor / promote people? Sr Staff is expected to.
• Failure and learning. Real failure with real cost, and what changed because of it.
• Conflict. Disagreed with a VP / Director / partner team and how it resolved.
• Why this company / why now. Specific to them, not generic.

The 8 stories you must have ready

Write each as a 3-paragraph STAR doc. Practice telling each in 4–5 minutes (not longer — let them ask follow-ups).

STAR template (use literally)

Situation (1 paragraph, 30s):
  - The team / product / quarter
  - The constraint or problem (with a number if possible)
  - Why it mattered to the business

Task (1-2 sentences):
  - What YOU specifically owned (not "we")

Action (the meat, 2 paragraphs, 2-3 min):
  - What you actually did, in technical specifics
  - The tradeoff you made and why
  - Who you had to convince / align with

Result (1 paragraph, 30s):
  - Quantified impact (metric, $, headcount, latency, etc.)
  - What changed downstream
  - What you'd do differently if you ran it again

Company-specific framings

Anthropic

• Behavioral round is heavy. They probe safety judgment ("when have you slowed down a launch for safety / quality reasons?").
• Look for clear thinking, calibrated confidence, low ego. Don't oversell.
• "Why Anthropic" should reference responsible scaling policy, model welfare, interpretability, or constitutional AI — be specific.
• Read Dario's essays (Machines of Loving Grace) and reference one if natural.

OpenAI

• "Move fast" culture. Stories about high ownership, late-night ship-it moments resonate.
• They will ask about your opinion on AGI timelines and your research taste.
• Behavioral often blends with research discussion ("walk me through a project you led, what would you do differently").
• "Why OpenAI" — what part of the mission. Pretraining? Post-training? Alignment? Products?

Thinking Machines

• Tiny team, very high bar. Read all Mira's / John's / Lilian's public writing.
• Heavy emphasis on first-principles thinking. Be ready to defend a contrarian technical view with rigor.
• "Why TML" should be substantive (not "small team where I can have impact" — everyone says that).

Meta-style competitors (Pinterest, Snap, Reddit, DoorDash)

• They'll probe metric-moving recsys/ads experience hard. Have 2-3 specific launches with numbers.
• "Why leave Meta" answer matters. Don't bash. "Wanted broader scope / different problem space / smaller team."

DeepMind

• More research-academic. Be ready to discuss a paper of theirs in depth and have a substantive opinion.
• Behavioral is lighter than Anthropic; technical depth is heavier.

Anduril / Defense AI

• Mission-fit matters. They'll ask if you're comfortable with defense applications. Have a real answer.
• Less pure-research, more deployed-systems. Frame your experience around "shipped to production at scale."

"Why are you leaving [current company]?" — sample framings

Write your version once and refine. Three honest framings to adapt — pick the one that's true for you:

1. Scope ceiling. "I've been Staff for X years on [my area]. To go from Staff to Sr Staff at my current company requires more ladder time. I'd rather earn that level somewhere I'm pushed harder on a different problem."
2. Mission shift. "What I've been doing at scale is well-trodden ground. I want to be on the LLM/AGI frontier where the hard problems are unsolved."
3. Founder energy. "I want a smaller-org bet. The optionality of pre-IPO equity at a frontier lab is a bet I want to take while I can."

Negotiation playbook

Before you have an offer

• Don't share your current TC. If pushed: "I'm focused on what the role is worth in the market, happy to discuss ranges." Then pivot.
• Don't share competing offer dollars early. Mention you're "in late stages with a few labs" — that's it.
• Take recruiter screens for companies you'd never join — practice the dance, get range data.

Once you have offers

1. Get every offer in writing first. No counter until you have the email.
2. Stack offers. 2 onsites in the same window so offers land within ~2 weeks of each other. If one offer is expiring, ask for an extension citing "still in late-stage with two others."
3. Counter every component: base, signing bonus, RSU, refresh schedule, vesting cliff, level. Don't only counter on TC — labs often move on signing or vest schedule when base is capped.
4. Be specific. "Company X offered $ABC, can you match or beat?" If they ask for proof, share the offer letter (yes, this is normal at Sr Staff).
5. Use the right anchor. Levels.fyi p75 for the level + a competing offer. Don't ask for p99 unless you have it.
6. Negotiate level. If you got Staff but the loop felt like Sr Staff, push the recruiter for a level review with debrief notes. Sometimes works, especially at mid-tier.

Numbers to anchor on (rough, 2026 Bay Area, Sr Staff/Staff IC, source: company tier list + Levels.fyi)

Source: aggregated tier-list data, April 2026. Verify on Levels.fyi at offer time — these move quarterly.

Equity calibration

• Anthropic / OpenAI: heavy private equity. OpenAI switched to RSUs (Jan 2026); Anthropic still on a private-equity instrument with periodic tender offers. Discount ~20-30% for liquidity risk vs Meta RSU.
• Thinking Machines: pre-revenue, world-record-Series-A $50–60B valuation Q1 2026. Equity could 5x or 0.5x. Take base + signing seriously.
• xAI / SpaceX merger: heading toward IPO mid-2026 per the company list — equity is closer-to-liquid than other private labs.
• Series-B/C startups (Cursor, Perplexity, etc): assume 10-15% chance of meaningful equity outcome. Don't take a base cut for equity unless you'd bet on the company at a personal cost.

Mock-interview routine

• Coding: Pramp (free, peer-to-peer), interviewing.io ($), or peer from FAIR / DeepMind.
• ML system design: hardest to mock. Find an ex-Meta E7 / Sr Staff at any FAANG. Pay if needed.
• Behavioral: read your stories aloud to a non-technical friend. If they zone out, the story is too long.
• Self-recording: zoom record solo, watch back at 1.5x. Painful, fastest feedback loop.