Accept / reject rubric
I fill this before I feel pressure from an exploding offer. Score each offer 1–5. Numbers stay offline. This page is the template, not the spreadsheet.
North-star warmth is a gate, not a post-sign hobby. If a contract kills ethical optionality, that is a real cost — I score it now, in chapter 1, not after I start.
Must-pass (any No → decline or renegotiate)
| Gate | Y/N |
|---|---|
| Real ownership of agent platform, RAG, governance, evals, or 0→1 AI product — not CRUD rebadge | |
| Level ≈ Staff+ / Sr Staff / Principal IC or true Founding with equity + authority | |
| Comp in good faith for Bay Area Sr Staff AI (compare total + equity to market; offer letters stay out of git) | |
| Geo workable (Bay Area / hybrid / remote / relocate) | |
| Does not block north-star search ethically (non-compete / moonlighting — I actually read them) |
Scorecard (1–5)
| Dimension | Weight | Score | Notes |
|---|---|---|---|
| Ownership depth (platform / agents / evals) | 25% | ||
| Total compensation + equity realism | 20% | ||
| Leveling / path to Principal or equivalent | 15% | ||
| Geo / hybrid fit | 10% | ||
| Learning density toward lab / FAANG / peer Principal | 15% | ||
| Team quality / manager / mandate clarity | 15% |
Weighted total ≥ 3.8 → strong accept candidate.
3.2–3.7 → negotiate or keep competing.
< 3.2 → decline unless runway crisis (document why offline).
I prefer
- Sr Staff / Staff+ AI Platform or agentic platform with org influence
- Founding with real equity, runway, and authority to set architecture
- Clear 90-day mandate tied to production agents + safety/evals
FDE / Applied seats count if the embed is real: discovery, wedge, irreversible tools, evals. A “Founding AI” title on a CRUD backlog does not.
I decline
- “AI” title on a non-AI backlog
- Founding label without equity or decision rights
- Demo theater with no production path
- Comp/level far below Staff+ without extraordinary upside
Multi-offer
Rank by ownership · total comp · leveling · geo · north-star signal — not logo prestige alone. Keep one competing process alive through sign when I can.
What I’d say on a deadline call
I can decide this week if we lock level, scope, and the written package. I’m scoring ownership first, then total comp, then whether I can keep doing this kind of AI work. I’m not optimizing for the loudest logo. I won’t bluff a competing offer I don’t have.
If I’m declining:
The ownership isn’t the work I want to be measured on in 18 months — [CRUD / no equity / demo theater]. I’m going to pass. Grateful for the time.
If they lowball level:
I’m happy to be leveled by your bar. The work I just walked is multi-program architecture and standards other teams reused. If this loop is Senior, we’re misaligned on scope, not on enthusiasm.
If they ask me to name a number first:
I’m scoring Bay Area Sr Staff AI total-comp for this scope — base, equity, sign-on together. I’ll be concrete once I see the written package. I won’t auction myself in the first screen.
If they want me to kill other processes before a written offer:
I can be exclusive once we have a written package I’m ready to sign. Until then I’m running one search honestly. I won’t ghost other loops on a verbal.
After I sign
I do not announce a search inside the company. I keep 2–3 north-star threads warm unless the contract forbids it. First 90 days on the new seat should produce referral-grade stories — isolation, evals, production incidents — so chapter 4 isn’t Lucid-only.
Finish-close checklist: Finish close.
Offer PDFs, exact asks, and recruiter threads stay out of git.