Stabilize and aim up (months 3–6)
When I enter: offer signed (or a clear bridge path) and the day-60 gate is done. Calendar is roughly months 3–6 from search start — I slide it if I signed earlier.
Two jobs that only conflict if I didn’t plan them:
- Become trustworthy on the seat so I get real ownership stories.
- Keep aiming up quietly toward labs / FAANG AI Platform / peer Principal.
The north-star threads already exist. I started them in the war room. This chapter is not “open the climb.” It’s: don’t go quiet, and don’t tank the seat.
People who only do (1) get comfortable and stall. People who only do (2) lose referral-grade wins and walk into chapter 4 with Lucid stories only — which is fine, but thinner than Lucid plus “here’s what I owned in the first 90 days here.”
Success (exit toward chapter 4)
- Written 90-day AI charter agreed with manager — agents / RAG / governance / evals, not “help with GenAI”
- Quiet north-star pipeline: 5–10 warm threads (intros that got a reply, or serious referred apps)
- 3 LinkedIn recommendations asked that can speak to AI platform / FDE embed / production outcomes
- One signal artifact shipped — not a new product brand
- Weekly scoreboard filled (tracker)
- I can tell AegisAI and Enterprise RAG in under eight minutes each without notes — including limitations
Mission in plain language
Stabilize so the next year of interviews isn’t “demo guy with a nice site.” Stabilize so someone inside Anthropic / OpenAI enterprise / Google AI Platform / Meta Agents / a peer Principal org can say: I’ve seen how he thinks about irreversible tools and false confidence — I’ll put a referral in.
Sign
→ listen 2 weeks, then charter (AI ownership in writing)
→ ship one deep signal (docs-first is fine)
→ refresh recommenders away from iOS-era quotes
→ 2 thoughtful intros / week (not spray)
→ chapter 4 when the exit gate is honestly green
Month-by-month
Staff+ first 90 days look like a plan that multiplies other teams, not “ship my first ticket.” I diagnose fulcrums — where a standard, a gateway, or an eval gate other people can inherit.
| Window | What good looks like | What failure looks like |
|---|---|---|
| Days 1–30 on the seat | Map who owns agents, RAG, tools, data; ask “what irreversible AI action scares this org?”; charter drafted and signed; ask 3 recommenders | Hero-ship a chat UI to look busy; accept “AI” title with a CRUD backlog |
| Days 31–60 | Small credible win on charter #1; 2 intros/week continuing (not starting from zero); one system I can demo under their SSO story | Meeting tourism; Substack binge while delivery slips |
| Month 5 | Signal artifact linkable; deepen only two systems in storytelling | Five half-finished Forge toys; catalog counting in panels |
| Month 6 | Gate review — pass → chapter 4; fail → fix one input for 4 weeks | “I’ll start aiming up next quarter” forever |
Time budget (employed)
| Bucket | Share | Reality check |
|---|---|---|
| Deliver on charter | 60–70% | If this drops below ~50% for three weeks, I’m already quiet-quitting toward a bad reference |
| North-star intros / apps / panels | 15–25% | Calendar blocks. Don’t steal from on-call chaos. These threads were already warm. |
| Signal artifact + rehearsal | 10–15% | One artifact, not a portfolio rewrite |
| Substack / brand | ≤5% | Pause the week the charter slips |
Unemployed / still searching after a near-miss close: invert — ~50% north-star conversion, ~30% signal, ~20% bridge apps. Still no new product brands. The failure mode is rebuilding the catalog because rejection stung.
Operating rules
| Do | Don’t | Why |
|---|---|---|
| Negotiate learning charter in writing week 1–2 | Accept “AI” title with CRUD backlog | Panels ask what I owned here; title alone doesn’t answer |
| Tell two deep systems well | Demo-farm the catalog | Labs discount breadth-without-depth |
| 2 warm intros/week with homework | Spray 20 cold apps | Referral portals reward low-effort packages from people who trust me |
| One signal artifact | Five half-ships | One linkable proof beats five README stubs |
| Reuse the same pitch/STARs | New personal brand per company | Per-company cards rot and sound fake under follow-up |
Scripts I need in this chapter
“Why leave Lucid / why this seat?” (30–45s)
I spent the last stretch standing up enterprise AI from zero — agents and RAG in production, with a control plane and evals, outcomes I can defend. I’m joining [seat] because the ownership is real: tools that can do damage, retrieval over messy corpora, a standard another team will actually inherit. I want the next chapter measured by that, not by another prototype.
I don’t trash Lucid. I don’t say “stepping stone.” Bridge seats hear that.
“What will you do in your first 90 days?”
Days 1–30 I’m diagnosing: who can call tools, where retrieval sits relative to authZ, what “good” means for the highest-risk agent path, and which team would adopt a standard if we wrote one. Then I lock three measurable outcomes with my manager — usually one isolation or authZ-before-retrieve path, one HITL/gateway on an irreversible tool, and one eval gate that can fail a build. I’m not optimizing for a flashy demo in week two. I’m optimizing for a control plane other engineers can inherit.
“Are you already looking?” (while employed)
I’m building optionality, not announcing a search inside the company. Timeline is real; I won’t broadcast it on Slack. If the contract says otherwise, I stop outreach and read it.
If pressed in a confidential recruiter screen: truthful about timing. Non-solicit and moonlighting are not optional. When unsure, pause.
“My manager asked if I’m interviewing”
I don’t lie. I also don’t volunteer a search on Slack. If asked directly: I’m exploring what’s next, confidentially, and I’ll give professional notice. I don’t recruit teammates out. I don’t use this seat as a named stepping stone in any other loop.
Weekly rehearsal
- Pitch once aloud — pitch
- Multi-tenant once — short + deep
- One Lucid STAR — pack
- One lab-style trade-off: eval false confidence, cost/latency, or isolation under Strict
Record every two weeks. Robotic cadence shows up faster on playback than in my head.
Exit gate → chapter 4
Open Close or upgrade only when all are honestly true:
- Charter mid-point review passed (or documented wins my manager would repeat)
- ≥1 recommender updated / AI-flavored quote visible or committed in writing
- Signal artifact linkable in a 90-second outreach sentence
- ≥3 north-star threads past recruiter screen or ≥5 warm intros with real replies outstanding
- AegisAI + Enterprise RAG fluent, with limitations, under eight minutes each
If I’m yellow on two items, I fix those for four weeks. I don’t “enter chapter 4” as motivation theater.