Nate's career advice moved through four phases across fifteen months. Early 2025 opens with historical reassurance: jobs are bundles of tasks, not fixed titles, and automation has always unbundled and re-bundled work (NASA's human computers, his own dead "E-Commerce Manager" title) rather than erasing it outright. He pairs this with a taxonomy of brand-new "finishing" job families opening up in AI's gaps — roles that stitch together what raw model output can't (Polanyi's Paradox: AI can't encode tacit knowledge humans can't articulate) — and a "seven motions" framework (sensemaking, deciding, creating, collaborating, executing, systematizing, reflecting) for seeing what's actually inside any job title.
Through mid-2025 the advice turns tactical: how to write an AI-honest resume bullet (name the tool, name what the AI did, name your own role, add a measured outcome — vague "AI experience" language is the single most common resume failure), how to run a STAR-C interview (Situation/Task/Action/Result plus Constraints, disclosing AI use rather than hiding it badly), and "spearfishing" — picking 5-20 target companies and building a custom artifact for each instead of mass-applying, because cold response rates were already collapsing toward 2%.
By fall 2025 the tactics collapse under their own success: once AI made every resume, portfolio, and cover letter free to generate, the credentialing system itself broke (his Shannon-entropy framing — infinite cheap signal carries zero information). His response was to try to build actual measurement instead: a five-dimension AI Fluency Assessment, a scored public platform (AI Cred), 17 role-specific qualification prompts, and a junior/mid-career/senior playbook (juniors push from production-task framing toward problem-solving framing and ship public artifacts; mid-career protects domain depth while proactively demonstrating AI fluency; seniors get a temporary "grace period" to learn tools on top of already-valuable judgment).
By 2026 this consolidates into his most durable thesis: a horizontal collapse (every job family converging into one meta-skill — directing AI agents with judgment) plus a temporal collapse (advantage windows compressing from years to months) mean waiting is now the single most expensive career move. Agents are "good at tasks and terrible at jobs" — they lack the organizational memory a human accumulates over years — so the paying skill becomes contextual stewardship: writing the evals and decision logs that encode judgment agents don't have. And because static resumes are worthless as proof, he pushes "transactions" instead — small, verifiable units of finished work, each carrying a four-question explanation artifact (what is this / why this approach / what would break / what did I learn) that travels with the deliverable, plus platforms (Nate's Network, Nate's TalentBoard) built to host that proof.
Key moves
Decompose your actual week into buckets (theatre / commodity / on-the-line / durable, or automate / augment / amplify) before deciding your job is at risk — most people undercount the "theatre" fraction by half.
Write resume bullets that name the specific tool, the task the AI performed, your own contribution, and a measured outcome; tool lists and inflated verbs ("architected," "spearheaded") are the top two failure modes.
Use AI on your own materials only after writing the honest, unpolished version yourself — treat the model as a junior editor sharpening framing, never a source of invented outcomes.
In interviews, disclose AI use proactively with your verification process attached; STAR-C (add Constraints) beats classic STAR because it shows judgment under limits, not just results.
Since resumes and portfolios are now free to fake, build verifiable "process" artifacts instead: decision logs, recorded think-alouds, paid work trials, and a short explanation note (what/why/what-breaks/what-I-learned) attached to anything you ship.
Don't chase a new "AI job" — decompose the job you already have into repeating workflows, prototype AI-augmented versions inside company-approved tools, ship one measurable workflow, then loop in platform/security before scaling.
Treat every "impossible" career obstacle as a skill issue you haven't solved yet, and collapse the gap between saying you'll do something and doing it — AI mainly rewards people who already have this bias to action.
Career-stage plays differ: juniors need public, completed artifacts more than polish; mid-career protects domain expertise while visibly demonstrating AI fluency; seniors should lead with judgment and treat tool-learning as the easy part.
Write evals for anything you deploy, not just prompts — encoding what "must never happen here" into a check is the highest-leverage and most senior-coded AI skill, and most organizations hand it to the least experienced person by mistake.
Mine your own exported data (LinkedIn and beyond) for warm paths before cold-applying; a relevance-and-warmth-scored second-degree connection now outperforms fifty blind applications.
Never paste confidential or proprietary data into an AI tool for resume, interview-prep, or work purposes — mask it or don't use it; one leak is permanent.
The skills that keep surviving every version of this framework: problem framing, taste/judgment, narrative persuasion, glue/alignment work, and contextual stewardship — none of them tokenize cleanly, which is exactly why they hold value.
The posts
2026-05-04 — 55-75% of your week is on thin ice... — A 90-minute method for tagging your last two weeks of work as theatre, commodity, on-the-line, or durable — and redirecting toward what compounds.
2026-03-21 — 55% of employers regret AI-driven layoffs — Agents ace isolated tasks and fail whole jobs because they lack organizational memory; the human role becomes writing the evals that catch it.
2026-01-22 — Why High Agency is a Non-Negotiable in 2026 — AI removed the entry-level "training rungs," so internal locus of control plus AI fluency replaces the traditional career ladder.
2026-01-16 — Why it's time to escape the application pile — Build your own AI-queryable portfolio interface, including a bidirectional fit tool that tells bad-fit employers not to bother.
2025-12-02 — PMs Have It Worst in the AI Era — PM is hit from four directions at once (asset generation, probabilistic products, glue-work automation, rising technical bar); the fix is technical fluency, meaningful work, intuition, and alignment.
2025-11-26 — Nobody Could Measure AI Skills, So I Built AI Cred — Launches a public scored AI-fluency platform because tool certifications aren't competency and LinkedIn's "proficient in AI" is meaningless.
2025-11-04 — I've Talked to Hundreds of Companies About AI & Jobs — What leaders say privately vs. publicly, mapped to distinct plays for juniors (production-to-problem-solving), mid-career (domain over skills), and seniors (a temporary grace period).
2025-10-21 — I Reverse-Engineered AI-Native Hiring — AI made resumes free to generate and therefore worthless as signal; the fix is verification (process, work trials, capability maps) over more polished credentials.
2025-10-13 — Which AI Job Actually Fits You? 17 Prompts — Role-specific qualification assessments across 17 AI job titles, because generic "learn AI" advice ignores wildly different entry requirements.
2025-09-22 — The Complete AI Interview Guide — STAR-C storytelling, an AI-maturity tiering system for hiring managers, and why expensive "answer the interview for you" tools get caught while cheap "organize your thinking" tools don't.
2025-09-20 — How to Break into Tech in 2025 — Real junior success stories built on extreme niche focus ("one of one") plus a list of companies still actively hiring AI-native juniors.
2025-07-10 — How to Get an AI Job in 2025 (Beyond OpenAI & Big Tech) — Argues Series-A companies beat Big Tech for equity upside, and introduces "spearfishing" — deep research plus a built artifact for a short target list instead of mass applications.
2025-05-30 — The Kids Will Be OK: How to Skip the Jobs AI Doom Spiral — Frames the Amodei-vs-Orosz jobs debate as irrelevant to individual strategy; the safe bet under either scenario is complex problem-solving with high agency.
2025-03-29 — Finishing What AI Starts — Names eight new job families (LLM-app finishers, human-AI collaboration specialists, agent managers, and more) emerging directly from AI's Polanyi's-Paradox gaps.
2025-02-14 — Help My Job is Changing! — Product, engineering, and design roles blur together across five industries; closes with a blunt FAQ on junior hiring, wages, and broken promotion ladders.