Loading Runway...
Sample Brief · representative example, not a real user
Code-completion and agentic tools merge feature PRs at rates approaching parity with junior ICs (GitHub Copilot 2024; Anthropic Economic Index 2025). The defensible work is on-call judgment, security under load, and the architecture decisions whose blast radius is too high to delegate to a model.
Code-completion + agentic coding tools (Copilot, Cursor, Claude Code) measurably ship feature-implementation and unit-test PRs faster than a junior IC, with merge rates approaching parity (GitHub Copilot 2024 internal study; Anthropic Economic Index 2025).
Hiring managers are explicitly tightening junior pipeline and consolidating into senior + AI-leverage roles.
Architecture, on-call judgment, security, performance under load, and org-specific systems context resist automation.
Signals quoted verbatim from Runway’s career-moves rubric (V11.1). Each references peer-reviewed research, independent benchmarks, or large-N labour-market data — see citations on each move below.
⚠ Placeholder · Sample brief — confidence band shown is a representative midpoint, not an evidence-grounded computation. Take the assessment for a calibrated number.
Verbatim from the career-moves rubric. Each move ships an observable artefact — not self-report.
Move 1 · Defensibility
Why it works · ADRs codify trade-offs in the language of the org's actual constraints; they become reference docs for years.
Time budget
4h / week
Prerequisites
at least one cross-team system in flight
Verifiable artefact
3 merged ADR markdown files in the repo with reviewer sign-offs, linked from the team's onboarding docs.
Citation: n/a
Move 2 · Defensibility
Why it works · accumulated context on one painful subsystem is precisely what an LLM does not have; it makes you the irreplaceable shepherd of that area.
Time budget
4h / week
Prerequisites
manager-level agreement on the scope
Verifiable artefact
written debt inventory with status, monthly burn-down graph, named domain-owner entry in service catalogue.
Citation: n/a
Move 3 · Skill build
Why it works · evals are the actual moat for AI products; teams without them ship hallucinations and regress silently.
Time budget
5h / week
Prerequisites
at least one AI feature in the product
Verifiable artefact
repo containing eval harness with ≥50 examples, scheduled CI run, dashboard publishing pass rate per release.
Citation: Liang et al. 2023 — HELM benchmarking framework
Sample data, not a real user. The tracking layer in the product follows actual move-completion artefacts and recomputes defensibility from observed signals.
Priya completed Move 1 — “Lead 3 architecture decision records (ADRs) on org-specific systems” — over 6 weeks. Tracking-layer artefact recorded: 3 merged ADR markdown files in the repo with reviewer sign-offs, linked from the team's onboarding docs.
Sample numbers. Real movement is computed from your task map + verified move artefacts when you complete an assessment.
Get your own brief
This sample is the average for a software engineer. Your specific tooling, environment, and tenure produce a different exposure profile and a different ordered list of moves. The assessment takes about 5 minutes.