The dashboard

Nine numbers that re-price this note

Every phase of the timeline carries "re-price when" triggers. These are the metrics behind them — each with its value as of July 2026 and what movement would mean. If you check back in a year, check here first.

METR time-horizon doubling time

~130 days (~4.3 months) from 2023 onward; suite saturating above ~16h

The capability clock. Settling above 4 months downgrades Phase 1's base case; renewed acceleration is the single strongest upgrade trigger for every phase.

METR Time Horizon 1.1 · Jan 2026

SWE-bench Pro under standardized scaffolding

~52–62% on the standardized leaderboard, vendor aggregates up to ~70% (vs. 88–95% on the saturated SWE-bench Verified)

The benchmark-to-reality gap made measurable. Clearing 80% under standardized scaffolding upgrades Phase 1; stagnation for four quarters points to the plateau branch.

Scale SEAL leaderboard · 2026

DORA delivery-instability correlation

AI correlates positively with throughput but still with increased instability

The verification tax in telemetry. Instability turning neutral-to-positive would signal harness engineering is actually working at scale.

DORA 2025, n≈5,000

Frontier inference cost per completed task

no standardized public metric exists yet — as of 2026 this is a line teams must instrument themselves (e.g. cost per merged PR)

The other half of the thesis: generation is cheaper, not free. Sustained order-of-magnitude drops shift agent-vs-junior unit economics and accelerate every adoption curve; cost plateaus cap fleet-scale automation.

provider pricing · internal telemetry

Outcome-based fee share at the majors

McKinsey ~25% of global fees; Bain ~30% AI/tech-enabled; BCG targeting ~40%

The billable hour's obituary, written in contract terms. Self-reported figures — watch for the first audited disclosure.

McKinsey — Business Insider · Nov 2025 · Bain · 2025 · BCG briefings

Headcount-based contract share (HFS)

49% today; leaders expect 16% within two years; AI-powered delivery 12% → 35%

The best-sourced structural projection in the deck. If 2027 data lands near 16%, Phase 1's consulting track is confirmed; if it sticks near 40%, the Evans calibration wins.

HFS Research, 1,002 executives · Nov 2025

Big Four UK graduate intake

2023→2024: KPMG −29%, Deloitte −18%, EY −11%, PwC −6% (competing dataset runs higher)

The apprenticeship pipeline in one number. Continued cuts make the Phase 2 mid-level shortage a certainty; a rebound would falsify the obelisk's sharpest edge.

Telegraph / Morgan data · Jun 2025

Indian IT headcount & Nifty IT index

TCS −23,460 in FY26; top-5 market cap −46% since Aug 2024; Nifty IT −20%+ in 2026

The market's verdict on headcount-led growth. Stabilization would suggest the repricing is done; further cuts at the majors say the model itself is ending.

FY26 filings · NSE

US developer employment, ages 22–25

down nearly 20% from late-2022 peak; Big Tech new-grad hiring down 50%+ from 2019, new grads just 7% of hires

The junior collapse, measured on payroll and hiring data. This is the input to 2029's mid-level shortage — recovery here would defuse Phase 2's labor squeeze.

Stanford Digital Economy Lab (ADP data) · Nov 2025 · SignalFire · 2025