Jul 16, 2026 Fixing the Bug Is the Easy Part. Figuring Out Why It Exists Is What Matters. Most teams fix defects and move on. The same defects come back. fagan-inspectionscausal-analysisdefect-analysisroot-cause
Jul 15, 2026 Pull Request Reviews Catch 15-30% of Defects. The Data Has Said So for 50 Years. Multiple studies across IBM, AT&T, HP, and Microsoft confirm informal code review catches roughly a quarter of defects. fagan-inspectionscode-reviewdatapr-reviews
Jul 14, 2026 A Generic Checklist Catches Nothing. A Structured One Catches 60% of Defects. Most review checklists are copy-pasted lists of good intentions. fagan-inspectionscode-reviewchecklistsstructured-review
Jul 13, 2026 An LLM Can Pre-Inspect Your Code. It Cannot Run the Meeting. Fagan inspections need four to six people and two hours to review 250 lines. fagan-inspectionsllmcode-reviewai-engineering
Jul 12, 2026 Fagan Inspections Found 90% of Defects Before Testing. Then We Stopped Doing Them. Michael Fagan's structured inspection process at IBM caught nearly all defects before they reached a compiler. fagan-inspectionscode-reviewsoftware-historyprocess
Jul 11, 2026 Your Best Reviewer Misses Most Defects. Fagan Measured It at IBM in 1976. Even senior engineers catch only a fraction of defects in unstructured review. fagan-inspectionscode-reviewhuman-factorsdefect-detection
Jul 10, 2026 Most Code Reviews Catch 20% of Defects. Fagan Inspections Catch 90%. Informal code review catches 15-30% of defects. Fagan inspections, a structured 50-year-old process, consistently report 60-90% removal rates. fagan-inspectionscode-reviewdefect-preventionsoftware-quality
Jul 9, 2026 AutoVerus turns 40 hours of proof writing into 3 LLM calls. The trick is knowing when to give up. AutoVerus uses a network of LLM agents to generate Verus correctness proofs for Rust code, automating over 90% of proof obligations through a... model-checkingautoverusrustverification
Jul 8, 2026 LLMs can generate Rust code. Formal proofs are a different problem entirely. Large language models write surprisingly good Rust, but when you ask them for a formal proof they hallucinate invariants and invent syntax that no verifier... model-checkingllmrustformal-proofs
Jul 7, 2026 Waiting for race conditions is a terrible testing strategy Why model checking beats days of production runtime for finding concurrency bugs, and how to apply it to your own code. model-checkingrace-conditionsconcurrencytesting
