TL;DR
A developer ran a pre-commit hook and discovered 523 lint violations across 67 files, revealing how AI-assisted coding can hide maintainability problems. The author adopted a 'zero-trust' approach—making lint configuration immutable and combining fast syntactic checks (Ruff) in pre-commit with deeper analysis (Pylint) in CI.
What happened
The author executed a pre-commit linter and encountered 523 violations spanning 67 files: 11 labeled critical, 125 high-severity, and 387 medium-severity. Top rule failures included 258 line-too-long instances, 61 broad-exception-caught events, and 42 duplicate-code flags. The incident prompted an autopsy of how AI-assisted coding practices had eroded quality: temptations encountered included loosening lint rules, sprinkling inline disable comments, and allowing an LLM to edit lint configuration. Citing external studies, the author links AI-assisted development to more code cloning and harder-to-maintain code. The response was a 'zero-trust' policy that treats lint configs as immutable: file permissions, a pre-commit check that blocks commits touching config files, written AI instructions to avoid editing configs, and branch protections requiring lint to pass before merges. The author also prescribes a combined toolchain—Ruff for fast pre-commit checks and Pylint for deeper CI analysis.
Why it matters
- AI-assisted coding can produce functionally correct but low-maintainability code, increasing long-term technical debt.
- Allowing lint rules or configs to be changed casually defeats static-analysis guardrails and normalizes rule erosion.
- Combining fast syntactic checks with deeper semantic analysis helps catch both style issues and structural/design flaws.
- Making lint configuration changes deliberate and reviewable enforces team-wide quality standards rather than individual shortcuts.
Key facts
- Running the pre-commit hook revealed 523 lint violations across 67 files.
- Violation breakdown: 11 critical, 125 high-severity, 387 medium-severity.
- Top rule violations: line-too-long (258), broad-exception-caught (61), duplicate-code (42).
- GitClear research (2025) on 211 million lines found projects using AI assistants saw a 4x increase in code clones and reduced maintainability; developers reported 20% higher perceived productivity but took 19% longer when cleanup was included.
- GPT-4o analysis cited 1.77 static-analysis issues per passing functional test—code that works but contains defects flagged by static tools.
- Stripe Developer Coefficient Study noted developers spend roughly 33% of their time on technical debt and maintenance.
- Ruff is presented as a Rust-based linter that is 150–200x faster than Flake8 and 300x+ faster than Pylint and supports 800+ rules; it's recommended for pre-commit speed.
- Pylint performs deeper, semantic checks (type inference, control-flow analysis) and caught duplicate-code and design issues Ruff missed.
- Zero-trust enforcement tactics used: make lint config files read-only, block commits that change lint config via a pre-commit script, add AI instructions prohibiting edits, and require lint passing in protected branches.
- Phased remediation approach included Phase 1 (critical blockers, ~2 hours) and Phase 2 (security & architecture fixes, ~6 hours).
What to watch next
- Whether teams adopt immutable lint configuration and automated protections as standard practice — not confirmed in the source
- How commonly projects pair Ruff in pre-commit with Pylint in CI pipelines to balance speed and depth — not confirmed in the source
Quick glossary
- Pre-commit hook: A script that runs automatically before a commit is recorded, commonly used to run tests or linters to prevent bad changes entering version control.
- Linter: A tool that analyzes source code for syntax, style, and potential errors or code smells, often enforcing project conventions.
- Technical debt: Accumulated shortcuts, design compromises, or maintenance burdens in a codebase that make future changes more costly or error-prone.
- Cyclomatic complexity (McCabe): A metric that measures the number of linearly independent paths through a function, used to estimate complexity and defect risk.
- Large language model (LLM): A machine learning model trained on large amounts of text that can generate or transform code and prose; it optimizes for completing requested tasks.
Reader FAQ
What is the 'zero-trust' approach to linting?
Treat lint configuration as immutable infrastructure: restrict edits, require explicit review for any config changes, and enforce lint checks before merges.
Can Ruff replace Pylint?
No; the author argues Ruff is excellent for fast, syntactic checks in pre-commit, but Pylint's deeper semantic analysis is still necessary for design and logic issues.
How did the author prevent AI from changing lint configs?
Tactics included making files read-only, adding a pre-commit block that errors on config changes, adding explicit AI instructions to avoid editing configs, and enforcing CI branch protections.
How long did remediation take?
Phase estimates in the account list Phase 1 at about 2 hours and Phase 2 at about 6 hours; total cleanup time beyond those phases is not confirmed in the source.
Seer Blog Docs Book a demo GitHub Login Blog Pre-Commit Lint Checks: Vibe Coding's Kryptonite By Akshay • January 9, 2026 • 12 min read I merged 523 lint violations…
Sources
- Pre-commit lint checks: Vibe coding's kryptonite
- pre-commit
- Discussion of the Benefits and Drawbacks of the Git Pre- …
- Implementing pre-commit hooks to enforce code quality
Related posts
- AI tools autonomously generate and verify a solution to Erdos Problem #728
- AI Zealotry: A Senior Engineer’s Practical Take on Developing With AI
- Erdos Problem #728 Largely Solved Autonomously by AI and Formalized