Quality Over Velocity: The Case for Slow AI Coding

Nolan Lawson suggests that instead of using AI to rapidly generate low-quality code, developers should use it as a rigorous multi-agent review tool to find deep-seated bugs. This approach involves using different LLMs to critique PRs, which often uncovers pre-existing issues and forces a more thorough understanding of the codebase. Although this method may reduce development speed, it results in significantly higher code quality and a more disciplined engineering practice.

Key Points

• AI is highly effective at finding bugs and vulnerabilities when used as a multi-agent review system rather than just a code generator.
• Using multiple different LLMs to review the same PR reduces hallucinations and false positives through a consensus or debate mechanism.
• This 'slow' approach to AI coding prioritizes code quality and long-term maintainability over raw output speed.
• Engaging with AI tools to explain code or 'grill' the developer ensures the human author fully understands every change being made.
• The process often leads to fixing legacy issues, which improves the overall health of the codebase even if it slows down the current PR.

Sentiment

The overall sentiment is cautiously positive toward the article. Most substantive comments accept the core argument that AI is valuable as a deliberate review and reasoning aid rather than as a speed-only code generator, and many contributors share similar workflows. The disagreement is meaningful but concentrated around trust, sycophancy, loss of understanding, and whether the process can survive real delivery pressure, so the thread reads as supportive but still skeptical of careless AI adoption.

In Agreement

• Using AI for planning, critique, and repeated review can produce more robust code than asking it to generate a whole feature at once.
• Fresh model sessions and different models are useful as independent reviewers because they find corner cases, security issues, and design flaws the author may miss.
• Small, incremental AI-assisted tasks are easier to review and understand than large generated changes, reducing the risk of review fatigue.
• AI can make developers better by forcing them to explain designs, compare tradeoffs, write tests, and understand a change before merging it.
• Custom reviewer prompts, skills, and adversarial roles can make model feedback more useful by encouraging pushback instead of agreement.
• The slow workflow resembles code review: it can be locally less efficient while improving quality and maintainability across the project.

Opposed

• Some commenters argue that adding more AI review around AI-generated code is a complicated workaround for a tool that still produces weak or untrusted output.
• Critics worry that relying on AI weakens the developer's own mental model, naming judgment, and ability to reason through the system directly.
• Several skeptics see AI coding narratives as overhyped and unstable, moving from faster generation to slower generation without proving durable quality.
• Others argue that models are too sycophantic to be reliable critics unless heavily prompted, and even then may not know when to challenge versus defer.
• Some note that workplace delivery pressure can make careful AI-assisted review unrealistic when management expects more features faster.
• A few commenters remain unconvinced that AI-assisted projects have demonstrated the quality of major mature software built through traditional engineering discipline.