L'IA fabrique vraiment du code de m*rde ?!

L'IA fabrique vraiment du code de m*rde ?!

Does AI really produce crappy code?!

Title history
  • L'IA fabrique vraiment du code de m*rde ?! — observed on 20261110-0254 UTC (YouTube)

by Eventually coding · View the channel on PSI

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Eventually coding 👥 6K 📅 July 26, 2026 ⏱ 13 min 👁 4K 📄 expert opinion 🧭 2026-11-10
Available in: English (current) Français

Keywords

AI code generationsoftware craftsmanshipCVE trendstechnical debtengineering practices

Summary

The video examines whether AI is responsible for a decline in software quality, a claim often supported by rising CVE numbers. The author argues that the increase in CVEs since 2017 can be attributed to multiple factors such as increased software pressure, expanded digital surface, bug bounty programs, and geopolitical tensions, not solely AI. He distinguishes between correlation and causation, noting that measuring CVEs is like counting thermometers rather than temperature. The discussion then shifts to the concept of ‘craft’ in software development. The author presents a study by Dave Farley that found no significant difference in maintainability between AI-generated and human-written code, provided proper engineering practices are in place. He emphasizes the importance of the ‘software factory’—a set of tools, tests, guardrails, and processes—rather than just raw code generation. The author shares his personal experience developing three products, noting that AI has allowed him to improve quality by addressing edge cases and reducing technical shortcuts. He concludes that code is becoming a commodity and that differentiation will come from the surrounding engineering infrastructure. While acknowledging that AI lowers the barrier to entry, leading to a flood of low-quality software, he believes that standards will rise due to competition, and craftsmanship will become a competitive advantage.

206 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers valuable insights by challenging a common narrative. The author’s use of the Dave Farley study adds credibility, and his distinction between CVE counts and actual software quality is well-argued. The personal experience with his own products provides practical perspective. The argumentation is solid, though it relies heavily on anecdotal evidence and the author’s interpretation of trends. The claim that AI makes code modification almost cost-free is debatable but presented convincingly. The analogy of the software factory is useful, and the emphasis on the human role in setting up guardrails and processes is a strong point. Overall, the video provides a nuanced discussion rather than a simplistic take.

Scientific Rigor, Source Quality, Title Accuracy

The video references a specific study by Dave Farley, but the study is not formally cited beyond a YouTube link. The author also mentions general trends and his own projects. The scientific rigor is moderate: the CVE analysis is based on graphical trends and speculation, not on statistical testing. The title is provocative but matches the content. The author does not claim to present a systematic review, so the expected rigor is that of an opinion piece. The sources are limited but relevant. The handling of the CVE data acknowledges the complexity of causation. The adequacy between title and content is good, as the video directly addresses the question posed. No comment data was provided, so public reception was not analyzed.

246 words

Title / Content Match

The provocative title matches the content well, directly addressing the question of whether AI produces bad code. The video explores this topic from multiple angles, including CVE trends, software craftsmanship, and personal experience.

Quality & Reliability

6/10

The video presents a reasoned analysis of AI's impact on software quality, citing a specific study and the author's personal experience, but relies heavily on opinion and anecdotal evidence rather than systematic research. The argumentation is coherent and nuanced, yet the scientific rigor is limited due to the lack of empirical data beyond a single cited study.

Chapters

Cited Sources

Contribution & Novelties

The video contributes a pragmatic perspective on the debate about AI-generated code quality. It argues that the real issue is not the code itself but the surrounding engineering practices, advocating for a ‘software factory’ approach. This reframes the discussion from ‘AI versus human’ to ‘how to integrate AI into a holistic development process’. The author’s personal experience with his products provides practical evidence that AI can improve quality if used within a well-structured pipeline.

Pour aller plus loin :

  • Software craftsmanship — The philosophy of software development as a craft, relevant to the discussion of ‘craft’ and its evolution.
  • Technical debt — The concept of shortcuts in software that accrue future costs, central to the quality discussion.
  • Continuous delivery — A software engineering discipline that emphasizes reliable releases, integral to the ‘software factory’ idea.
  • CVE database — The standard for vulnerability disclosure, used as a proxy in the video.

149 words

Radar Profile

The radar profile shows moderate to high scores across information quantity, quality, and technical depth, but a lower overall reliability due to the opinion-based nature and limited empirical evidence. The video is informative and well-structured but not groundbreaking in scientific rigor, reflecting a balance between practical insight and anecdotal support.

Reliability 6/10