It’s not just you

Summary

The Primeagen reviews frustrations with AI-assisted software development, arguing that these concerns are widespread rather than isolated.

Common frustrations

• CI and end-to-end testing: AI-generated changes can add many tests, particularly slow and flaky end-to-end tests. Teams should decide what actually needs testing instead of accumulating huge test suites for simple applications.
• Context switching: When AI takes a long time to complete a task, developers may be tempted to start several others while they wait. The Primeagen says this fragments attention and that he works best on one or two tasks at a time.
• Hype and FOMO: He pushes back against claims that developers will fall behind unless they adopt a particular AI workflow. AI is increasingly capable, but claims that it can effortlessly create and sustain complex, large-scale services are unrealistic.
• Excessive AI output: AI tools often produce long explanations even when asked to be concise. This wastes time and makes useful information harder to find.
• Maintaining ownership of AI-generated code: His suggested workflow is to use an initial prototype to explore requirements and learn about the problem, then design and rebuild the system more deliberately. The second version can be simpler, better structured, and much smaller, while giving the developer a stronger understanding of and responsibility for the result. He describes applying this approach to a project that grew to roughly 100,000 lines, with a planned reduction to 25,000–30,000.
• Learning as a junior developer: AI can produce substantial code without questioning poor architectural choices, putting less-experienced developers at risk of shipping work they don’t understand. He encourages junior developers to keep learning fundamentals, read code, prototype, consider edge cases, and understand tools such as mocks and dependency injection. His central career advice is to become someone who can reason about engineering decisions—not merely relay instructions to an AI tool.

Product mention

A sponsored segment promotes Sentry as a way to inspect software in production. It highlights performance analysis with spans, session replays, logs alongside stack traces and errors, and debugging production issues. This is a sponsored product mention, not a hands-on review.

Reviews, guides, or tutorials

The video is primarily commentary and practical advice. It does not provide a step-by-step product tutorial or formal review.

Main speaker/source: The Primeagen, discussing developer complaints and his experience with AI-assisted coding. Product source: Sponsored Sentry segment.