May 23, 2026
Ctrl+Alt+Del the Code Review?
- -dangerously-skip-reading-code – olano.dev
Devs Are Fighting Over Whether Humans Should Stop Reading the Code Altogether
TLDR: The article argues that companies may soon stop expecting programmers to read every line of bot-written code and instead judge software by written plans and tests. Commenters were split between "that’s the future" and "this is just coding with extra steps," with skeptics saying the key promise still doesn’t actually work.
A spicy new post on olano.dev basically asks the question that makes software people break out in a cold sweat: what if humans stop carefully reading the code and only review the written plan and the tests instead? The author argues that if company leaders knowingly choose speed over hand-checking every line, then maybe the code itself becomes more like machine output—something you run, not something you lovingly inspect. In plain English: let the bot write the guts, and make humans responsible for the blueprint.
The comments? Instant drama. One camp nodded along and said, yes, this is where teams are headed: the spec, or written description of what the app should do, becomes the real source of truth, while the code turns into a disposable artifact. Another camp was not buying the hype. One commenter dropped the party-pooper line of the thread: the idea that software can be automatically checked against a written spec is basically still unsolved. Ouch.
Then came the philosophical hot takes. One person insisted this is old news and even dragged in Elon Musk energy, saying we’re just moving from “how do we build it?” to “what should we even build?” Others pushed back with a brutally simple reality check: if people still have to type detailed instructions somewhere, then maybe this is just coding in a trendier outfit. That became the running joke of the thread—Markdown as the new programming language, with all the same headaches but better branding.
Key Points
- •The article argues that if leadership explicitly accepts the risks of heavy LLM use, software teams may need to stop treating source code as the primary artifact humans must fully read and understand.
- •It says LLMs generate non-deterministic code faster than humans can effectively review, making traditional diff-by-diff code review increasingly impractical.
- •The author states that productivity gains from faster code generation require organizational changes, not just individual adoption, and cites Amdahl’s law to support this point.
- •The article argues that reducing human bottlenecks would require less coordination and bureaucracy, broader autonomy for engineers, and a larger flow of requirements.
- •It proposes standardized Markdown specifications and associated tests as the main unit of project knowledge, with automated checks verifying both test results and code conformance to the specification.