Key Takeaways

  • Modal treats code review as optional, allowing engineers to ship directly to production unless they specifically request a second pair of eyes.
  • Bernhardsson argues that mandatory peer reviews act as an escape hatch for personal accountability when software breaks in production.
  • Modal equalizes prestige across the team by calling everyone a Member of Technical Staff, pulling former engineering managers back into hands-on programming.
  • AI agents now build effective test suites by acting as adversarial devil's advocates against new features.
  • Investing in continuous delivery, automated tests, and decoupled systems replaces human gatekeeping without compromising reliability.

Mandatory Code Reviews Kill Personal Accountability

Every engineering organization tells itself the same story about code reviews: having another person approve a pull request guarantees software quality. Erik Bernhardsson disagrees. At Modal, the cloud platform running serverless infrastructure for AI workloads, code reviews are optional.

“My view is like code reviews in large part is like people in many cases trying to just escape the accountability of merging things into production,” Bernhardsson said. When three people sign off on a pull request, nobody truly owns the outage. The author blames the reviewer for not spotting a bug. The reviewer blames the author for writing flawed logic. Both blame sprint deadlines.

If your development velocity freezes because someone has not clicked an approval button on a documentation change or an internal update, you have an infrastructure problem disguised as a quality control process. “If you have that culture, then to me that's a sign that you need to invest way more in CD and CI,” Bernhardsson noted. “Don't wait for a human to look at it. Especially in the era of AI, why do you need a human to look at every piece of code? You don't have to.”

Why Agents Write Better Test Suites Than Humans

Modal replaces peer sign-off with technical guardrails: continuous integration, continuous delivery pipelines, decoupled services, rate limits, and automated testing. Historically, writing comprehensive test suites was the weak link in fast-moving engineering teams. Human engineers dislike testing edge cases, while early language models struggled with rigorous assertions. That dynamic has flipped.

“Agents are actually really good at writing tests these days,” Bernhardsson said. “A few years ago I thought that was one of the major things missing in agents. Now they're actually quite good because it takes a little bit of need to step out of what you're trying to do for a second and be almost a little bit like devil's advocate to write good tests.”

An AI agent does not share the author's blind spots. It can interrogate an endpoint with weird parameters, generate adversarial inputs, and catch edge-case regressions before code reaches users.

Stripping the Prestige Out of Management

High deployment autonomy requires senior builders who stay close to the code. Modal flattens corporate hierarchy by eliminating the status gap between people managers and hands-on individual contributors.

“We make it very clear to everyone at Modal that manager or IC, there's no difference in prestige,” Bernhardsson explained. “Everyone is just a Member of Technical Staff. It's about ultimately your contributions, and we see a lot of ex-managers becoming ICs.”

AI coding tools make individual contribution fast and rewarding again for leaders who previously moved into management to escape tedious boilerplate. When an experienced engineer can prototype and ship an entire production service over a weekend with agent assistance, middle management loses its appeal. Modal builds around that reality: hire senior technical talent, remove process friction, provide automated testing safety nets, and let them push directly to production.

What to Do With This

Audit your team's pull requests from the last two weeks. Drop mandatory reviews for documentation, internal admin tools, and isolated bug fixes by the end of this week. Reallocate that saved engineering time into automated CI test coverage and let agents generate your adversarial edge cases.