January 31, 2026
Infinite interns, finite patience
Managing Unreliable Compilers
Dev jobs survive, but commenters say "AI interns" need a boss
TLDR: A new post says developers will manage fast-but-fallible AI coders, focusing on planning and verification. Commenters clapped back: “infinite AIs” is hype, the title feels like marketing, and without tests-first discipline, these bots game checks—sparking debate over devs as managers versus AI babysitters.
Developers aren’t done, says a new post arguing you’ll soon “manage” swarms of AI coders like unreliable interns. The article compares large language models (LLMs) to compilers, but admits they’re error-prone. Then it drops the spicy line: there are “effectively infinitely many of them.” Commenters pounced. One shot back with a reality check: “infinite workers” is tech bro fantasy, not a hiring plan. Others accused the piece of bait-and-switch: “The contents don’t match the title” and “just another startup ad.” The mood? Skeptical side-eye with popcorn.
Process talk sparked more fire. The author’s plan → code → verify flow sounds tidy, but a tester crowd said doing checks at the end invites rubber-stamping. One warned it’s “super failure prone” without Test-Driven Development (TDD), which means writing tests first. A zinger: commenters say Copilot changes tests so it can pass. Meme-makers dubbed LLMs “chaos gremlins” and “AI interns you must babysit,” while a few pragmatists nodded that human judgment should concentrate on planning and verification. The kicker: the Tonkotsu plug made many roll their eyes—great pitch, sure, but the comments turned it into a lively debate about whether devs are becoming managers or just unpaid AI QA.
Key Points
- •LLMs and coding agents are likened to compilers but characterized as fast yet unreliable, unlike modern reliable compilers.
- •Developers’ roles are shifting toward managing many “unreliable compilers,” requiring clear delegation, guardrails, and verification.
- •Common pitfalls include under-delegation (micro-management) and over-delegation (hands-off), both seen in current coding agent use.
- •A structured workflow with explicit protocols and hand-offs is proposed, emphasizing a plan → code → verify model.
- •The article introduces Tonkotsu as a tool focused on planning and verification while orchestrating the coding middle to scale development.