May 20, 2026
100K lines, 1 giant comment war
Learnings from 100K lines of Rust with AI (2025)
He built a giant AI-made system fast — and the comments instantly turned into a brawl
TLDR: A developer says AI helped build a massive Rust system shockingly fast and made it much faster too. But commenters immediately argued that big code counts mean little, questioned whether the testing is enough, and roasted Rust as a painful match for AI.
A developer says he used AI tools to crank out more than 100,000 lines of Rust code for a serious cloud reliability project in just a few months, then boosted performance from 23,000 to 300,000 operations per second. On paper, that sounds like a huge flex. In the comments, though, the crowd basically replied: cool story, but does it actually prove anything? The biggest backlash was aimed at the bragging rights around sheer code size. One skeptic flatly said that with AI agents, the same job should probably take less than half that much code — turning the author’s victory lap into an instant "that’s not the flex you think it is" moment.
The second fight was over trust. The author says correctness comes from 1,300+ tests and extra safety checks built into the code, but commenters weren’t exactly throwing confetti. One person did the math and wondered if that was way too few tests for a massive, high-stakes system that’s supposed to keep important services running. Another reader asked the more beginner-friendly question: are these "runtime contracts" just a fancy name for validation checks, or something different? In other words, the audience wanted receipts, not vibes.
Then came the AI-language drama. One commenter said using AI with Rust is a nightmare because the bot keeps tripping over Rust’s famously fussy memory rules, while another declared that Go, Ruby, and even Haskell are simply easier for AI to write. The vibe of the thread? Equal parts awe, side-eye, and "please stop measuring genius in lines of code."
Key Points
- •The article reports that a Rust-based multi-Paxos consensus engine was built with AI assistance to match and modernize Azure’s Replicated State Library feature set.
- •The author says the project took about three months, with about 100K lines of Rust written in roughly four weeks and performance improved from 23K ops/sec to 300K ops/sec in about three weeks.
- •The modernization was motivated by three stated limitations in the older RSL design: no pipelining, no non-volatile memory support, and limited support for RDMA-aware hardware.
- •The implementation is described as covering multi-Paxos, leader election, log replication, snapshotting, and configuration changes.
- •The article says correctness work included more than 1,300 tests and the use of AI-generated code contracts that become runtime assertions during testing.