Perry Rajnovic

41% of all code shipped in 2025 was AI-generated or AI-assisted. The defect rate on that code is 1.7x higher than human-written code. And a randomized controlled trial found that experienced developers using AI tools were actually 19% slower than developers working without them. Devs have always written slop. The entire software industry is built on infrastructure designed to catch slop before it ships. Code review, linting, type checking, CI/CD pipelines, staging environments. All of it assumes one thing: the person who wrote the code can walk you through what it does when the reviewer asks. That assumption held for 50 years. It broke in about 18 months. When 41% of your codebase was generated by a machine and approved by a human who skimmed it because the tests passed, the review process becomes theater. The reviewer is checking code neither of them wrote. The linter catches syntax, not intent. The tests verify behavior, not understanding. The old slop had an owner. Someone could explain why temp_fix_v3_FINAL existed, what edge case it handled, and what would break if you removed it. The new slop has an approver. Different relationship entirely. Arvid’s right that devs wrote bad code before AI. The part he’s missing: the entire quality infrastructure of software engineering was designed around a world where the author and the debugger were the same person. That world ended last year and nothing has replaced it yet.

🚨BREAKING: Alibaba tested AI coding agents on 100 real codebases, spanning 233 days each. the agents failed spectacularly. turns out passing tests once is easy. maintaining code for 8 months without breaking everything is where AI collapses. SWE-CI is the first benchmark that measures long-term code maintenance instead of one-shot bug fixes. each task tracks 71 consecutive commits of real evolution. 75% of AI models break previously working code during maintenance. only Claude Opus 4 stays above 50% zero-regression rate. every other model accumulates technical debt that compounds over iterations. here's the brutal part: - HumanEval and SWE-bench measure "does it work right now" - SWE-CI measures "does it still work after 6 months of changes" agents optimized for snapshot testing write brittle code that passes tests today but becomes unmaintainable tomorrow. Alibaba built EvoScore to weight later iterations heavier than early ones. agents that sacrifice code quality for quick wins get punished when consequences compound. the AI coding narrative just got more honest: most models can write code. almost none can maintain it.

@liz my MbP is an Intel '19, i've definitely noticed lots more lag in unlock and icon re-caching with macOS 26. but i do appreciate the extra level of greyscale i can have in the OS chrome. but i'll probably pull the trigger on a new one this year anyways.

ok so I know lots of people have complained about it, and my personal MBP is from ~2021 and I notice absolutely no problems. but I mean I spent like many years modding my macs with OS 8.5/9/X to do this this, its why I have this account in the first place, this is amazing

The ones that get it, get it 🤝

At least once a week i think of that tweet about how millennials are the one generation born after they took lead out of paint and before smartphones.

Genders of France’s rivers in French

The Pittsburgh Post-Gazette is shutting down.

Real nerds write _everything_ in markdown by default, regardless of whether it will be rendered, knowing that the target audience can be trusted to mentally render it while reading.

@Citymapper your listing of mta issues is excluding planned skipping of stops, which is sort of important (northbound R in BK)

@SwarmApp latest update definitely addressed this, thanks!

@psylum we've identified a performance bug in adding a caption to a checkin that we're investigating right now. are you finding any other features to be running slowly?

@SwarmApp latest update is running extra slow on the iphone 13 mini. still fully supported, hop this isn’t a neglect of older devices coming to pass.

@SwarmApp also, not perf related, but i went to a sporting event recently and the “you’ve seen Team A x times and Team B y times” had the x/y values swapped.

@SwarmApp there does also seem to be a lag before i see that i have notifications badged after starting the app.

@AndrewSolender 4 or 1 are my choices. if you're taking 5 with nobody in 4, you're a jerk blocking a seat, and if you're taking it with someone already there, most people lean away and give you space anyhow.

picking 4 is diabolical. 5 or 1 are the only correct answers

“Take Mass Transit, Get Laid” is perhaps an angle Amtrak and others should explore again.

I just built a base system (rootfs) with Coreutils, Binutils, Bash, and curl, those are all compiled using Fil-C (a memory-safe C, created by @filpizlo, fil-c.org), everything works properly, it's really awesome! The building script: github.com/hemashushu/fil…

@BillLaboon i was going to recommend the perrin funicular in prague, but it’s apparently closed. the astronomical clock tower, and any local brewery are wins

What are some things I should see/do while in Sofia and Prague? I won't have much time in either one, so something quick.

I cannot stress enough there are four different ads in this screenshot

All the roads in Canada. Reveals much to think about. (map by @EarthArtAus)

The largest clock in the USA, located on #Pittsburgh's South Side, received a much needed makeover this week as @IronCityBeer is now displayed on it. I shot a bunch of photos of it this morning but this one with it framed by these two old bell towers was definitely my favorite!