Labor market impacts of AI: A new measure and early evidence

1. Performance-critical code to featurize byte slices for use in a ML model. Claude kept trying to take multiple passes over the slice when the featurization can obviously be done in one. After I finally got it to do the featurization in one pass it was double-counting some bytes but not others (double counting all of them would have been fine since the feature vector gets normalized). Overall it was just very frustrating because this should have been straight-forward and instead it was dogshit.

2. Performance-critical code that iterates over lines of text and possibly applies transformations, similar to sed. Claude kept trying to allocate new Strings inside of the hot-loop for lines that were not transformed. When I told it to use Cow instead so that the untransformed lines, which make up the majority of processed lines, would not need a new allocation, Claude completely fucked up the named lifetimes. Importantly, my CLAUDE.md already tells Claude to use copy-on-write types to reduce allocations whenever possible. The agent just ignored it, which is _the_ issue with LLMs: they're non-deterministic and any guidance you provide is ultimately just a suggestion.

This is what's so frustrating about the hype bros for me. In most cases, everything AI spits out are code smells.

We're all just supposed to toss out every engineering principle we've learned all so the owner class can hire less developers and suppress wages?

I'm sure it's working great for everyone working on SaaS CRUD or web apps, but it's still not anywhere close to solving problems outside that sphere. Native? It's very hit and miss. It has very little design sense (because, why would it? It's a language model) so it chokes on SwiftUI, it also can't stop using deprecated stuff.

And that's not even that specialized. It still hallucinates cmdlets if you try to do anything with PowerShell, and has near zero knowledge about the industry I work in, a historically not tech-forward industry where things are still shared in handcrafted PDF reports emailed out to subscribers.

I'm going to leave this field entirely if the answer just becomes "just make everything in React/React Native because it's what the AI does best."

(1) LLMs are basically Stack Overflow on steroids. No need to go look up examples or read the documentation in most cases, spit out a mostly working starting point.

(3) Learning. Ramping up on an unfamiliar project by asking Antigravity questions is really useful.

I do think it makes devs faster, in that it takes less time to do these two things. But you're running into the 80% of the job that does not involve writing code, especially at a larger company.

In theory, this should allow a company to do more with fewer devs, but in reality it just means that these two activities become easier, and the 80% is still the bottleneck.

When you're unemployed, it doesn't matter. When executives cargo cult, it doesn't matter.

The leading AI exposure indices (Anthropic, Eloundou et al.) focus on which jobs get automated. They treat low exposure as “safe.”

But the least exposed workers—cooks, roofers, dishwashers, construction laborers—are often in the worst jobs: low pay, high physical toll, short career spans, and little upward mobility. Safe from AI, but not from burnout or injury.

I built JQADI (Job Quality-Adjusted Displacement Index) to combine AI exposure with job quality. It surfaces three kinds of risk:

High AI exposure → classic displacement risk Low AI, low quality → “trapped” workers in grinding, unsustainable jobs Moderate AI, low quality → partial automation strips cognitive work and leaves physical drudgery (the “task residual” effect)

Findings: 83.5M workers are in low-AI, low-quality jobs. Customer service reps, data entry keyers, and medical records specialists sit at the intersection of high exposure and poor quality. Meanwhile, chief executives and lawyers are both low-exposure and high-quality.

The index uses ONET, BLS, and Anthropic exposure data. Code and methodology are open source. LINK https://github.com/quinndupont/JQADI