disfictional

Solving the hardest problems doesn't always result in the best outcomes unless the answers are shared equitably and implemented compassionately.

I agree that intelligence at cost is exciting right now, especially if you view AI as a tool. The clock is ticking on subsidized tokens and cheap or free local inference will be the future. I think people want AI to be an oracle for prediction and discovery, which is where the sota models come in. But each release seems more iterative and underwhelming than the last. When the latest models regularly reveal unexpected insights, like how to get my execs to stop demanding hand-wavey 10x productivity gains, somebody let me know.

Currently, I'm using custom instructions plus reinjecting the writing cues Opus 5 ignores most frequently via a UserPromptSubmit hook. Again and again, I'm reminding the model what voice I want. Again and again, Opus 5 ignores it.

> Even a correct guess leaves you worse off than not committing.

Similarly, this is also confusing. If I scaffold a highly-likely feature and everything lines up, I ship the feature faster. My team isn't guaranteed to grow or even maintain our headcount, so scrambling to account for YAGNI close to the deadline feels worse than congratulating ourselves on our restraint.

I misread the title as "write bitter error messages", and that resonated with me deeply.

The key to the essay is that "ordinary people will be fine." Software Engineers will be highly impacted, though not in the way most commenters seem to think. Management isn't going to arbitrarily decide that, "AI can do 65% of the job, so we'll lay off 65% of the engineers." They won't hire. Attrition? New projects? "Just use AI tools to be more productive. Find the bottlenecks and automate them. Focus on your core value." AI isn't going to be a fast slash to the workforce; it will be a constantly accelerating drain. Yes, ordinary people will be fine, but those of us who aspired to be artisans of building these systems will be stretched further and further until all we do is maintain AI code full-time for a discounted price.

As someone who spent a year writing an SDK specifically for AI PCs, it always felt like a solution in search of a problem. Like watching dancers in bunny suits sell CPUs, if the consumer doesn't know the pain point you're fixing, they won't buy your product.

LLMs as a body double for executive dysfunction is a great insight. I see chronic examples of corporate-sponsored executive dysfunction: striped calendars, constant pings and interruptions, emergency busywork, fire drills. It's likely that LLMs aren't creating productivity as much as they're removing starting inhibition and helping to maintain the thread through context switching. What's presented as a magical tool, which LLMs can be in the areas you mentioned, is also presented as a panacea for situations that simply don't promote good programming hygiene.

Laziness, or job search, or parenting, or health issues, or caregiving, or something else. It's not a binary stay-current-or-you're-lazy situation, it's that the entire industry is moving to shorter timelines, smaller teams, and more technical complexity for web projects simultaneously. LLMs are a huge dopamine hit for short term gains when you're spinning plates day after day. The question is what the ecosystem will look like when everybody's been using LLMs as a stopgap for an extended period of time.