Hacker News Reader

Comment by Aurornis | Hacker News Reader

My strategy is to skim the workshop to find suspicious sections, then prioritize debunking at least one part that makes the creator look bad.

Then respond with that addressed, with a firm but polite explanation that the document has some problems that need to be addressed carefully before we invest time on it.

None of this works if it’s coming from your boss, but it’s very effective at making people think twice before sending you slop.

People only do the workslop thing when they believe the benefits of showing the work outweigh the reputational risks. If they get caught every time they do it and exposed on an email chain, they start doing less.

View on HN

Sadly not my experience. When I give this kind of feedback on a document, highlight a gap, or give constructive feedback, most of the time this feedback is fed back to the AI and a new revision of the document shows up.

BUT because it is AI it's not that section that's addressed, it's about 90% of the document that now has changed, and I need to spend another gargantuan effort to review it. It's like by trying to be helpful I actually lose more time.

The AI has made it so that the thinking before writing is mostly gone, and shifted that step to the reviewers.

Left this way, every collaboration can turn into what used to be the worst nightmare type with that one unavoidable stakeholder or coauthor who likes to procrastinate and then create an insane fire drill in the final moments before a deadline.

Now, no matter the amount of preparation work, those last rounds can completely rewrite something with no hope of review. No iterative improvement. No ratcheting towards a known quality. Just a bunch of cargo cult review followed by YOLO-style, vibe-everything absurdity.

Review burden is increasingly higher for many people, with Tech likely being patient 0 for the rest of the economy.

The ratio of verification capacity to generation capacity, V/G, has broken with LLMs. It’s not simply an issue of more generation or less review.

The impression seems to be that individuals are more productive, but that productivity is someone else’s review burden. So the team/firm as a whole is not better off.

The cheap generation of content does mean that reviewer capacity is now a limited resource.

Unless your firm is aware and is measuring time spent on reviewing slop, there is no incentive or structure to ensure that time is respected and valued.

This is a management and awareness problem since the typical response is “use a bot to review it.”

I believe this is most clearly visible in mathematics. Who is going to review the 700 proofs put out by OpenAI in a single day ? And even if mathematicians could, how will they handle the exponential growth of LLM-generated proofs ?

Not to say these proofs are slop, even if the LLM-generated work is of good quality, what happens when no one knows how it works anymore ? Even if you ask the LLM to explain, which it does quite badly, the time to understand the explanation is incompressible.

So in the end, productivity will probably reach a ceiling that we can estimate as the product of humans, their cognitive capacity and their time. And that ceiling may be lower than what AI companies valuation expect, regardless of the compute and they can pump out and the RSI level they can reach.

Sounds like Claude. I haven’t made this experience with the large OpenAI models.

Reply on news.ycombinator.com