hysan

> One thing I've noticed is that people will engage in thread arson to kill submissions they don't agree with. When I see some of these low value comments, I check their comment history and often they have reasonable comments on other things.

Interesting to see that others are doing this. I started this habit because I got curious to see what bad/downvoted/dead commenters would say elsewhere and noticed this same pattern. It’s not on all of them (some accounts seem to post comments that get killed in every thread they reply on which makes me think those are troll accounts), but it happens enough that it was noticeable. Not sure how I’d describe this category of user. They aren’t trolls but they are clearly doing this purposefully.

I don’t know much about apple variety names, but if you love apples and plan to visit Aomori prefecture in Japan during apple season, I highly recommend stopping by a market or grocery. I used to live in northern Akita and there was a farmer from Aomori that would come by with freshly picked apples to sell at one of the supermarkets. They had many varieties and they all tasted amazingly sweet and unique. I have yet to find an apple anywhere close to those since moving back to the states. The most recent new variety that I tried that I liked was Kiku. It’s like a sweeter version of a Fuji apple with a very similar texture. Based on your preferences, I think this would a fun one to try.

Fair. I should have reigned in my emotions and written that first comment with a lot less conviction and not jumped to my conclusion. It’s a topic I’ve been spending a lot of time learning about and got a bit too excited when a relevant thread popped up where I could chime in.

True which is why I said anecdata. But the point is that it’s a useful signal when it’s across enough sessions where it quotes back contradictions to you. If you’re willing to burn the tokens, the first thing it does is use a tool call to search your rules to see what it quoted wasn’t there. That leads to the LLM claiming two common conclusions (sometimes after a lot of back and forth) - it’s in the harness or it’s in the model. Over the 30+ sessions where I’ve tried this, the overwhelming claim was some variant of a harness instruction. It might not be there but given the consistency and how all my rule tests have failed in the same way that others describe, it at least makes it reasonable to conclude that it’s baked in somewhere and in a way where agent rules aren’t able to strongly affect the behavior. My bet would be on the harness because the class of undesired behavior follows that of a structured response.

Relevant anecdata because I've burned many a Claude sessions on this.

If you're using Claude Code, then it's in the harness. At the close of many sessions, I would start a meta conversation over why the LLM would consistently break certain rules. What it found when debugging itself is that some of the "contradicting" rules that I had were in fact, not from my rules. Instead, the instructions from its own harness had phrases telling it to do things like that. When something contradicts, its own instructions would outweigh any custom ones you write. Every rule variant I had tested (including the one that says it overrides the harness instructions - and yes, I've actually tested all the ideas in your comment too) has ultimately been unsuccessful due to this according to the LLM.

I feel like the increased reviewing time is consistently understated. I’m just an IC, but it seems obvious to me that you cannot cut staff and achieve increased output. There literally aren’t enough eyeballs to go around reviewing code when everyone is 2-3xing their output. I spend so much more time reviewing code; reviews that are sorely needed because I regularly catch batshit insane “fixes” that work but would quickly turn the codebase into a mess (the most recent one being a multi-hundred line diff that I went and fixed in 2 lines in 15 min). Maybe I’m underthinking it but it seems obvious that you either maintain the same output with fewer staff or you gain increased output with the same staff. All the companies that are attempting to cut staff and gain increased output are chasing an impossibility and throwing away their opportunity to accelerate.

I’ve had a longstanding belief that big tech purposely over-employed because it denied resources to competitors. This gave them a velocity edge despite being big and bureaucratic. In the new AI world, denying resources doesn’t slow competitors down anymore because AI raises the productivity floor. Now that over-employment brings no competitive advantage, we’re seeing huge cuts.

Probably nothing and it’ll also do nothing for the many people whose lives were drastically impacted by those cuts.

Turnstile feels bad as a user. Every site that I’ve seen it long will lock up Safari hard while it’s doing whatever it’s doing. But at least I haven’t run into more than 2 refresh loops.

Same. Google Maps quality has gotten noticeably worse these past 2 years for me. It routinely tries to navigate me to making impossible turns or taking weird and sometimes more dangerous routes just to shave off a potential minute. I started using Apple Maps at the advice of a colleague and it’s given better directions. This is all local. I have no baseline comparison for using maps while on trips.

Took long enough. Maybe I missed it, but I didn’t see them say how invested they are in tackling this. Promoting a rule is one thing, but everything SEO related becomes a cat and mouse game. I don’t have high confidence that this will work.

Oh, so I’m not imagining this. Recently, I’ve tried to up my LLM usage to try and learn to use the tooling better. However, I’ve seen this happen with enough frequency that I’m just utterly frustrated with LLMs. Guess I should use Claude less and others more.

Does it? It only takes like 2 min for my electric kettle to boil. If I was a more avid coffee/tea drinker, I’d get one of the always heated hot water dispensers that are common in Japanese households (def one of the appliances I miss since moving back to the US). Then you never have to wait.

Dunno how long this is normally supposed to go but it took me 10+ min of actually seriously considering the fonts at each choice and the final suggestion is a font that I actively dislike. I’m curious how it’s narrowing things down because I noticed that it started to give me only serif fonts which I don’t like. But the sans serif ones that it was using to try and narrow things down had distinct characteristics that I didn’t like like very narrow stems or very narrow or wide characters. But it wasn’t doing that with serif fonts. I’m guessing it began to think I preferred serifs because of that but in reality, I wasn’t picking the lesser of two evils most of the time.