Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

What Claude Code chooses

amplifying.ai

Post reply on HN
Print view View original

191–200 of 266 posts

Re: What Claude Code chooses

#191

Post by nindalf »

jaunt7632 wrote: post #124

[dead]

How did you come to the conclusion that it was blogs that made it change behaviour? Look at the examples where Claude shifted behaviour dramatically between Sonnet 4.5 and Opus 4.6. Drizzle ORM went from 21% to 100%. Was there an avalanche of Drizzle related blog posts that we all missed? Celery went from 100% to 0%. Was there a massive but invisible hate campaign against Celery?

Blog posts almost certainly helped. But dramatic shifts like these to favour newer tech indicates that there's some other factor in play.

Re: What Claude Code chooses

#192

Post by jascha_eng »

conception wrote: post #115
killingtime74 wrote: post #113

Earlier quoted context omitted.

I use a skill that addresses these short comings, it basically forces it to plan multiple times until the plan is very detailed. It also asks more questions

Share?

Probably referring to superpowers or gsd. But imo these are asking way too much stuff and are just annoying. It's useful for realy vibe coders though that don't have any idea what they are doing. It will ask you: Should I handle rate limiting for the slack-api? Before you have written a single line of code.

Re: What Claude Code chooses

#194

Post by gjm11 »

hinkley wrote: post #61
dvt wrote: post #50

Earlier quoted context omitted.

> Claude Code likely is correct that I should start to use NeonDB and Fly.io which I have never used before and do not know much about I wouldn't be so sure about that. In my experience, agents consistently make awful architectural decisions. Both in code and beyond (even in contexts like: what should I cook for a dinner party?). They leak the most obvious "midwit senior engineer" decisions which I would strike down…

How do you make an LLM that’s was trained on average Internet code not end up as a midwit? Mediocrity in, mediocrity out.

If you take thousands of photographs of human faces and average them out (even if you do it just by roughly aligning them, overlaying, and averaging the pixels) then what you get is a (perhaps blurry but) notably more attractive than average human face image.

LLM output could be like that. (I am not claiming that it actually is; I haven't looked carefully enough at enough of it to tell.) Humans writing code do lots of bad things, but any specific error will usually not be made.

If (1) it's correct to think of LLMs as producing something like average-over-the-whole-internet code and (2) the mechanism above is operative -- and, again, I am not claiming that either of those is definitely true -- then LLM code could be much higher quality than average, but would seldom do anything that's exceptionally good in ways other than having few bugs.

Re: What Claude Code chooses

#195

Post by tomhow »

jascha_eng wrote: post #173
jascha_eng wrote: post #166

Earlier quoted context omitted.

@dang this accounts comments smell like LLM slop. They are mostly on topic and its more claude than chatgpt but it's slop nontheless. is telling didn't win... It won ... Look at their other comments they are also fishy I know you guys don't want us to call it out because of negativity. But there needs to be awareness in the community, this is the top comment somehow right now. It feels like it happens every other thr…

Note I might be wrong on this one but it's just extremely annoying that I even have to consider if I am being manipulated by an AI while reading HN comments. If I want to read AI stuff I go to Clawdbook or OpenAIs Sora app.

Sure, and we've banned the account, but please email us with these hn@ycombinator.com. @mentions don't work on HN; I only saw it because I was looking through the thread. We're also asking people not to make these accusations publicly, partly because they take longer for us to see than an email, and also because a false accusation is more harmful than a valid accusation is beneficial.

Re: What Claude Code chooses

#196

Post by esailija »

PeterStuer wrote: post #176
piokoch wrote: post #164

Earlier quoted context omitted.

"Tailwind didn't win because it's the best CSS solution. It won because it has the most tutorials per capita in the training set." Obviously. People keep forgetting that "Artificial Intelligence" does not think and is not intelligent. It just statistically predict next token in a sequence. It is all statistics. So, Django 6 has new task framework, but LLM does not care, as Celery has better stats. Side note: it is no…

"does not think and is not intelligent. It just statistically predict next token in a sequence. It is all statistics" Technically correct, but pretty useless as a working model. Like sayin humans are not intelligent. It's just biochemical and bioelectric reactions. It's all physics. How would you, from a Searlian perspective argue against "humans are just statistical next token predictors"?

We don't know what humans are because they are a black box, we use some imperfect models that have limited usability in specific contexts.

LLM is white box that we know for sure is just a statistical next token predictor and nothing more. It's not a just a model of some black box we are trying to understand but the whole actual thing. That people think it's something more or could be something more is on them. If you understand that then you understand the flaws, limitations and vulnerabilities which is very useful.

Re: What Claude Code chooses

#197

Post by jaunt7632 »

OJFord wrote: post #172
jaunt7632 wrote: post #160

Earlier quoted context omitted.

[dead]

That's an incomplete story though, 'revenue has plummeted due to LLMs', 'revenue is from people sponsoring the project', so... what, people that formerly liked and sponsored Tailwind stopped , figuring they can just ask AI now? Bit surprised that would have happened in significant volume (I'd have thought the LLM using non-sponsors would have far more overlap with the prior non-sponsors) but maybe.

[dead]

Re: What Claude Code chooses

#198

Post by properbrew »

jcims wrote: post #66

Interesting to me that Opus 4.6 was described as forward looking. I haven't *really* paid attention, but after using 4.5 heavily for a month, the first greenfield project I gave Opus 4.6 resulted in it doing a web search for latest and greatest in the domain as part of the planning phase. It was the first time I'd seen it, and it stuck out enough that I'm talking about it now. Probably confirmation bias, but I'm gene…

> to do great things in the context of the right orchestration and division of effort

I think this has always been the case. People regularly do not believe that I built and released an (albeit basic, check the release date - https://play.google.com/store/apps/details?id=com.blazingban...) android app using GPT3.5. What took me a week or two of wrangling and orchestrating the LLM and picking and choosing what to specifically work on can now be done in a single prompt to codex telling it to use subagents and worktrees.

Re: What Claude Code chooses

#199

Post by alex_suzuki »

aryehof wrote: post #143

I fear we are heading to less innovation. Are paradigms, techniques and practices that are not popular (or recent) likely to be increasingly forgotten?

Or the other way around… are more recent approaches significantly disadvantaged because of the huge inertia of existing solutions by virtue of them having existed in the training data both broadly and for a long time?

Re: What Claude Code chooses

#200

Post by AlecSchueler »

awad wrote: post #74
actionfromafar wrote: post #44

Earlier quoted context omitted.

That's called LLM SEO now I believe.

There are competing terms currently being decided on by the market at large: AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) Candidly I am working on a startup in this space myself, though we are taking a different angle than most incumbents. While it's still early days for the space, I sense a lot of the original entrants who focus on, essentially, 'generate more content ideally with our pa…

> I am working on a startup in this space myself

Do you see it as a positive contribution or just riding the gold rush?

Post reply on HN