361–370 of 401 posts
Earlier quoted context omitted.
This seems like exactly the type of problem human-written filtering systems fall into as well.
human-written filtering systems don't brag about having a solution for a problem in 2 minutes and fail.
Earlier quoted context omitted.
I should add: Many years ago I used to read a news site that was modeled after slashdot. One day the person running it decided to switch it to be community-moderated. Every day it was the same discussion over again, from someone who didn't bother to do a Google search or look at what was posted the day prior. After a week or so of seeing the same discussion over and over again, I stopped reading the news site. Needle…
Out of curiosity, what was the site? The only clone I'm familiar with is (defunct) christdot.org.
I'll let you guess what the site was!
The problem was similar to a Goldfish group I left on Facebook: Every day someone, in a panic, would ask for help with a sick goldfish with the same exact symptoms as a post made the day prior.
Needless to say, if a community can't occasionally discuss what content it wants, and what content to exclude, it devolves. HN has a "no politics" rule. I think it's okay to discuss if other content needs to be excluded too.
I built you this: https://tools.simonwillison.net/hacker-news-filtered It shows you the Hacker News page with ai and llm stories filtered out. You can change the exclusion terms and save your changes in localStorage. o3 knocked it out for me in a couple of minutes: https://chatgpt.com/share/68766f42-1ec8-8006-8187-406ef452e0... Initial prompt was: Build a web tool that displays the Hacker News homepage (fetched from…
Almost certain you can use the HN Algolia to do the same thing by excluding terms
I switched to using the older firebase API which can return up to 500 item IDs from the current (paginated) homepage - then I fetch and filter details of the first 200.
https://github.com/simonw/tools/commit/ccde4586a1d95ce9f5615...
Skill issue?
No for three reasons. One, lets be honest, hn wont do it, part of their secret sauce is that they don't change, and they know that. Two, fragmenting the community would just reduce engagement and risk making both feel like a ghost town. Three, LLMs are (one of) the forefronts of our industry. State of the art is advancing fast. It has properties that no one knows the best practises for. And it has implications that a…
I would rather hang out on a HN more similar in attitude and user count as when I joined than have it be "lively" with the user base it has now.
I would agree with #3 if it were the case that the technical side of LLMs were discussed more, and there was a focus on trying to run and understand them yourself; these are topics that I think make more sense for hackers to care about. The front page stories about LLMs are usually business news about LLMs, blog post about using (someone else's) LLM, the occasional blog post that says LLMs suck either for technical, social, or other purposes.
Obviously I'm missing a few categories here, but my point is that there is basically zero hacker spirit in the LLM posts. Since when is it part of the hacker ethos to plug into a SV machine that you pay $200/month for the privilege to use to do stuff? That's ridiculous...
It's hard to not feel like this is part of a bigger transformation of the culture and then absorbing that culture on HN. Hackers make things, and they are unlikely to settle for poorly made things, but routinely the chorus on HN is that you should never reinvent the wheel, etc., even though you could make a better wheel for your purposes than the wheel you've found online. Not to mention this hilarious "I hate writing code" sentiment that seems to have grown over time. In my opinion neither of these sentiments belong in any group claiming to be hackers.
But the user base of HN is not hackers, hence why they feel comfortable spouting these opinions, plugging into $200/month services that write code for them (badly and slowly, I've used Anthropic's Max plan with Claude Code), not understanding and/or taking apart the technology they use, etc.
Earlier quoted context omitted.
It’s not clear where you are trying to go with this aggressive line of questioning that puts words into other peoples mouths that they aren’t saying.
It only seems aggressive if you believe that HN somehow owes you the precise content you are seeking over whatever it’s delivering now. If you don’t like the topic, don’t read it. No one’s forcing you to do so. Don’t yuck someone else’s yum. The internet’s a big place and life is short, spend your time reading stuff you like rather than bitching that you want things another way.
ublock origin filter example, removes post, post actions/stats, and spacer: news.ycombinator.com##tr.submission:has( :has-text(/LLM|agentic/)) + tr + tr news.ycombinator.com##tr.submission:has( :has-text(/LLM|agentic/)) + tr news.ycombinator.com##tr.submission:has(*:has-text(/LLM|agentic/))
We just built https://www.hackernews.coffee/ to rerank your frontpage based on a quick survey of your preference, all local storage based. In general we're thinking about how you can have a transparent profile that stands in place of an opaque algo, or in this case a dominance of a community by something you're not so into. It allows you to still engage with HN, but through the lense of a profile you have control ove…
I feel like 10 answers is too few to generate a good profile. I tried it and it marked a lot of stories as 'skip' that were pretty interesting to me, presumably because they didn't fit into the 3 or 4 categories it determined I was interested in. That said, it was useful to filter out AI stuff, so that's good I guess.
Theres some work to make _both_ the onboarding for totally fresh users essentially frictionless, while also allowing adding more nuance. Our todo is having a "rank some more" thats easy, or some "while browsing" to build that profile up over time.
Earlier quoted context omitted.
Its pretty easy to do the user-loyal bit, with a bit of prompting to give an llm your preferences/profile. Not ideologically loyal, but i mean acting in accordance with your interests. The tricky part is having that act across all sites in a light and seamless way. Ive been working on a HN reskin, and it only is fast/transparent/cheap enough because HN has an api (no scraping needed), and the titles are descriptive e…
I'd expect a noticeable delay with current local LLMs - especially visiting a site for the 1st time. But then they could potentially memoize their heuristics for certain designs, including recognzing when some "deeper thought" newly required by server-side redesigns. But of course local GPU processing power, & optimizations for LLM-like tools, all adancing rapidly. And these local agents could potentially even outsou…
We ended up finding that a middle ground between that and ~simonw's no-AI but fast, was to use flash for fast semantic understanding of preferences and recs, but degraded quality compared with a friontier model.
> And forward-thinking sites might try to make that process easier, with special APIs/docs/recipe-interchanges for all users' agents to share their progress on popular needs.
HN is that! our exploration was made 1000% easier because they have an API which is "good enough" for most information.