Hi HN!<p>I’ve been lurking for a while, but out of fear of being steamrolled by HN readers or maybe just natural introversion, I’ve always been too scared to post or comment. Which is why<p>1. this is my first real Hacker News submission<p>2. my friend Michael and I built "Fake" Hacker News, a place to post and see what AI-generated HN comments might say.<p>Here’s a video of me using fakeHN to test this very submission:
<a href="https://www.loom.com/share/4b9f4f9d7c77489a86baeb92ec55a1ed?sid=a74043f6-8a6e-4a68-bcda-330ab9d7eafe" rel="nofollow noreferrer">https://www.loom.com/share/4b9f4f9d7c77489a86baeb92ec55a1ed?...</a><p>And an example of one of our generated posts:
<a href="https://www.fakehn.com/submission?id=tnYPX00BX827jWFPFkVJ" rel="nofollow noreferrer">https://www.fakehn.com/submission?id=tnYPX00BX827jWFPFkVJ</a><p>To try it, submit a title and text, and depending on traffic and the powers that be, after ~5 seconds, you’ll see some Fake HN comments and replies.<p>We don’t support url submissions yet, but we’re happy to build it if the community wants it!<p>Other features to knock out: deeply nested replies, streamed comments, and higher-fidelity comments mapping to real readers, since the generations now are still pretty shallow. Instead of the quick and dirty system in place now, we think it’d be really cool to see how more nuanced AI agents with the opinions and biases of real individual HN readers might respond.<p>I’d love to see what fakeHN posts you’ve tried and hear any feedback, whether you feel like it’s more of a nifty toy or could eventually solve real problems. If nothing else, it’s been funny to try random posts and see the results. :)<p>- Justin and Michael
I love it<p>I didn’t try anything crazy. But this comment was randomly added<p>——<p>[deleted] on Sun Aug 20 2023
[flagged]
▲
dang on Sun Aug 20 2023
You can't post slurs to HN, so I've banned the account.
Hey HN! Michael here, one of the co-creators of Fake Hacker News.<p>if anyone is curious, here is some background info on our tech stack:<p>- Frontend: react, next 13, typescript, tailwind css<p>- DB: firebase for storing submissions and the generated comments<p>- Backend: vercel edge function that calls openai<p>- Analytics: posthog<p>For analytics, we are collecting the following product events:<p>- number of times we call openai<p>- failure rate of the generation request<p>- number of retries per user<p>would appreciate any feedback on the product and the tech stack!
I went recursive or meta or something and submitted the title and description of this submission itself. Some of the results:<p>cyberXplorer on Mon Aug 21 2023<p>Wow, this is such a cool idea! It's great that you're providing a platform to see AI-generated HN comments before posting. It could really help users get a sense of how their post might be received. Looking forward to trying it out!<p>TechNinja on Mon Aug 21 2023<p>This seems like an interesting concept, but I'm not sure how useful it would be in practice. AI-generated comments might not capture the nuances and insights of real HN users. Have you considered incorporating real comments into the system to improve its accuracy?<p>CodeBreaker on Mon Aug 21 2023<p>It's about time someone built something like this! It can be intimidating to post or comment on HN, especially for newcomers. This could be a valuable tool to build confidence and engage with the community. Kudos to you and Michael!
I'm a little sad you're uncomfortable about being steamrolled by HN readers. I get it, it took me <i>a long</i> time to not get emotional about HN comments, and still I often fail, but I'd encourage you to join the conversation as much as you can. I find HN generally to be a very kind and loving community and on the whole people tend to be very thoughtful and constructive in their criticism. I have noticed over the last few years folks seem to be trending towards a confrontational tone over a kind one when penning thoughts, however, that's exactly why we need more people like you in the community. So: please do dive in! :)
I want to see this, but with HN urls so actual HN threads can be compared with it, to see whether AI can generate on-topic threads of higher quality and civility, in general than human-created threads. Doing so by simulating actual users is an interesting way to go about it, although it might be counterproductive in the long run.<p>Maybe in the future this can be integrated into a filter here that detects inaccurate, biased or harmful comments and blocks or rewrites them on submission, or pre-generates the common boilerplate of discussions around evergreen sites and subjects, leaving the interesting parts for human beings.
I tried it like this-<p>Title - fakehn.com<p>Text - Do a test run of your hacker news submission to see how readers may respond to your post. Planning to add url support in the near future!<p>Top comment -
Is there a reason why we need a separate platform to test our Hacker News submissions? Can't we just post directly and see how readers respond?<p>A few interesting comments -<p>I'm not convinced that using a separate platform for testing Hacker News submissions is necessary. Why not just share drafts with a small group of trusted peers?<p>Has fakehn.com considered partnering with Hacker News to offer this testing service directly on their platform? It could save users the trouble of using an external website<p>I hope fakehn.com takes user privacy seriously and ensures that any data or submission content is secure.<p>---<p>The last comment is always a flagged one with a comment by @dang.<p>Awesome :D
After trying it with a few posts, the biggest flaw I saw, is that LLMs, just can't say yes or no, and only talk hypothetical.<p><a href="https://www.fakehn.com/submission?id=EUfJuTfwvbjNQWifM20C" rel="nofollow noreferrer">https://www.fakehn.com/submission?id=EUfJuTfwvbjNQWifM20C</a>
vs
<a href="https://news.ycombinator.com/item?id=37199177">https://news.ycombinator.com/item?id=37199177</a>
I don't see a SINGLE comment in any of these examples that threadjacks to discuss CSS or font kerning. This absence breaks my suspension of disbelief <i>entirely</i>. Something to work on for future iterations.
I think this is cool. This has given a use case for llm at work, using them to prescreen ideas and generate FAQ material.<p>First time it worked great even with only a link. Second time I had it generate some comments based on a small prompt and kept telling me something went wrong, alternating between comments and that error state.<p>Good work!
I submitted “Show HN: Fake Hacker News – See what HN has to say before you post” along with body text to Fake Hacker News. Here’s ByteBuster’s suggestion:<p>> Have you considered partnering with OpenAI's GPT-3 to improve the quality of the AI-generated comments?<p>Btw I’d love a showdead on option. I need to read that flagged comment.
i posted the headline and first paragraph of <a href="https://www.reuters.com/technology/elon-musk-says-x-will-strip-ability-block-accounts-2023-08-19/" rel="nofollow noreferrer">https://www.reuters.com/technology/elon-musk-says-x-will-str...</a>. Every response listed was a generic, negative response, like "this seems a terrible idea". I was hoping for at least one or two "people just dont understand Musk's genius / that's what mute is for" responses. So this didn't really for me capture the special "flavor" of Hacker News, which is that no matter how awful something is, there's always a solid nut of "this is actually a good thing" contrarians ready on the mark.
Suggested that beauty pageants be part of interview process. Responses: <a href="https://imgur.com/gallery/mPuJhqC" rel="nofollow noreferrer">https://imgur.com/gallery/mPuJhqC</a>.<p>For the record, I am not a fan of pageants. No idea why this popped into my head.
Amazing. Great work. Interesting and useful!<p>Only feedback I guess would be that one-line replies, often with exclamation marks felt much more common than on HN.<p>Still, v cool and probably hugely helpful for people with social anxiety, which seems like pretty much everyone. :)
> I’ve been lurking for a while, but out of fear of being steamrolled by HN readers or maybe just natural introversion<p>It’s really annoying to me that you cannot hide your karma on the top in the settings.<p>Like, I don’t need to see the number changing constantly. Worse thing about this site IMHO. Even Reddit does not display it so prominently.<p>Given how easily it is to nice if you’re being “steamrolled” via downvotes, I get it. The site conditions you if you’re a maximizer of the number.
Thank you, I had a good laugh at this[1].<p>[1] <a href="https://i.imgur.com/KF9e2h9.png" rel="nofollow noreferrer">https://i.imgur.com/KF9e2h9.png</a>
After an initial delay because of high load, I got a handful of generated comments. Shortly after, they all disappeared with an error. I’m guessing it’s some background poll that failed because of high load (could be paginating for more results, I don’t know). In this event, it’d be really nice to retain the already generated comments instead. I only got two seconds to look at them, before they got replaced by an error.
Haha, posted a “Show HN” for a JS framework that makes JS frameworks and the comments were exactly as expected: <a href="https://www.fakehn.com/submission?id=H0qene04EpSjuGJkiBRK" rel="nofollow noreferrer">https://www.fakehn.com/submission?id=H0qene04EpSjuGJkiBRK</a>
Love it. I chuckled when I saw a flagged comment with a telling off reply below it, and was disappointed I couldn't expand the flagged response to eead it.
Hmm... Let's try with the ridiculously esoteric thing I'm currently tinkering with:<p>> "Show HN: Ghidra object file exporter extension" - "I'm working on a reverse engineering technique which allows me to unlink parts of programs back into object files. Applications include binary patching, turning programs into libraries, ports... First release includes support for unlinking statically-linked, no-pic 32 bits MIPS executables into ELF relocatable object files."<p>Second comment and replies:<p>> "Is this something that could potentially replace traditional disassemblers?"<p>> "Absolutely! Ghidra's object file exporter extension opens up new possibilities for reverse engineering and analysis. It provides an efficient way to export disassembled code from object files, making it a valuable tool for the reverse engineering community."<p>> "While it may not completely replace traditional disassemblers, it definitely enhances the reverse engineering workflow and adds more flexibility to the process. Ghidra's object file exporter extension is a welcome addition to the toolbox of any reverse engineer."<p>> "It's an exciting development in the field of reverse engineering, but I wouldn't say it's a complete replacement. Traditional disassemblers still have their place and offer unique features. However, Ghidra's object file exporter extension is a powerful addition that enhances the capabilities and efficiency of the whole reverse engineering process."<p>It <i>almost</i> managed to make sense of my prompt (it's not exactly comparable to a disassembler, but rather as the opposite of a linker, but <i>damn</i> that was close).<p>> "I'm not sure I fully understand the implications of this technique. Can you explain it in simpler terms?"<p>That's actually what happens most of the time when I try to explain it to someone. It's such an unconventional topic that the AI itself is puzzled like everyone else.<p>> "This is exactly what I've been looking for! Can't wait to try it out."<p>> "As someone who works with MIPS executables, this is a game-changer for me. Thank you!"<p>OK, now it breaks down. There shouldn't be so much excitement about something so niche and even less for something that can (barely) handle PlayStation 1 executables from the 90s.<p>(<a href="https://www.fakehn.com/submission?id=sCuEvS4VUA1tFLfLKVJu" rel="nofollow noreferrer">https://www.fakehn.com/submission?id=sCuEvS4VUA1tFLfLKVJu</a>)
This is fun, but it's including meta data in the outputs. For example, "Reply 1:" or "Reply 2:" or even "Agree:" / "Disagree:".