Ask HN: How can I improve how I Show HN?
The main project I have shown here is highly technical, and one using LLMs that I got into far before the current hype, in 2020 onwards.
Despite that, I have some successful Product Hunt launches over the years.
The product I have tried to Show HN here (especially whenever new features others don't have are released, like conceptual knowledge gap analysis of test results), has public interest, reaching 10k+ users by organic sharing this month and increasing twitter sharing like here: https://twitter.com/Artifexx/status/1658381998740627460
But when I share it here I get crickets. This feels unusual to me, because I usually am releasing something not available elsewhere and love going into the technical details a bit, something that I thought would be popular on HN.
Any advice on my approach, or why this might occur? Is it because on HN, people are after dev tools, not studying tools? What do you want to see in a Show HN?
8 comments
[ 3.9 ms ] story [ 28.1 ms ] threadPretty much. It's mostly the wrong audience.
HN is interested in things that are either aimed at a technical audience, or where you can show some really unique way of building something so that even if the end product isn't relevant to a technical audience, that the process used to build it was.
There'll always be exceptions, but that's the general rule I've observed. And most of the exceptions are things where there's some other aspect of the product that makes it interesting to HN or some aspect of how it was built that makes it interesting to HN, or where the target audience happens to overlap with HN type users to a high degree.
I looked at the show HN in your history and I would've skipped over it and not read it. If it was a technical blog post about how you solved something that was both interesting to me and seemed novel, then I might've clicked. (But if you had that information buried inside something that seemed like "just" a product for learning, I wouldn't have clicked)
Try product hunt, twitter, tiktok, and other places.
> If you want to appeal to Hacker News readers, tell them how something complicated works. Especially if most people either don't know, or are mistaken about, how it works.
Source: https://twitter.com/paulg/status/1553708274465140736
A repost once in a while, appears to be acceptable though.
Here are some statistics: http://readcodelearn.com/notes/when-to-post-to-hn.html
I tried a demo quiz on studywand, but it seemed worse than typical manually generated quizzes. Here's an example:
You could know nothing about psychology and choose B because it's the only one that's an answer to a question of that form. A good quiz question tests for a more precise understanding.If the tech is doing something that couldn't be done before, you should draw our attention to it!
The quiz question (though not answer, since overwritten) from the demo you posted was generated in 2020, with T5 not GPT-3. Other and more complex questions were also possible with our finetuned tech on a bespoke carefully collected dataset. It was at the time much more impressive than it is now with GPT models, and we had it out before competitors even launched(the closest thing was old style GOFAI and a few online tutorials on finetuning T5 for similar tasks that had inferior data).
For something more offbeat the technical part isn't going to sell it (unless you write a blog from the language viewpoint: "How I used Rust to beat an LLM into submission" or something).
You might want to get someone else to read up and play with it and write the post for you, as coming in from the out side.
And even then, doing everything right, you likely will languish in the new queue unless the lightning strikes and everything lines up.