marcotm

Nice idea!

For quickly skimming the list, I think it would be useful to add a short, standardized summary for each project. These summaries should be easy to generate automatically with any LLM of your choice.

I have a similar tool for organizing the “Who Is Hiring” threads [1] that uses GPT to provide a quick overview. It’s still running on an old model version from 2023, yet it has been working surprisingly well for over two years now with basically zero maintenance.

[1] https://www.hacker-jobs.com/

I built a site that transforms the monthly Who is hiring posts into a structured table [1]. It also allows for semantic sorting to find similar jobs to a given one. It originally used GPT-3, which already worked quite well.

[1] https://www.hacker-jobs.com/

I think that a huge part of Tailwind‘s success is not so much the underlying concept but a combination of several aspects: You get a kind of generic design system (margins, paddings, matching colors, etc.) that are restrictive enough to help make things look good and consistent but flexible enough to not result in a every-site-looks-the-same situation (at least not to the extent many other CSS frameworks did); Tailwind sits at a low level of abstraction above raw CSS that is just enough to make some CSS concepts a little bit easier to use (flex, grid, etc.); the documentation is quite good and getting started is relatively easy; …

Given these (and probably some more) aspects, Tailwind makes it easy to get some good results quickly which helps to convince users that their whole approach is the right way to go. Add very good marketing and early traction and you are where we are now.

So I agree with the article that Tailwind‘s conceptual approach is not the holy grail it‘s often said to be, but it maybe also is not even the reason for its success compared to many auxiliary things it did/does right.

Sample size one, but I actually found that the iOS ecosystem works surprisingly well for having a finished app. I have one, created over six years ago with almost no updates over the past years. I intentionally built it in a way that does not require any kind of maintenance: It works offline, i.e., does not require a backend; it's a one-time payment through the App Store (no subscription -> no expectation of continuous updates); the free and paid versions only differ in how many items you can put in the database, i.e., everyone can try the full feature set before buying -> reduces complaints / refunds; etc.

While these are deliberate design decisions, a nice benefit of the iOS ecosystem is that the app has kept working on every new iOS version without me updating anything in the code. Things like installation flow, payment processing, etc. are definitely components that would fall apart a lot sooner when not offered by the platform itself.

Sure, platform lock-in and all that stuff, but so far the app just keeps generating happy users every month (and some pocket money).

If you like the format, you might also like https://www.hacker-jobs.com/ Full disclosure: I run that site.

HN‘s „Who is hiring“ threads can be a good source for remote opportunities.

Self promotion: I run hacker-jobs.com [1] which indexes these threads and lets you filter by remote opportunities.

[1] https://www.hacker-jobs.com/

Just to present a balanced view for our international audience: a significant number of locals sincerely enjoy Oktoberfest. Within my circles in the Munich tech/startup scene, a visit to Oktoberfest with the team is often considered one of the year’s highlights in terms of team building and socializing. People attend with friends, family, etc. The infamous images of inebriated tourists merely represent the most attention-grabbing outliers, overshadowing the majority who simply have a splendid time there.

Munich does have its share of issues, rent being a notable example. However, most of these problems are likely a result of the general lack of serious troubles, relatively speaking, which in turn boosts the city's attractiveness.

Indirectly, in a way. A few months ago, I built a small site [1] that parses and structures the "Who is hiring" posts using GPT. It was/is just a personal project to deepen my actual experience in building things with LLMs [2]. After the launch [3], I was actually contacted a few times for LLM-related projects. I am not looking for a full-time job at the moment, but it led to a small consulting/contracting engagement.

PS: The site is still up. Some people think it's a nice alternative interface for quickly skimming the "Who is hiring" posts.

[1] https://www.hacker-jobs.com [2] https://marcotm.com/articles/information-extraction-with-lar... [3] https://news.ycombinator.com/item?id=35259897

Yes, sorry, you're right of course. I wanted to say that you need to use the more developer-oriented tooling (API, Playground) if you want to have the parameter options.

For many use cases like summarization or information extraction, you can get deterministic and mostly non-creative results by adjusting the parameters (temperature, top-p, etc.). This is only possible via the API, though. And it work's most reliably when providing the whole input which should be worked on ("open book" as another commenter called it). I run a task like this for Hacker Jobs [1] and am quite happy with the results so far (there is also an article detailing how it works [2]). If you ask for facts that you hope are somehow remembered by the model itself, it is a different story.

[1] https://www.hacker-jobs.com [2] https://marcotm.com/articles/information-extraction-with-lar...

Thanks for mentioning the Firebase-based API. I knew it existed, but somehow I went with the Algolia API by default. I use their HN search quite a bit, so that's probably why I stuck with them. (no affiliation)

You're welcome! For the chat model, it definitely helps to let it know that you want valid, parsable JSON (and nothing else). Otherwise it tends to get chatty. ;-) Depending on your use case, you might even ask it to fix the JSON if it's not parsable.

Thank you very much! I will definitely try out your suggestions! However, at least with GPT-3.5 and the amount of data I have to deal with in this case, my main concern is the quality of the extractions. With about 500 posts per month, the cost is manageable. But for larger datasets, saving tokens is definitely important.

Thank you very much! Filtering by location (and role) is on my todo list, but it is trickier than it seems at first. And I totally agree that the buttons are confusing. Actually, the "sort" button does sort the jobs. It sorts by semantic similarity to the job you selected (using the GPT text embedding). As for the buttons (and probably other parts of the site) not being accessible: I apologize. This shouldn't be an afterthought.

Exactly. I am sure you can get similar results with some "traditional" NLP skills, but the good (bad?) thing is that they are not required when using one of the newer LLMs.

The core ideas for extracting the information with GPT are already available in the blog post linked above. Those are exactly the prompts I'm using. The rest is just a pretty simple Nuxt web application. So I'm not sure if open sourcing my mediocre frontend code would be of any value. Is there anything in particular you would be interested in?

Thanks! Haha, you're right. There's already a kind of rule for that in the original prompt I'm using, but the "competitive" thing somehow still slips through. Will fix it in the next version.

Thanks for the suggestion! It's on my todo list. For now, you at least can sort jobs by similarity to a selected job. It's the middle icon to the left of each entry (maybe not the most intuitive way how to do it, though).

I wanted to share a little side project of mine that I created while tinkering around with GPT-3.

The project uses the Algolia HN Search API [1] to retrieve the "Who is hiring?" posts from HN and then parses them with the help of GPT-3 / GPT-3.5 (I do not have API access to GPT-4, yet, but it already works quite well even with the older models). It then puts the job postings into a structured list that is hopefully easier to skim than the original postings. There are some additional features like sorting jobs by semantic similarity (based on the text embeddings from OpenAI). Filtering, sorting and saving favorites is implemented client-side, so your data and preferences remain local to your browser.

Originally, this wasn't even meant to be a public product, but if people find it useful (and HN is fine with it), I'll try to keep it running. I've also written a short article about how the parsing works behind the scenes [2]. It's quite amazing how easy many of the classic NLP tasks have become with the newer LLMs.

Happy to answer any questions about the project!

[1] https://hn.algolia.com

[2] https://marcotm.com/articles/information-extraction-with-lar...

There's a Kickstarter for startups. It's Kickstarter. But looking at the current project guidelines, I think you are right that Kickstarter is not (or does not want to be) the right platform for pure software / SaaS / etc. startups (which I guess you are thinking about). So the equivalent would be something like backers pay x amount of money to get a lifetime, one-year, whatever license of the final product (in contrast to shares like in crowd investing). Funds will only be paid to the project initiator if they get above their pre-defined threshold (similar to Kickstarter). I definitely think that this could be valuable, but I'm also not sure if that space is not already covered by existing companies. Seems relatively straight forward. Somehow similar: Early access titles on Steam, etc.

+1. Due to the small size of our company, my role still is somewhere between lead engineer and a pure managing position, but it‘s enough of the management schedule that I can no longer be a reliable partner for working on the core product. So I try to automate/optimize internal workflows with my coding skills, i.e., I build tools that improve things if they exist but do not break or block things if they‘re delayed due to too many meetings creeping in.