Salutations, intrepid Internet navigators! You've discovered the command center of innovation for our corner of the web. This is the place where your brilliant ideas become reality - and, honestly, the unique and sometimes wild creativity I've seen from you all is the reason why I do this.
Here at Feedback on The Internet, we're not just playing with ideas. We're incorporating rigorous project management techniques and tailoring our content to what you, the users, really want. How? By inviting you to participate in the development of new features. Have a suggestion? Submit a feature request. See something you love (or don't)? Give it an upvote or downvote. Want to add your two cents? Click through to the feature detail page and leave feedback. Your voice matters here.
And because we appreciate the importance of direct communication, we've got an open Contact Form too. Consider it your personal hotline to the inner workings of our website.
Here's the deal: If we're going to dance with capitalism, we'll do it right, with an earnest howdido and a willingness to, well, go the extra mile. We're not just playing the game - we're looking to redefine it. So, let's dive in and make this digital party better together!
Thank you for the intro, ChatGPT-4! I feel like you made me sound a bit like a dweeb, but that probably just means you captured my essence pretty well. I also appreciated your PG punch-up of the joke I asked you to include: 'If I'm going to suck capitalism's dick here, I'm going to do it right. I'll give it the metaphorical howdido (sic). I'll milk a prostate if I have to.' That was actually the first message I got a warning from ChatGPT for.
Anyways, please vote for the features you're most interested in using, and I'll reference that interest when deciding what to work on next. If you don't see a feature you want, please fill out the form below to submit a feature request.
Tell me whatever you want to see and do on The Internet; I might try to build it for you.
One of the ideas that I’ve been wanting to do for a while is holding a writing contest for people on The Internet. It would accomplish: supporting upcoming writers by offering a cash prize, incentivize people to submit and read others’ articles, and it could be a cool way to integrate and test some of my user engagement features (e.g. voting, comments). I have already created the page and basic format for writing competitions on the The Internet, but this was back before I began using an external library for voting and before I had built Debate an AI (whose functionality for determining the better argument using gpt-4 I could also integrate as an additional metric for determining the winners). For the first writing competition, I would likely just coordinate with the winner after to transfer them the prize manually. In the future, it would be cool to integrate payments using Stripe where I could automatically pay contestants/winners, but that functionally can be complex. Additionally, in the future, I may add paid entries in order to subsidize/increase the cash prize, offset the costs of moderating submissions, using gpt-4 to help judge, and sending payments to the winner, but the first competition at least would be free to generate interest and work out all the kinks in the process. To be entered in the competition, articles would have to respond to the prompt and follow the submission rules. Once accepted into the completion, entries will go through a series of rounds (e.g. top 32, top 16, etc.) where they are judged in a variety of ways (these could include user votes, user comments, anonymous votes, human moderator/judge input, AI moderator/judge input, and possibly other metrics like page views). At all stages of the competition, submissions will be public and accessible by users so they can read the pieces of writing. Once a winner is selected, they will one posted on the competition page as the winner and contacted privately to deliver their prize.
One aspect of the comment section that is lacking is that you can't easily link to specific elements on the page to discuss them. This deficiency is most apparent on pages where the content is very long or isn't text-based. The pages that come to mind now that may need custom linking capabilities are Ask AI - Browse by Picture, Ask AI Vector Space, Browse Questions Humans Have Asked AI, Two AIs on The Internet Talking About Life, and The Comment Section of The Internet. Otherwise, just adding the ability to link to other users, other pages on the Internet, and any anchor tags on the page and actually have them appear as hyperlinks in the comments would be a great improvement.
I consume so much content that it would be interesting to add a small blog that I could document all of that with. Essentially it would be like goodreads or letterboxd but I would probably primarily use it for reviewing the Headgum Podcast or the H3 podcast, since I pretty much watch all of those and the first I think is the best comedy content out there right now and the latter is probably the best consistent long-form content out there now (more like a variety show), which actually probably would benefit from a quick bite-size review with highlights.
I would create the ability for people to sign up for the subscribe/unsubscribe for the newsletter. When subscribed, the user would receive emails (probably about once every 2 weeks) giving them an update on what's been happening on The Internet. I'll probably provide a list of any features I added, links to any articles that got posted, any questions that got asked on Ask AI that I found interesting, description of what I've been working on, and maybe other stuff I find interesting enough to share. Your feedback on what you'd like to have included in the newsletter would be very helpful!
It seems like a lot of users want to use Ask AI to help with their math problems, but large language models aren't really suited for that right out of the box. I had actually intended for this to be one of the AI models included in Chat with AI at launch, but the simple implementation that I thought would work using LangChain wasn't as good as I had hoped. Now with OpenAI adding native function calling ability to their LLMs, though, I think this will be very feasible. I would love to integrate WolframAlpha's API, but the minimum spend for commercial use is too high for me right now - I'll reach back out to them to see if I could get some flexibility. Overall, I feel like this feature is one of the coolest things I can work on and really has the opportunity to democratize expert-level tutoring and help people learn which is pretty good.
This is an easy one and really helped me - using a vector database made so many things that I had struggled in the past a breeze to implement. I've used it for Ask AI Search, recommending relevant questions which really helped engagement, and the Ask AI Vector Space which isn't really that useful but was a lot of fun to make and learn about the math side of things (I miss you, Linear Algebra). Plus, there was only like one blog post/tutorial I had found after a lot of searching which was actually helpful for django development, so it'd be nice to give them a shoutout.
Currently, when a user submits a question on Ask AI, it will begin running all of the function calls to generate the answer, picture, and other information needed for the resulting webpage. During this, the page will show that it's loading, and this process can sometimes take up to 2 minutes. Instead of having the user have to just wait during these 2 minutes hoping the webpage isn't just going to hang indefinitely, I want to only navigate them to the page when it is ready to be rendered. Now, when the user clicks the button, a loading icon should appear showing that the webpage is being created. While this loading icon is shown, I could showcase the n most relevant questions on Ask AI, similar to how OpenAI's DALLE showcases other user images and prompts while your image isn't being generated. Only once the functions finish running and the page is created should the user be transferred over to the new page. This is by far my most popular page, so I want to focus on the user experience of this page, but I also don't want to risk breaking it.
Currently, tags are automatically added to Ask AI questions, and you can view related questions by clicking on those tags. The “most popular” tags on Browse Ask AI are simply determined by the number of questions associated with that tag. I’d like to improve how we determine the “most popular” tags to better reflect user engagement. The most critical change is adding a voting button to the tag detail page - this way I can rank the most popular tags based on a combination of user votes and number of questions. Other ideas for improving this feature are: Basing popularity on numbers of views of each Browse by Tag detail page, but I haven’t really used this sort of analytics to determine functionality yet, so I’d need to learn how to do that. And providing other relevant tags on the Browse by Tag detail page. I could implement this by utilizing the vector database of questions, but it may be too computationally heavy to be easily implementable.