HN user
This is ridiculously fun. Congrats on the launch! I took inspiration from “There I Ruined It” and grabbed lyrics from various popular songs to have the AI sing them in the style of other artists. It sometimes took a few attempts, but it honestly did a great job. You got a chuckle out of my friends and family. Also loved that I didn’t have to enter a credit card in order to try it out.
Is this limited to PDFs or could the same chunking and parsing be applied to plain text, html, and other input file types?
There is some interoperability between other clients today. For example, I have a Jellyfin server but in order to connect to it on my Apple TV I use the Infuse Pro app. I think that works more because of a standardized file structure that each app creates its own indexes for, so an actual protocol that they share would still be an improvement.
There are some really cool geoguessr metas that aren't as much about memorizing the street view cars that I love. (Although, a few car metas are interesting such as the handful of African countries that required a security follow car to tail the Google car, or the “Kenya snorkel” on the more off road capable street view vehicle for Kenya.) One of these is that there is a geostationary satellite just south of Texas that most US satellite dishes will point to, so if you can find a house with a satellite dish on it then you can combine its angle with your compass and get a rough sense of the US region you are in.
Are the testimonials also AI?
My current setup is Stellarium + Clear Outside + windy.app and with that you really do need to know ahead of time which objects are visible for your particular location at that time of year or spend a while browsing. Your app looks amazing for my main use cases, but still lacks some information that I rely on. First, it is useful to know the direction of the cloud coverage throughout the night to be able to compare it with where the object you’re shooting will be. I also like to know the brightness (magnitude) of each object to help with planning. The last thing would be to get a list of recommendations rather than using lists of favorites. “What deep sky object do I not know about that today would be perfect for shooting?” is a question I have only been able to answer well with the ASIAR’s list of suggestions, manually curated calendars online that are not tuned to my location, or social media. It seems like you might have all the data to do a better job than all of those options in answering that question. Super cool app! Excited to see where it goes.
As a native speaker, it was my first time seeing the word, too. Not common at all.
I’ve been working on a project [1] to do just that from within a Chrome extension. The idea was that as an extension, it could make use of the context menu and feel more like a native feature of the browser. I’m always hesitant to link to my things from comments but in this case I think it’s a perfect fit for what you’re describing.
Richard Hamming’s The Art of Doing Science and Engineering is one that’s really shaped my philosophy on CS. It pushes for the importance of reasoning from first principles, experimentation, and taking on extraordinary work. There’s also a fascinating and prescient section on AI and the limits of computers and how we think about them. Stripe Press makes a nicely bound hardcover edition of the book, too:
https://press.stripe.com/the-art-of-doing-science-and-engine...
Heartily recommend!
One of my biggest concerns when saving lots of things to an app like this and spending time curating bookmarks and other content is what happens if the app one day disappears. Is there a data export mechanism built in or other considerations for these sorts of events?
These theories are fun. My first instinct was that maybe it’s for “yeoman”. But that’s not the most kid-friendly or likely word to appear on a ball. “Yellow” seems a lot more plausible, and the spray paint can idea is especially clever.
Almost every day for my 5+ years of college, and fairly regularly since (although I encounter the types of complex math problems it excels at far less frequently today). I think there are many others like myself who use it.
This is so intuitive. Huge props for coming up with the first ui for time zone comparison that has actually made sense to me immediately. Clicking a zone to make it the base, which all others get compared to, was a great choice. Bookmarking this for sure.
This leans on the LLM’s latent space to hopefully give you better results. The idea is that it’s able to use those key words to form connections that may have been absent otherwise.
One idea is that the url makes it immediately shareable. You can send your note to your phone in a text message without any uploads due to it being encoded in the url.
This pops up again later when he says 81 is 10000001. Or that width 8 corresponds to 04. I don’t know enough about the gif format to know if I just misunderstood these parts or if they were written incorrectly, but it was a bit confusing.
Just at a cursory glance, this looks pretty verbose and clunky. Existing React ecosystem styling solutions are many, so it’s not immediately evident to me when you would accept the trade offs introduced by this tool over any existing ones that seem much nicer to not only integrate, but also have to work with day to day.
The idea that computer code can't be a company name is just begging for clever company names to skirt this rule, especially with so many languages that are light in syntax.
SQL is a natural contender with potential queries like “select customers from store” but I'm curious how far this can be taken and what other “computer code” company names other languages would make possible.
For anyone looking to dip their toes into Yabai and tiling windows, I recommend the following setup. Enable full tiling only on a couple workspaces. Desktop 1 can remain in floating layout, just as though you’d never installed Yabai in the first place. Then tile Desktop 2 and as many others as you want.
The one other tip I have is to have an easy hotkey for toggling full screen mode. Toggling an app into full screen mode doesn’t lose its tiled position or create a new desktop the way native Mac full screen mode does, which is especially helpful when working on a smaller monitor.
It’s cool to see how certain prompts and themes stay relatively stable, like the gnome example. But then “cat lecturing mice” quickly goes off the rails into weird surreal sloth banana territory.
My best guess to try to explain this would be that “gnome + art style + mushroom” will draw from a lot more concrete examples in the training data, whereas the AI is forced to reach a bit wider to try to concoct some image for the weird scenario given in the cat example.
I like the concept and would really like to have something like this (especially the combination of recipe + shopping list—that’s great), but the ux is a bit off. Having to fill out a form and then click “buy now” before I get a sense of what I’ll actually be getting or even what the pricing is feels wrong. Are the meal plans handmade or AI generated, or pre-generated and just matched to your preferences? What exactly does “personalized” mean in this context? There’s very little that gives me trust (reviews, detailed product information, a free demo or maybe a video example) so you might consider adding at least some of those elements. The background gradient in dark mode also makes the text hard to read when you get to the bottom of the page. Ironically it sort of becomes a light mode theme if you scroll far enough. And at least on mobile, the hamburger menu with only a single link in it is a funny choice. Why have the menu at all? Hope you find this helpful and not too critical because, again, I think the concept is totally awesome and there’s a lot of potential here.
What’s the use case for passing in an array of numbers? Typically when generating an ID my input is either a single random number, a string that’s being hashed, or nothing at all.
This list excludes the way 80%+ of the commands that the leadership at my company issues tend to be done. Sure, small one-off requests from someone senior might be phrased in some of the ways mentioned in this post. Or, more commonly, something a bit more conversational like “hey can you do X”, or “we need Y by Wednesday, are you up to the task?” But the vast majority of commands from higher up aren’t issued as commands at all. Bosses don’t just tell people what to do next, it’s a collaborative process and a discussion. The team is aligned on our future goals and there’s a degree of self selecting projects to be assigned to, where the things we work on will flow naturally from our project goals. Trying to pick the best wording for what really amounts to opaque non-collaborative decision making from the top just seems like a terrible company culture.
For the Assistants API with unlimited context length it’s not clear to me how the pricing works. Do you pay only for the incremental tokens per message for follow up messages, or does each new message cost the full prior context amount + the cost of the message itself?
I think most of this would be resolved by just reducing the left and right padding on small screens. If the author is monitoring this thread, please consider making that update!
I think most of the Jellyfin ecosystem is open source. Was curious if the code for your desktop client is, too. Would be fun to poke around the GitHub for a project like this.
Cool! Just curious, why do you query 10 times per second for a feed? This seems kind of excessive and maybe a bit bad for battery life, especially since this could be running in several tabs at once.
This is a great use of the sidepanel API. Really like the idea and how you implemented it. The Firefox equivalent to that API isn’t identical to Chrome’s, and I believe that there is currently no way to trigger the side panel to open programmatically unless it’s in response to a user action. So it may take a bit of work, but would be exciting to see support for FF in the future.
A bit tangential but I was curious what you used to record the demo video on your landing page with those zoom-in animations during critical moments. I’d like to record something like that for some of my side projects and thought your video looked rather polished.
Not only this, but I imagine some of the less tech savvy end users (kids, grandparents) would be the ones who found the game from one of these parasitic sites inundated with ads. The goatse image doesn’t help those users. They probably won’t realize what that’s about and then just stop playing altogether.
This is pretty funny and reminded me of the first website I had ever made (probably sometime around 6th grade, back when <marquee> was cool). I’d found that most of the flash games I enjoyed had either embed codes or could just be embedded manually using an iframe. So I made a library of my favorite embedded games in a small folder of HTML files, which I FTP’d onto some server I had access to for some reason. If this goatse alternative to CSP had been used I am sure I would have gotten in a lot of trouble, which is funny to think about today.