Its for purchasing AI summaries. By default each user gets 1000 article summaries to begin with, which should suffice for several months for most users.

Initially I didn't want to put any in-app purchases but decided to add it to prevent abuse.

Glad that you are liking the app! Import/Export feature for bookmarks is on my roadmap.

OpenAI just released Codex, which is basically the same as Claude Code.

It looks the same, but for some reason Claude Code is much more capable. Codex got lost in my source code and hallucinated bunch of stuff, Claude on the same task just went to town, burned money and delivered.

Of course, this is only my experience and codex is still very young. I really hope it becomes as capable as Claude.

Part of it is probably tgat claude is just better at coding than what openai has available. I am considering trying to hack in support for gemini into codex and play around with it.

I was doing this last night with open-codex, a fork. https://github.com/ymichael/open-codex

These datasets would definitely have a lot of Text => Sketch pairs as well. I wonder if its possible to extrapolate from Text => Sketch and Text => Image pairs to improve Sketch => Image capabilities. The models must be doing some notion of it already.

Can this be used to clean podcast audio? I don't see a reason why it can't be...

Adobe already has such a product: https://podcast.adobe.com/enhance

I have used it for quite sometime. The v2 works fairly well now but still not _great_.

You can use krisp.ai during recording to already remove a lot if not all noise. Not sure about cleaning up in part but I suspect there are a bunch of those tools available as open source as well.

Elevenlabs have a good noise reduction tool for that

That is an interesting observation. I wonder what percentage of the training text data for LLMs contains proper dashes, since a large part of it is user-generated content.

All self-respecting journalistic outlets use proper symbols. Where does the LLM get their opinions on “foreign affairs” from? Probably from the likes of New York Times like a standard lib...

And it shouldn’t be hard for an LLM to learn to use proper symbols when synthesizing content from the everyman. It’s not like it works on the level of literal copy and paste.

> Where does the LLM get their opinions on “foreign affairs” from?

Yeah that's a big problem with LLMs at the moment - people are having an argument online, and one of them posts a piece of text from an LLM and says "see, Grok/ChatGPT/Claude/Gemini agrees! You're wrong!", which only reveals how much they did not do their homework and that they're arguing on the basis of hear-say, just like the LLM; the LLM has no mind of its own. If 90% of the internet says A, it will defend A to the death. What's not part of the equation is how much of that 90% has the same source, financed by a certain group.

We have a LOT of work to do soon.

People are coming to ChatGPT with an intent to type. Its much harder to introduce text-based interface to otherwise click-based interfaces because of the friction. Thats why chatbots often fail to gain adoption, because they are a byproduct of the product team being lazy and not thinking through the UX.

I 100% agree that grafting a chatbox on an existing app that wasn't meant for it sucks. (I'm working on an AI app where I set a "no chatbox" constraint, see my profile.)

However, I think a lot of people are willing to type if it will do something useful they can't otherwise accomplish. And: typing gives you user intent, very valuable.

Thats why I go for the nuclear option (the one that requires higest attention, like account cancellation) when presented with a list of options by a bot. More likely to get a person that way.

I have always assumed that I could just kick the window to break it open in a situation like this.

When I was a kid I worked in a double glazing factory. Some of the old timers demonstrated just how hard it is to break a piece of toughened glass to me by whacking one with a bit of wood right in the middle of the pane, really really hard. Nothing happened. Then the dude tapped it really quite gently with a glass breaker in the corner (where it can't flex as much he said) and the whole thing exploded into those little cubes you see on the ground in dodgy car parks.

Car windows are probably harder to break than you think...lol. Also - consider the situation where the car is in, or partially in water, and pressure differentials.

(There was actually another HN thread about this recently - https://news.ycombinator.com/item?id=39691780).

I keep a Resqme in the car glovebox - my other half used to do a lot of interstate driving, and I was always worried she'd be trapped in the car. The Resqme has both a seat-belt cutter, and also a centre-punch for easily breaking the side windows.

This is a genuine question: how useful is the resqme in a glove box? Do you think the driver could reach it in the event of an accident, or do you keep it so the passenger can use it to free the other occupants?

I'd clip it to the seat belt pulley. But round here you'd have difficulty reaching speeds to have a crash, what with all the minis doing 20 under the limit

Side windows in most new cars are also laminated. The quarter windows are the one you should try break.

There is a mythbusters episode about exactly this that is definitely worth the watch.

windows are normaly tempered glass. You can't kick them out.

They're also unbelievably strong when they're completely rolled up and incredibly weak when they're not.

Although a tiny piece of ceramic at just the right speed will shatter any window.

And I recommend a spring-loaded window breaker, as opposed to having to, like, physically hit the thing against the window.

This is what we need.