The page accompanies a blog post about how many images have alt text on Bluesky so empty lines are included to give a feel for how many images do or don't have the descriptions: https://digitalseams.com/blog/image-descriptions-on-bluesky

> This demo is part of Image descriptions on Bluesky: not bad, could be better - see the blog post for more details!

Surprised then that so many posts are just images.

The pricing page mentions how many credits you get but not what a credit does or gets you. Could you elaborate on that?

But with AI it never tells you that. I pay $xxxx for the Claude API, half of the time it breaks it's own code and just does things explicitly told it NOT to do; I think I want to be refunded for the time wasting and frustration, but that's not happening. So you pay credits and you get 'stuff' that 'might work'; that's the reality. I'm gonna say this is basically not even allowed for consumers in the EU, but I don't know that for sure.

Pricing credits - dollars is explained below. +We were thinking about not charging credits whenever one of our agents fails because of a random error on our side (not too frequent, but still). This makes it more reasonable to start exploring and building with us.

Makes sense, thanks.

It's simple! For $50 a month we give you a thousand smeckles which you can use to crank the floba. A very good deal.

Just explained below what you can do with the 10K smeckles - that's 10K URL scrapes or 5K agent steps or 10K minutes of browsing :)

To reiterate, "what a credit does or gets you" is not explained by your pricing page.

I'll operate on the assumption that it takes roughly 500 credits to scrape a single page for now.

Oh - missed that too. We had this somewhere in the docs and got lost.

- 1 scrape URL : 1 credit ~ $1c - 1 agent step : 2 credits ~ $2c - 1 min browse-time: 1 credit ~ $1c

Meaning if you run an agent for a 10-steps task for a minute it'll cost ~ $21c

Update- https://www.notte.cc/#pricing

I just wish their realtime audio pricing would go down but it looks like GPT-5 does not have support for that so we’re stuck with the old models.

See also: https://github.com/drathier/stack-overflow-import

    >>> from stackoverflow import quick_sort
    >>> print(quick_sort.sort([1, 3, 2, 5, 4]))
    [1, 2, 3, 4, 5]

I believe you need to use the audio element specifically. The Web Audio API is subject to different restrictions than the audio element. I used a similar approach on Audjust: https://www.audjust.com/blog/unmute-web-audio-on-ios/

(nice site you created btw! I love seeing audio stuff for the web)

Thanks mate! I'll implement your solution:)

I think I know of some of the other MCP posts you’re referring to. I saw multiple posts recently about AI controlling browsers using MCP and thought that was interesting. I’m not sure if it will be a fad or not, but I’ve been trying to model browsers as an HTTP resource instead as a side project. This allows you to HTTP DELETE a tab or POST /browsers to create a new browser. I think it might be a more natural way of using an API plus it can just use the classic auth in the form of API keys.

relevant: TabFS <https://news.ycombinator.com/item?id=34847611> <https://github.com/osnr/TabFS> (MIT) and <https://github.com/balta2ar/brotab> (MIT) also showed up while I was trying to remember the name of TabFS

In case you haven't seen it, the MCP repo actually has Puppeteer as one of their example servers: https://github.com/modelcontextprotocol/servers/tree/2025.3....

Thank you! I had seen TabFS before, but the others are new to me. Will take a look! It's nice to see demand for this.

The article mentions that "The names do not affect outputs using OpenAI's API systems or in the OpenAI Playground (a special site for developer testing)."

I have been working on Audjust (https://www.audjust.com/) on and off in my spare time. It's a service to manipulate (shorten/lengthen/loop) audio for video editors and music producers.

I had a Show HN a while back that was well-received and kicked things off (https://news.ycombinator.com/item?id=36480687). Since launching I have changed the name and added paid accounts which have brought in enough money to cover costs and make some profit!

I agree, the German translation is pretty bad (phrases like "Hier ist, was sie über Quicklang.", "Es ist ein Durcheinander. Sie wissen Bescheid.", and "Du kannst all meine Programmierprojekte auf X" are incomplete or awkward).

It also doesn't sound like this can handle dynamic phrases like "buy {{count}} items"?

When you use machine translation, you must be aware that the fact that text is not in the original language does not necessarily mean that someone else speaking another language can understand it.

When I took a look at using DO last time I decided against using them because you had to pay for the duration your clients are connected to them when using WebSockets. As far as I understood it’s not execution duration but flat out how long your clients are connected. You’ll burn through seconds quickly that way. I decided to go with a polling mechanism instead.

They have hibernating durable objects now that don’t charge unless there is a request through them: https://developers.cloudflare.com/durable-objects/api/websoc...

Oh that’s neat! I’m not sure if it didn’t exist or if I didn’t find this last time. Looks like this would address the problem.