gkbrk

> whereas AI generated media is problematic

It's not problematic to 99% of the population that's not part of a very vocal minority on Twitter and Bluesky. Everyone else is enjoying AI art and making funny family pictures and memes together.

Memes I share with friends on Signal and verification codes from my bank with my life savings require very different levels of trust.

Why would I give any app permission to read my messages, no matter how secure it claims to be?

Frida can usually un-pin those pretty easily.

AI works much better when it gets immediate feedback from a good type checker

Top contestant in the Hutter Prize uses a neural network for compression. So fair to say, LLMs would perform pretty well compared to gzip.

Official Deepseek v4.1 Flash API costs are more than GPT 5.6 Luna. Deepseek v4 Pro performed worse than Luna, so I wonder if 4.1 Flash will justify the cost.

Clearing cookies when all you want to do is read static content is not "breaking expected behaviour on websites".

What's inconsistent about their versioning?

DuckDB is still OSS btw. It didn't go proprietary like Redis or ElasticSearch.

Hint: It's not 15%. And there are many reasons, such as not having to keep up-to-date billing details in 70 providers, and not wasting money because most providers want you to prepay a balance that gets stuck in there if you switch to another provider.

It also has way better uptime than the underlying platforms, even for proprietary models like Claude. When Claude APIs are having issues, OpenRouter Claude still keeps working because they can route to AWS Bedrock instead of Anthropic etc. This effect is even bigger with open-weight models because they typically have 5-10 providers.

It's such a bad idea to use a random website to generate passwords, surely this has to be bait.

> Strong passwords, generated only on this device. Nothing is sent or stored.

> Private by design

And the only way to verify this is to read and understand 288 KB of obfuscated Javascript code every time you launch this app.

The Offline test ranking of Tracking AI lines up much better with my experience of using the models.

https://trackingai.org/home

This is more app instances than 99% of deployments. Most apps and websites run a single server.

Really glad open-source password managers are resisting the bullying and not implementing DRM.

If it's designed well, your application also opens a pool of connections and re-uses them each time the code needs a DB connection.

It's the #6 most popular website in the world. Chances are, someone will fix the libraries or investigate and write about any changes done to the internal API pretty quickly.

It still has an internal GraphQL API. How would the official web app and mobile apps work without an API?

It's still open-weight models that you can download and run locally. You don't need to send anything to China.

Most people would consider Europe a continent. What kind of strange definition of continent are you using?

https://en.wikipedia.org/wiki/Continent

If the model cannot figure out simple and ubiquitous tools, how is it supposed to figure out complex problems? All of the good models basically work with any harness, including giving them a single "shell command" tool. They can just figure things out.

You’re insane if you think supporting Russia is not “more wrong” than supporting Ukraine during a war Russia started to invade Ukraine.

Wine has better backwards-compatibility than Windows these days.

It's always on The Pirate Bay. That's the streaming service everything is on.

The best advertisement for these companies would probably be running a public Polymarket account so everyone else can quickly check how accurate they are.

I wouldn't classify getting random malware as "nothing happening".

A monitor cannot install software on your computer by the way. It's Windows installing this software automatically (for some reason), so the blame should be on Microsoft.

Autorun of malware when you plugged in a USB drive was also a Windows issue, I'd classify this as the same security problem.

The same server that's responsible for sanitizing garbage JS out of user content is also responsible for sending the Content-Security headers. Why would you trust it with one, but not the other? If it's buggy garbage it will also send the wrong headers.

All the codecrafters content is available on Github too, for free.