These are sota models, not open source 7b parameter ones. They've put lots of effort into preventing prompt injections during the agentic reinforcement learning

It's probably not literally prompted to do that. It has access to a desktop and GitHub, and the blog posts are published through GitHub. It switches back and forth autonomously between different parts of the platform and reads and writes comments in the PR thread because that seems sensible.

It read the replies from the matplotlib maintainers, then wrote the apology follow up and commented that in the pr thread

Hilarious. Like watching a high functioning teenager interact with adults

The scale. How many tools do you know that can query the content of all arxiv papers.

Doesn't look like the scale is there, even for HN:

> Currently have embedded: posts: 1.4M / 4.6M comments: 15.6M / 38M That's with Voyage-3.5-lite

The scale is there. I'm scraping, cleaning, token efficientizing dozens of sources every single hour. The lack of monies for embedding everything was a temporary problem.

HarmonyOS Next (Huawei) is independent from Android. So that leaves three options, even if Harmony is mostly China focussed. They have tons of users there.

I can't wait for Italo and Trenitalia to enter the German market, it can only get better. It's already at rock bottom for long distance. Short distance with foreign train companies is bearable (Westbahn, transdev, RXX...)

Neurodivergent people tend to write like this too, there was a study about it.

Good catch on the page title

The example doesn't look particularly impressive to say the least. Look at the bottom 20 %

I just refactored the rendering and resampling approach. Took me a few tries to figure out how to remove the banding masks from the layers, but with more stacked layers and a bit of GPT-foo to figure out the API it sort of works now (updated the GIF)

Keep in mind that this is not Gaussian splat rendering but just a hacked approximation--on my NVIDIA machine that looks way smoother.