Sebastian Deutsch

Since Claude is down we probably can see a traffic spike on Pornhub for all the vibe coders.

YouTube video by GameTrailers Cast n Chill - Official Launch Trailer

This is a catchy game trailer: www.youtube.com/watch?v=qclh...

Cool. Do you have a mix online (SoundCloud)?

Nice one! Just checked out your website... out of curiosity: What kind of music are you DJing?

Apple has a private CSS property to add Liquid Glass effects to web content I have an incredibly boring summer hobby: looking at the changelog for the WebKit Github repo. Why? Because I spend a chunk of my professional life working with webviews inside mobile apps and I like ...

TL;DR Even Appe is so lazy that they're using Webviews in their apps for displaying liquid glass - but you can't:

alastair.is/apple-has-a-...

Apple’s new Liquid Glass: stylish effect with fluid animations - reviving early-2000s Aqua vibes. Critics see issues: accessibility, wasted GPU power, and nostalgia instead of innovation. Apple's goal? Reinforce native apps over web. But real fix needed is App Store economics, not just UI tweaks.

Mistral launched Magistral, a reasoning model. Glad to see a european model provider. A bit sad that only the small one is open weight and that it's lagging far behind Deepseek. The medium one is quite good.

Announcement: mistral.ai/news/magistral
Full Scoop Paper: mistral.ai/static/resea...

Apple Glas :-) Who of you is old enough to remember Aqua?

OpenAI released Codex; I’ve been underwhelmed by Sonnet lately: I asked an agent to implement Rails/Turbo drag-and-drop, but it produced subpar code with a nasty page reload instead of Hotwire. Only step-by-step instructions yielded an acceptable solution. Let's see how it measures up to Codex.

Very welcome!

To all the vibe coders out there. A nice compilations of blindspots that an AI has while developing software:

ezyang.github.io/ai-blindspots/

#llm #mcp #ai

Hallucinating like a LLM

gemma3 The current strongest model that fits on a single GPU.

Google just released their open source Gemma 3 LLM - you can access it easily via Ollama ollama.com/library/gemma3

Canadians answer right now: All american users of pr0nhub must register if they they visit the site longer than 30 seconds.

Mistral OCR | Mistral AI Introducing the world’s best document understanding API.

I've dabbled with a lot of vision models (ChatGPT Vision, LLAMA Vision) to replace our OCR chain (Google Document Cloud and then an LLM to clean and process). Unfortunately they were hallucinating too much or gave suboptimal results.

Mistral OCR looks really promising:

mistral.ai/fr/news/mist...

Thanks <3

Thanks <3

What do I need to do to get added?

@joshuawood.net kudos to the Honeybadger EU region. Big fans of your service. BTW: Any chance to get on your Rails Starter Kit list?

@rsmwall.dev hello - what would I need to do to get listed in your Rails starter pack?

Happens :-) aber danke.

Danke, aber wie kommst du auf Bielefeld?

Package Version MCP Server - For AI Tools Help AI tools like Cursor to automatically use the correct package versions when adding dependencies to your projects. Supports Hex, RubyGems, and NPM packages.

Yesterday, I released a tiny MCP server that allows Cursor Editor to get up-to-date version information for external Hex dependencies. Someone suggested doing a version for #RubyOnRails, so I built a more generic version that supports Ruby Gems, Hex, PyPI, and NPM.

dep-mcp.9elements.com

To be fair: There was a sign that stated "Use code with caution".

Request for developer feedback: customizable select  |  Blog  |  Chrome for Developers An early look at the new customizable select feature.

Never seen this before - a request for developer feedback:

developer.chrome.com/blog/rfc-cus...

Now is the time to get rid if all the JS libraries that are necessary to style a <select>.

I'm switching from Laravel to Rails I have been using Laravel since version 4 in 2013. Over the years, Laravel has evolved significantly. I initially chose Laravel over Rails due to...

I'm glad that this one was settled once and for all :-) old.reddit.com/r/rails/comm...

I'm trying to do importmap only. But it's really hard. Especially when you need some js at build time.

Are you using importmaps or vite/esbuilt?

DeepScaleR: Surpassing O1-Preview with a 1.5B Model by Scaling RL | Notion Michael Luo*, Sijun Tan*, Justin Wong†, Xiaoxiang Shi, William Tang, Manan Roongta, Colin Cai, Jeffrey Luo

Interesting read: Towards Democratizing RL for LLMs

pretty-radio-b75.notion.site/DeepScaleR-S...