... Quick update since launch. I have spent most of this week improving reliability, Music.app integration and large library support. Playback with missing local files is now cleaner and should no longer trigger repeated Locate prompts. LMO can now detect moved or replaced files more reliably

... jagged frontier" to describe how these systems excel and then fall off a cliff with no warning. Their example is almost too good: an AI model can win a gold medal at the International Math Olympiad and still not reliably tell time from a clock face. Genius and incompetence living in the same box, and no obvious seam between them. That gap between capability and reliability

... tree for bulk actions. Larger PDF uploads: upload files up to 250 MB (previously 100 MB). Growing chat input: the chat box now expands as you type longer messages, up to 12 lines. Improved reliability: This has and will continue to be a core focus, ensuring that sharing content across devices continues to work smoothly. Going offline while saving content no longer loses your work. Saving recovers automatically once you are back online. Recall appears more reliably

... real shift, the kind of thing that quietly makes a whole class of orchestration scaffolding unnecessary. The thing I'd still want to know is whether it holds at the exact worst moment, not on average, because average-reliable and reliable

... frequently encounter 401s and 403s when requesting an HTML page with fetch or curl. pure.md is a REST API that lets developers reliably access web content. With pure.md , you can: Avoid bot detection by mimicking real user behavior Render JavaScript-heavy websites , PDFs, images, and files Scrape web pages into markdown optimized for an LLM Crawl search engines for up-to-date knowledge Extract JSON from web pages using natural language It's almost too easy to get started: simply ... ... with ` https://pure.md/ `. I use it myself to easily pipe webpage content directly into an LLM, knowing that all the superfluous HTML will be stripped out so that the input token count remains low. You can use it reliably