← Feed

Posts

• Post #2852723

  Working on an exciting project that touches hundreds of transcripts this week (alongside at least three other 'normal' context size projects in parallel) has put me in claude jail repeatedly. Agent teams are amazing, and they have made it clearer to me than ever that more tokens=more speed. I have *way* more ideas than I have tokens, so if we desired we could simply pay to move faster *without losing quality of output*, it's wild.

• Post #2852722

  RE: https://mastodon.social/@arstechnica/116059409468374734 This must have been part of the Gemini deal?

• Post #2852721

  One sneaky but powerful aspect of skill.md’s I’m reflecting on is that they’re as effective as the model leveraging them. So as long as the skill is well crafted, you should expect better results ‘for free’ every time a new model drops

• Post #2852720

  IYKYK

• Post #2852719

  if I upgrade to max 20x will they let me have multiple chat/cowork/code windows open at once like I can with code in ghostty?

• Post #2852717

  “compression and decompression, otherwise known as codec.” I have worked in video production for essentially 30 years and I did not know this is where codec comes from 🤯 https://daringfireball.net/linked/2026/04/02/buck-quicktime-history

• Post #2852716

  There’s a certain poetry in Adobe messing with my hosts file today as a fully paid user when years ago I would edit my own hosts file to black hole their license checks

• Post #2852714

  auto mode on max when