@neogoose_btw yesterday I hooked up fff mcp to opencode, but I never seem to see it actually use fff for grep / file search. It just keeps on using normal grep for searching. MCP status healthy etc. I'll try extra instructions to use fff mcp for that next
@thdxr Not sure whether this was to do with opencode or the model, but when I started using haiku over sonnet 2 weeks ago, it started to write code in plan mode, suddenly committing changes (it never did that before) and ignoring instructions.
@dhh I love what you did with Omarchy, but I don't like the bleeding edge nature of Arch. So, I worked on a compatible installer to install Omarchy on Fedora 42. Early screenshot so far!
Automatic dark mode in GNOME desktop environment. Put script in a cron to run every minute or so. Location based on public IP geolocation. Caches location, so will work in offline mode as well.
#GNOME#darkmode#freedesktop
https://t.co/HJGAXx3e2X
@jhuskisson@SlackHQ I really wonder who is going to end up paying all the AI hardware in the end. I feel like the AI bubble bursts when the ROI is negligible for too long and it doesn't make sense to make any further investments in AI hardware/resources
@Hasen_Judi Completely agree. One side note I'd like to add is that sometimes a cache can be used with precomputjng. You fill the cache beforehand (either on save or by a background process) and let the query only look in the cache. If it's not found, then it's simply not there!
@falconerd Haven't had an invite for the jai beta yet, so only seen online examples so far.
Things like play_sound vs sound_play or add_entity vs entity_add. Sometimes the prefix sounds / reads better, so then I ponder to not be consistent with the prefixing