Dude, this is the funniest thing ever
It made up a feature that I can close ghost processes, right
so I instantly go back to my computer I don’t look at the conversation
I go and just click close all, then I go to the conversation for the first I’m saying, but please don’t use the close all button🤣🤣
Didn’t know roofr was a yc company!
Roofr (YC S17) Raises $3M To Make Roofing Transparent: ow.ly/5kqt30nByfb
I really hate when The NASA scans don't include rotor inertia data, so I have to estimate F100 rotor inertia from published engine size/weight
The next wave is coming
Any opensource model lab, I would love to test your new models please dm! (Only if benchmarks are frontier os
This doesnt matter anymore when you stopped people from using anything other then grok, terrible change imo
Scary i love it
Camera works: 1280x720, face found in 59/60 frames, irises readable. Now the rest of the stack.
I hit my session limit (irl). My reset is in 8 hours (sleep), but I still have many active chats (things I need to get done).
I need a reset button (a white out monster) or to get another plan (just deal with it).
Anyone else have no clue what this means
I will die on the hill that file-based agents are the best agent architecture.
NO ONE wants to set up all their different agents every time a new model drops.
- No moving agents around harnesses
- No moving agents around computers
- No managing skills in multiple harnesses/computers
- Skills only apply to the agents that need them, reducing context
- Custom system prompts for different agent tasks/roles
- Each agent role can use different subagents
Literally all I have to do when a new model drops is open that specific agent's settings on any 1 of my machines and select the new model/harness.
It automatically propagates to every future thread across all my harnesses/devices.


















