How can @ChatGPT be so confusing? Chats on the phone take super long time to sync to the desktop app. I can see MCP on desktop but not on the phone?? Why don't they merge chat and work modes together? So damn confusing.
CI was running on Karabiner, my home server. With ~10 parallel threads in Nexure (my agent harness), it became the bottleneck.
Moved to @UbicloudHQ today: auth + change the runner label. With some test fixes: 20 min β 3.
In a week: seeing what it costs π
@tinchoex Eso pense yo. Y en realidad es una mierda porque no puedes mover el mapa a tu campo de vision y pierdes toda la visiΓ³n de la partida. Aunque parezca que veas mucho mΓ‘s mapa solo prestas atenciΓ³n a lo del centro.
Yo termine jugando en la resoluciΓ³n estandar por eso.
I keep thinking about how much better does a new model have to be before it's worth paying more for? I'm on Opus right now with practically unlimited usage, and it handles 80β90% of what I need. I think this new model it's such a sweet spot between quality and price.
@MrB2Beast I guess patience and giving the best customer support you can. I've the feeling the relationship you grow with the customer is probably the most important factor for a good retention.
Nah, they were really not flaky but it looks there were still some performance improvements we could do. Plus let the agents figure out the best server for parallel testing. Turned out 16 cores is not much expensive than 4 since timing drops significantly. But after that is not so worth it.
Another good thing about @UbicloudHQ - you change one string and you get different powerful machines.