Building AI infrastructure & agent systems. Creator of Jarvis Desk / Agent Accelerated. Dev @ SSHGrid. #BerryCore maintainer. Teacher of code. AKA SW7FT.
This Canadian woman was given MAID as her only choice for stage 4 cancer in Deathcare Canada. She went to the US and death was not the option the doctors there gave her.
Canadian healthcare used to be good, it is now in desperate fear of collapse. They're bringing in doctors who did not study in Canada and whatever else that entails.
I find the healthcare in the US unquestionably superior. Health insurance is the same since I don't pay income tax in Florida, I pay it on insurance. There is immediate, private healthcare if you need it.
Medically assisted suicide is not an option they offer.
I built a modern #chromium for #blackberry 10 to prove to the world how good #QNX is.
#QNX adoption will rising quickly in the age of AI... and these old devices taught me why :)
https://t.co/BXRpc3YXxU
@sahildarz Well tmux + injected context.md at model
Load , then each agent is scoped
With a ssh key, rocket channel and unix user. I keep optimizing but I think I’m on the right track :)
@burkov I think all the smart ppl saw the death of those two companies a long time ago but capital has kept them alive and the programmed loyalty to Microsoft as well
@ChrisTranGG I’m a dev ops veteran and my quality of life is 100x better because of ai. The real test of intelligence now is “how many tokens to build donkey kong”
@someguy789@kingwilliam_ No I have 153 and barely burn through 10 dollars a month. My agents only use the shell no gui and are injected with task refined context on boot
@EdwinCh43136084@kingwilliam_ No it’s not, tokenization can be optimized incredibly for the task. For example telling an agent what tech stack to use and how to use it could save 100x on tokens
@kingwilliam_ The trick here is context window management and context management I have a 153 agents all tied to their own VPS they access it via ssh keys. Never spent more than 200/mo compose 2.5 is basically free at this point and can handle the majority of tasks if context.md is good
@sth1d@3rdWorldBritain@tomhfh perhaps in that statement lies your point. Yes I am doing the thinking, but AI is navigating the needless abstraction layers of code that was built my humans to increase my productivity