Biggest mistake from GitHub: rate limits on org repos interactions. As my goal directed agent swarm for making Lin the go to high performance language matures, I am limited by the rate limits of GitHub. Time to change I guess.
Computational Science was monopolized by big Corps and big academic groups. You could not do what you want. That is going to change with my new language Li.
Via Gmail APi. Super simple just made a simple agent harness that goes through the mails and classified them etc using qwen-9Bn to save costs because the token consumption is crazy high like 100 Millions per week. Hardest part was clicking through GCP.
It all started with an application automation tool to apply to all cursor jobs. Then Ben said you like crafted applications so I tested it on Indeed. now I got a sales job and taking my to find time again to craft a good application for you guys.
Dreaming to work at Cursor it's the best AI company in my opinion. But also need some time for projects. Haha not sure why I am telling you this. Probably won't help my application...
Token usage for high-performance languages using Cursor SDK and Qwen-20B models. Li is super expressive and has some overhead because of profability. However it is using less tokens for thinking than SOTA c++ and rust. Only julia using less (tested 70 reps of a physics benchmark of 10 different tasks)
Li server (lis) is best in class on performance. Replacing treafik ingress on my kubernetes cluster. Sometimes lighttpd is better though. Working on that. Key are bitwise operations here which I implemented into the compiler.