Remember your first Hello, World?
I felt that again today ,but not in python , rust or c .This time in my own language.
fn main() {
print("Hello, World!");
}
Brix just compiled its first program to a native binary.
Just getting started. Miles to go.
https://t.co/fMWrmejEsy
Building Stash, a small vault for all the UI snippets, patterns, and interactions I keep reusing while building.
Live previews, clean organization, and copy-ready code.
Still polishing it, but it’s slowly becoming the tool I always wanted for myself. ✨
Also, all the beautiful shadows on these buttons were created by @rohitmehta_twt.
Phase 1 (lexer) is now 100% complete for the EEL compiler.
The final edge cases for token scanning are resolved, and type declarations are stabilized.
Phase 2: building the parser and constructing the AST.
#eBPF#Compiler#CProgramming#Linux#SystemsProgramming#BuildInPublic
Advanced Certificate in Data Science from IIIT Bangalore done! Advanced Certificate in Data Science from IIIT Bangalore done! Specialized in NLP and ready to build something meaningful with it.
#DataScience#NLP#IIITB#AI
we’re cooking something big & applying for @Arcium RTG launch coming sooner than you think.
Don’t get left behind hop on the waitlist here:
https://t.co/xxxXC0KRLE
cc : @SuperteamIN@superteam
just submitted rpc-stats to @colosseum Frontier.
the idea is simple: every RPC provider claims they're the best at landing transactions. none of them can prove it.
so we built a system that tests helius, triton, and alchemy every 30 seconds. on-chain verified via geyser. live public leaderboard.
no marketing. just data.
built with @Subhdotsol 🤝
Introducing Skema → a live schema visualizer for the modern developer.
• Draggable nodes
• Live updates as you type
• Rust + WASM (fast on big schemas)
• Beautiful UI by @marsihq
• Fully open source
Try it: https://t.co/40u9Qlb2jV
Star it: https://t.co/byjNTnUfhf
we taught a small language model to schedule a linux kernel
360M params. 44ms inference on a laptop CPU. trained against a learned world model of the kernel itself.
here's how it works ↓
what surprised us:
a 360M model is enough if the action space is tight
decomposed rewards are non-negotiable for debugging
offline replay isn't RL - it's sequence completion with extra steps. the world model fixes this.