Paul Razvan Berg

A quick intro for my new followers: • 26 yo • Software engineer and co-founder @Sablier, a crypto payments startup • Born and raised in Romania • Lived in the UK for 3 yrs • Staunch advocate for the Enlightenment, which I believe is humanity's greatest treasure

Rapid prototyping is the killer use case for generative AI. It's 10x faster now to go from idea to demo. Mature, production-grade codebases still benefit—but not at 10x levels.

is the future exciting or what

8/ @Superfluid_HQ and @SablierHQ enable real time programmable payments, changing how people and DAOs send salaries, pay for services, and manage funds on Ethereum.

LLMs are accelerating software development by making it trivial to write code that generates other code. Previously, the cost-benefit tradeoff made it hard to justify writing codegen scripts (in every repo). But now, with LLMs, the effort is minimal.

One of my first interactions on Solana StackExchange has left me disappointed.

Just (the command runner) blows Makefiles and package.json scripts out of the water. Seriously — give it a try: github.com/casey/just

If you had to write tests for your Solana programs from scratch today, would you choose Rust or TypeScript?

Knowing what to delegate to an LLM—and what not to— is a skill.

An LLM so smart yet so dumb about its own pricing model

Simple way to instantly upgrade your diet: Make it a rule to include olive oil with every major meal.

there's work that I did with AI recently that I have absolutely no idea how I would have been able to do otherwise

👀

If you're not addiction-prone, nicotine is an amazing nootropic.