🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

In many details, Rust is even more respecting of maximum performance goals. A Box<dyn T> is two pointers, so you save on an indirection when dispatching calls dynamically.

Deterministic Simulation Testing (DST) is a powerful verification technique and now with 10 months into doing it, I think I have better answers to how it fits the overall verification process. 1/

So if I tell an LLM model which of two steps worked so that it keeps that in its context, does that count as RLHF?

Sarah Catanzaro Firebolt TU München Fun Fact: During a CIDR 2024 talk, DuckDB announced they just reached #1 on ClickBench. Thomas was sitting in the audience with a PR containing new Umbra results that put him back at #1. He waited until the talk was over and then immediately pushed the PR. Cutthroat AF.

This is one anti-Rust take I can get behind Build enough tiny crates like this, force transitive dependencies and you end up with either a leftpad style fiasco or a supply chain attack

the absolute state of my life right now until an llm raises a pr like this, it's not a real engineer

Quitting programming as a career right now because of LLMs would be like quitting carpentry as a career thanks to the invention of the table saw.

Alright this has gone too damn far

the perennial tradeoff between safety and performance

Databae, don't shard my heart.

Zaid Humayun v Debasish (দেবাশিস্) Ghosh 🇮🇳 I think it’s even simpler. PBT is the formalism, DST is a specific execution model with some properties that make it very useful. That’s why I don’t like to call the work I do on the Turso simulator DST, because I’m specifically working on the PBT part, not fault injection.

Cats 90% of the time