Moritz Groß

Algorithms, AI, Open-Source 🦀🐍

[Moritz Groß @Moritz__Gross]

@alexandr_wang SaaS - Slop as a Service

[Moritz Groß @Moritz__Gross]

@Miles_Brundage the link to his AI alignment blog at https://t.co/3Pa5maNkCZ being broken is also hilarious.
the Romans would see this as an omen.

[Moritz Groß @Moritz__Gross]

@michael_chomsky makes me wanna write blog posts

[Moritz Groß @Moritz__Gross]

@1a1n1d1y from my experience, using Codex for Rust is particularly juicy

[Moritz Groß @Moritz__Gross]

@jxnlco adversarial robustness

[Moritz Groß @Moritz__Gross]

@bryan_johnson only took the American a few years to touch grass instead of microplastics

[Moritz Groß @Moritz__Gross]

@fchollet onwards to the next 6 months

[Moritz Groß @Moritz__Gross]

@scaling01 afaik, the important difference is that Work gets billed to the same token usage as Codex, and Chat does not

[Lazarz @Laz4rz]

Can’t believe I ever had to write pandas data manipulations by hand and it was actually taking most of my workday

The second half was taken by matplotlib

[Moritz Groß @Moritz__Gross]

@theo I didn't want this to be used as a shot against Bob, I was just thinking about him with a smirky look in his bathrobe and the hot chocolate writing this.

[Moritz Groß @Moritz__Gross]

@daytonmills

[Moritz Groß @Moritz__Gross]

@paularambles what's going on in biology that basically all roles I've seen are payed so shitty

[Moritz Groß @Moritz__Gross]

@NeelNanda5 wild that the optimistic long term scenario is still set so relatively close to now

[Moritz Groß @Moritz__Gross]

@finkd Thank you Mark, very cool

[Moritz Groß @Moritz__Gross]

@signulll @ywjjza I thought this post was a fake at first lmao

[Moritz Groß @Moritz__Gross]

@jxnlco Did you add the typo to make the post look human-written? ;)

[Olivia Moore @omooretweets]

We have achieved AGI as measured by the “average corporate Zoom meeting” benchmark

[Moritz Groß @Moritz__Gross]

Imagine seeing this pre-COVID

[DoorDash AI Research @AIatDoorDash]

Models out of the box have trouble reviewing DoorDash’s codebase, even when feeding them appropriate context. Single-pass AI reviewers only caught 30% of the real issues in our PRs.

Instead, we built a multi-model (Sonnet 4.6 + Opus 4.8) code review agent that caught 53.6% issues at $3.91 a PR (most enterprise code review solutions perform significantly below 50% recall on DoorDash PRs).

Today, we’re proud to share DashBench, our internal benchmark to compare multi-model/harness code review systems against each other. DashBench has given us confidence in maximizing the intelligence per dollar we’re spending on AI tokens to ship code safely.

[Moritz Groß @Moritz__Gross]

@growing_daniel it sounds like something Kanye would say

[Moritz Groß @Moritz__Gross]

@signulll sounds like a Kanye quote