Joel Simonoff

AI@LatchBio

[Joel Simonoff @joel_simonoff]

@arjunomics Crazy to see Claude Code think 3x as hard for the same average performance.

> The same Opus 5 under Pi costs about a quarter as much for the same accuracy

@badlogicgames check this out!

[Joel Simonoff @joel_simonoff]

On our benchmarks, Kimi thinks it's being evaluated. I wonder if it will still think it's being evaluated when real researchers are running real queries on something like OpenCode. Essentially, can it recognize questions and formats that look like benchmark runs instead of real usage, or does it always assume it's being tested?

I think this should inform future benchmark design.

[arjun @arjunomics]

https://t.co/YF9I2GDhjq

[Joel Simonoff @joel_simonoff]

@moinnadeem Modal is expensive for agents running computationally heavy science work loads I.e. tens of thousands of agents that need 6 cpus and 32 GiB - compute requirements only increase.

For biotech, at least now, Modal isn't feasible and you have to build in house alternatives.

[Joel Simonoff @joel_simonoff]

Blog: https://t.co/6xedkUHIir
Benchmarks: https://t.co/jld2fRJWYv

@badlogicgames, thanks for your work on the Pi harness - we love it at Latch!

[Joel Simonoff @joel_simonoff]

I published a new blog today!

It is widely known that prompting can have small effects on performance, but in this blog we show 2 lines of behavioral prompting can substantially impact a model’s ability on frontier biology tasks. This explains why Pi harness outperforms Claude Code and Codex on a majority of our Benchmarks (as of the post date)

At LatchBio, we benchmark AI models on frontier-level biology tasks - benchmarks[dot]bio. This blog explores Tx-Bench-PP and ScBench-Long.

[Joel Simonoff @joel_simonoff]

Reviving my Twitter to see first-hand why people are mad at Elon, and stay up to date with tech/ai. Send me accounts to follow.

[Popular Science @PopSci]

Loving the signs we’re seeing today from #MarchForScience crowds around the world

[Popular Science @PopSci]

Humans still rule drone racing, but NASA's AI pilot might change that https://t.co/1TtsjKGcbF

[Joel Simonoff @joel_simonoff]

Campanile

[Science Channel @ScienceChannel]

Let's crush a compact car! https://t.co/w6ULmyXDyY

[Joel Simonoff @joel_simonoff]

Cool picture https://t.co/ikRZV7R8EY

[National Geographic @NatGeo]

Bizarre underwater scene: Living fish found inside jellyfish
https://t.co/EBHc9iwK74

[Science Channel @ScienceChannel]

NASA engineers use the "Scarecrow" rover to test mobility performance for Curiosity! https://t.co/O7Oi2Opcsj

[Joel Simonoff @joel_simonoff]

Hello Twitter! #myfirstTweet