Handshake

This account is verified. Learn more

@joinHandshake

Building the future workforce of the AI economy 🤝

Handshake AI now has its own home on X @joinHandshakeAI

United States joinhandshake.com Joined March 2014

364 Following 12K Followers 3.2K Posts

Latest posts 5

Pinned post

Handshake @joinHandshake 2d

What do frontier labs actually buy?
@latentspacepod went inside Handshake AI for their first deep dive with an AI data company.

Full episode 👇

Garrett Lord Verified account @GarrettLord 2d

What does it take to push frontier models forward?

@latentspacepod's first deep dive with an AI data company. Our team walked through our own datasets, life sciences, finance and safety, step by step.

Worth your time if you care about progress and alignment of AI.
View quoted post on X

Handshake @joinHandshake 20h

The answer is the 1 µg column. The model said 5 µg. It’s easy to see for a human, even one who’s not an expert.

The model didn't count the faint band at 20 kDa, which any scientist would call a band.

Nick walks through it 👇

3:12

Handshake @joinHandshake 22h

"AI is not going to cure cancer before it can interpret these very basic types of images."

@jomulr spent 15 years in AI research, a lot of it in computational biology. On VIALS, Handshake’s Life Sciences benchmark, even the best frontier models score 50% or worse on tasks a lab tech routinely does in under a minute.

2:18

Handshake @joinHandshake 23h

@ArtificialAnlys now runs our VIALS life sciences benchmark. Fable 5.2 launched yesterday and it takes the crown.

VIALS is 161 real-world lab imaging tasks, like the “western blots” scientists read daily.

Artificial Analysis Verified account @ArtificialAnlys 2d

Claude Fable 5.2 leads scientific visual reasoning benchmark VIALS-AA; it makes more tool calls than any other Anthropic model, unlike OpenAI models which make fewer as they get smarter

VIALS-AA is our implementation of @joinHandshakeAI’s VIALS benchmark for visual reasoning in life sciences research, across artifacts such as gel blots, plasmid maps, and phylogenies. Claude Fable 5.2 (max with fallback) scores 67% on VIALS-AA, ahead of Gemini 4 Carbon (high) at 64% and GPT-6 Astra (max) at 63%. Fable averages ~37 tool calls per task, while Astra averages ~8, the fewest of any model we’ve tested.

On turn 4 of this protein interaction task, Fable creates crops and asks to view them. On turn 10, it builds a color mask by thresholding color channels. On turn 17, it fits a circle to each arc and unrolls the arc into a straight strip. On turn 31, it counts the bars as peaks in an intensity profile. And on turn 41, it maps every detected spoke back on the original image to check its work. Fable batches calls on 28 of its 51 turns, averaging 1.75 tool calls per turn, compared with 1.12 for Sonnet 5.5.

View quoted post on X

Handshake @joinHandshake 1d

In which column below do you first see a band appear in all three of the pink highlighted lanes?

This is the type of routine task a life scientist performs in the lab, and frontier models still get it wrong all the time.

Answer at noon…