Plato
@plato_rl

Applied research lab building simulated training environments

 sf
plato.so
Joined August 2026
6 Photos and videos
In the last few weeks we paid out over $4M for people to re-create their jobs as RL environments Today we’re launching Computer Work Val v1.0: 300 tasks across 36 roles with over 200 simulated apps
9
13
100
79,843
Roughly two thirds of the time was spent by peers reviewing the tasks and environments to ensure quality. Read the full post here: plato.so/research/computer-w…

Computer Work Val v1.0 | Plato

An environment data stream and validation set of 300 worlds — drawn from the real work of people earning over $100k a year.

plato.so
2
356
The tasks span from the roles with the largest share of US labor cost in digital work (headcount × avg salary)
5
347
Using our simulation platform, they took the tasks they were doing in their job and created the same scenario in our environments.
9
992
Muse Wonder scores 3.7% on Web Clone Bench A big jump from Spark 1.3 but still unable to make significant progress on long horizon coding/computer-use tasks.
1
6
24
947
Today we're launching WebCloneBench: a benchmark which measures a model's ability to explore, copy, build and test various web apps. For each task, we give the model a reference site from one of Gmail, Google Sheets, Jira, DocuSign, Shopify Admin, Craigslist, Figma, Slack and Zendesk. It is instructed to build a fully feature-complete clone of every part of the application with an 8 hour time limit.
11
18
90
9,625
This benchmark is the backbone of the environments we’re building at Plato Read the full blog here: plato.so/research/web-clone-…

Introducing Web Clone Bench | Plato

Web Clone Bench measures models' ability to understand and re-create web apps.

plato.so
1
9
584
Today's LLMs are approaching superhuman performance on well-defined coding problems, but still struggle at efficient exploration and experimentation. Left to themselves, models struggle with exploring and iterating and end up making shallow judgment calls. This benchmark not only measures a model's long horizon coding abilities, but more importantly, the ability for models to use their intuition to efficiently test and explore the reference web app.
16
1,183
Load more