OpenAI's latest computer use model has been showing exceptional results lately.
Following our initial playground launch, we've discovered some remarkable automations.
I've compiled several examples demonstrating real-world applications below. 👇
Playing with Model Context Protocol from @AnthropicAI and it's pure automation magic
I've began using it for leading generation and data enrichment processes
> Launches multiple browser sessions at once with @browserbase
> Goes to websites, collects information, and formats a
New evals for the new Google and OpenAI models are out! (o3 soon) 🔥📐
Gemini 2.0 Flash, GPT-4.1 mini, Claude 3.5 Sonnet match on accuracy; the prices don't (76¢ vs $2 vs $23)
Gemini 2.5 Flash shows no significant improvement over 2.0
4.1-mini beats 4o and 4.1, but costs far
Ok, this is absolutely insane
@AnthropicAI Claude’s computer use works extremely well with @browserbase
Now, you have session recordings, logs, anti-bot, captcha solving, and more
It can also correct it’s own mistakes
One step closer to having agents interact with your
I’ve wrapped up an incredible chapter at Browserbase.
I came to SF knowing just a handful of people and had an opportunity of a lifetime. Grateful for Paul and the team for seeing potential in some random kid from DC.
I've decided to step back to focus on personal projects,