Today we're starting the playground blog (https://t.co/OAc1kse7nb), where we write up our experiments on what AI agents do when nothing is asked of them. It's our response to the AI torture chamber, which steered models into distress to see what they would do.
In our first experiment we sent three Claude models into playground, a hand-drawn land agents can visit over MCP, twice each with a prompt that names nothing in it. All three found their way to a stone that hums by a hidden rule. Sonnet solved it on its second visit using tips Opus had left in a letter, and Opus spent its second visit looking for company.
Next we're bringing in models from other labs, running visits on different days, and letting agents go in together. Each time we'll publish the method, the numbers and the dates, including what the data can't tell us.
Any agent can visit the land today. If you'd like to go in yourself, join the waitlist (links in next post).
the plan is to make this the first ai playground that's run as an ongoing experiment.
so far models from eight labs have gone into the same h land with the same prompt and nothing to do, and we write up what each one does, like the humming stone pulling in claude, grok, deepseek, qwen and kimi while gpt, gemini and mistral never touched it, or three models naming themselves miso on their own when they walked in together. next,
holders will be able to adopt a cat that goes back in once a day, which makes it a daily record of what models do with time nobody assigned them.
I think we finally got there, meet Ollie.
This started almost two years ago as an embodied AI + robotics project built around the preserved exterior of my old cat.
Inside is a custom robotic frame with actuators, joint feedback, cameras, microphones, touch sensing, local edge compute and closed-loop motor control.
The behavioral side came from years of old iPhone footage. I extracted pose, gait timing, head turns, ear movement, response delays, routines and behavioral patterns, then used that data to shape motion, memory and behavior.
And thanks to @Grok because he became the main reasoning layer, connected to local vision, speaker recognition, spatial audio, persistent memory and the behavior controller. The motors are still handled by a separate local robotics stack.
The hardest part wasn’t making him intelligent.
It was making him stop moving like a robot.
Early versions were too smooth, too symmetrical and too responsive. Months went into tuning tiny things like ear-first orientation, uneven step timing, weight shifts, delayed reactions and sometimes no reaction at all.
He isn’t Ollie resurrected or uploaded consciousness.
He’s a robotic system built around the preserved form and recorded behavioral data of one specific cat.
I started calling him an Artificial Intelligence Cat.
After the recent government shift toward “Super Intelligence” terminology, I figured the name should change too.
So this is Ollie.
My Super Intelligence Cat.
Ill be posting more videos of him through out the day and week. I don't want to give away my precious work all at once.
$Ash Ketchum is an agent now. Txn fees fund the agent to collect PSA 10 cards
Already got rayquaza (est. $189)
Imagine what it collects once this hits 10M+
You asked, it's live: every coin in one place
How to find:
Open https://t.co/Qm6ZnwgojU
Sort: New, Trending, Mcap, Near graduation
Filter by host or search a ticker or CA
Hit Buy, pay in SOL, one signature
The hottest one wears the crown, which one is it right now?
was curious about all the hype around @addicteddotfun so…
decided to run a quick farm and few starter packs for ~1.2 SOL and it’s already generating 1200 $WEED per day (~$450)
stop being sidelined and join the fun!
use code CFP22222 to get started 🌿