Not bad at all ! Could you validate if you know geez?

LLMs turned out to be not bad with geez Last week I was testing Gemini with kine like kibr yieti and it's translation and getting the gold from the kine and it handled it quite well even better than I expected plus it suggested to create mewdis out of the first kine and seemed to know a lot and i hv to say I'm impressed.

A thought I can't shake.

The most honest dataset about us might be our saved folder.

The hardest part of building a personal AI agent, call it OpenClaw, Hermes, Claude co-work, or anything else, is describing ourselves to it.
When it asks, "Tell me about yourself" or "What goals do you want to achieve?", we give it the version of ourselves that we can explain.

And when we describe ourselves, we perform. We simplify. We pick the version we understand, or the version we want others to see.

But our saved Reels, TikToks, "watch later" lists, and the quotes we screenshot. that's where our taste lives.
Our ambition. Our insecurity. Our future self.

So in a way... our saved folder might know us better than we do.

Because every save is a tiny confession.

Every time we bookmark something, something in us says:
"I don't fully know why, but this matters to me."

Just tried it, and tbh, this is where my assistant starts to feel real on a different level; when it starts to understand what keeps pulling us back.

@b1n1yamBuilds

Never had to run two emulators at the same time until today

My CPU was screaming for help.

EXPOSING THE ENTIRE TECH COMMUNITY

Coming soon.

I saw this on code night someone was claiming it would give a free Claude API and it actually worked.

Verify with telegram and it not gonna cause u anything it just send /start command to bot that's all .