Meet Kricko β Your Everyday Offline AI Assistant.
What if you could use AI without depending on the internet?
With Kricko, experience AI directly on your device.
- No internet required for supported features..
- Chat with documents.
- Ask questions, write and learn.
Kricko v1.2.0 is here!
π Introducing Chat with Documents on Android.
Upload your documents, ask questions, and explore your files with AI.
One more step toward making AI more useful, right on your device.
#Kricko#Android#AI
Kricko v1.2.0 is here!
Meet Chat with Documents π
Upload your documents and chat with them using AI. Get insights from your files and make information easier to explore.
We're building Kricko to make AI more accessible, one update at a time.
The journey has just begun.
What if your AI worked without the internet? π
We're bringing offline LLMs to mobile with Kricko.
Our goal: 1,000 users exploring on-device AI.
No cloud dependency. Just AI on your device.
Who's in?
#Kricko#OfflineAI#LLM
@yevloop Fair question! Right now you can download a model and run Kricko fully offline: no API costs, private, low energy. Next version brings small task-specific models, each built to nail one job. Which task would you want first? π°
Just shipped Kricko π°
An open source AI companion for coding, creating, and getting things done.
Desktop. Android. CLI. Python. npm.
pip install kricko
npm i @cortiqa/kricko
What should we add next? π
#AI#OpenSource
Built Kricko, an open source AI companion that runs everywhere:
desktop, Android, terminal.
Try it in 10 seconds:
pip install kricko
Feedback welcome. Brutal honesty preferred. π°
#BuildInPublic#AI
4/4 π¦ Open Frontier Research:
Strictly for research & educational exploration. Weights, standalone PyTorch code & configs are open on Hugging Face:
π https://t.co/kA2Ot4NN91
#AI#DeepLearning#LLMs#OpenSource#Cortiqa
1/4 π Introducing ReverX-1-100M: An open-weights 100.45M Causal LM exploring Reversible Transformer architectures (RevNet).
Trained from scratch on a single 16GB Tesla T4 GPU (130,391 steps) with zero activation caching.
A research breakdown by @CortiqaAI π
AI works for humans, yet understands human strategies and games better than most people. In analysis, speed and pattern recognition, it is already smarter. Use it as a powerful tool under clear human control.
Why we built Falin-300M: We didn't just want to plug into someone else's API and call it AI. We needed to prove Cortiqa can actually build its own LLM β from scratch, no shortcuts. Falin-300M is that proof. Still early, still rough β but it's ours, end to end. In the future
@yevloop@AlokMehtaa Solo founder β built everything myself, from architecture to training. Data mix: ~40M custom tokens, mostly general conversational data + a dedicated identity dataset so it self-identifies as Falin. Still early-stage β refining the mix before scaling to the 1B model.
The platform's ready to launch. Keeping the servers, compute and API costs running is on me right now. If you've followed the journey and want to help, scan the QR and support β every bit helps.
No pressure, just an honest ask. π