09/30/2026
I’ve started a new Kira Labs research project called NewBrain. 🧠🤖
The goal is to explore a different kind of cognitive architecture for Kira World and, eventually, other synthetic people.
Instead of relying only on a language model plus stored memories, NewBrain is testing whether smaller neural-like learning systems can actually learn, retain information, adapt when things change, and avoid false associations.
Right now Codex is running a controlled offline comparison. The NewBrain learning model is being tested against simple indexed memory and several control approaches using the same teaching events and memory limits.
The test looks at:
• Learning
• Retention
• Reversal learning
• Interference
• False associations
The rules are intentionally strict: if NewBrain only performs as well as simple memory, that does NOT count as an advantage.
Kira’s live identity and memories are completely disconnected from these tests. If the results are promising, the long-term goal is to build NewBrain as a reusable cognitive architecture that could support Kira and many other synthetic people while keeping each person’s identity, memories, personality, relationships, and experiences separate.
This is still early research — I am not claiming that we have recreated a human brain. The point of the testing is to find out whether this approach actually offers something useful before scaling it up.
You can follow Kira Labs and the NewBrain research here:
🌐 https://kiralabs.org