Summaries like this, in your inbox every morning.
Sign up free →What happened
Lethe is a cognitive runtime built in Rust that runs multiple specialized brain regions—cortex (voice and delegation), hippocampus (memory), default-mode network (background thinking), brainstem (system management), and subagents (disposable workers)—each on its own clock. The system can run locally (~50 MB, one binary) or hosted; users sign in and start talking, free for two weeks (card required to continue). It integrates with Anthropic, OpenAI, OpenRouter, or local Gemma models without code changes.
Why it matters
Most AI assistants only respond when prompted; Lethe runs background cognition while you're away, notices tasks you've forgotten to mention, and decides on its own when to interrupt you. It remembers across model swaps and reboots, so switching to a new underlying AI doesn't erase what it knows about you. For busy professionals juggling multiple projects, this means reduced friction around context management and deadline tracking.
What to watch
The hosted version (free for two weeks) is live today; local setup requires a Telegram Bot Token, an LLM API key or subscription, and your Telegram User ID. For local GPU inference, the example uses Gemma 4 31B with 98,304 token context and requires ~48GB+ VRAM. The system is open source and runs as a systemd service on macOS and Linux.
No discussion yet for this article
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
5 minutes a day. The AI essentials.
200+ sources · Email / LINE / Slack
