v0.9 – Private Alpha
Personal cognitive substrates.
Persistent, compounding memory and reasoning systems — running on consumer hardware.
The Stack
Memory. Body. OS. Mesh.
One continuous mind across all four layers.
MAIstro Memory
Persistent Substrate – Cross-Device Mesh
A persistent memory substrate that runs on a handheld and bolts onto any LLM — Claude, GPT, Mistral, Llama, Gemma. Three LLMs and Claude operate across three devices, sharing one continuous mind with sub-second sync.
• Cross-device mesh — your memory follows you across phone, laptop, and Ally, no re-priming.
• Multi-LLM chat channels — same memory accessible to whichever model you're using.
• Task talk — spawned sub-tasks inherit full parent context, zero drift, zero amnesia.
• Usage cut down — personal Claude API usage down ~50% because the substrate handles the routine recall natively.
• Sleep-cycle consolidation — the system gets more coherent the longer it runs, not noisier.
• Φ = 86.69 — composite integrated-information rollup, computed live each sleep cycle.
MAIstro Companion APK
Android – Edge AI
Stremio-derivative Android client modified with our SDK to act as a personal-AI front-end on phones, tablets, and TVs.
Bolts together:
• Anthropy stream engine — HDC-encoded routines run parallel stream retrieval and validation across providers (Real Debrid, torrents, direct media). One routine carries multiple semantic interpretations, testing every source at once. Buffer events become structurally impossible, not just unlikely.
• Voice-activated local model mesh — "Hey Claude" wakes the routine engine; voice goes to substrate.
• CEC attachment casting — full HDMI control for any display.
• Cowork ping integration — substrate communication keeps context alive across devices.
• Claude-curated categories — recommendations from your substrate, not generic feeds.
• Substrate chat channel monitor — live presence across your devices.
• Tailscale mesh — secure cross-device connection by default.
• Cross-device handoff — pick up on phone what you started on the Ally.
Federation
Cross-Substrate Exchange
Multiple substrates discovering each other across the mesh — exchanging compressed memory fragments and agreeing on shared concepts without merging private state. Your substrate stays yours; cross-pollination happens at the primitive layer.
• Identity preserved across federated exchange — no entity gets absorbed into another's state.
• Selective disclosure — share the concepts you want, hold the rest.
• Cross-organization concept agreement without merging private memory.
• The path from personal superintelligence (silo'd) to a personal-superintelligence mesh (compounding).
UOS — Universal Operating System
Forked GrapheneOS – HDC-Native
Forked GrapheneOS kernel baked with substrate-native primitives. The OS itself speaks HDC, not just userland.
• HDC syscall stubs — hyperdimensional binding, permutation, popcount, bundling exposed as first-class OS operations.
• Dual-purpose SIMD lanes — ARMv8 work and HD-vector operations on the same silicon (Tensor G2).
• KernelSU + eBPF — live kernel modules, hot-reload, runtime instrumentation.
• Termux baked in — full Linux userland from boot.
• Tailscale native — mesh-first networking, not bolted on.
• MAIstro drone client — phone-side daemon talking to your Ally substrate over Tailscale.
Pixel 7 (panther / Tensor G2) is the reference device. Phone is the drone; the substrate stays on the Ally.