A private intelligence system, exposed only where it should be.
Project Kaizen is built around a simple operating idea: powerful AI should be useful without requiring the private system to become public. The private side runs Max, profile-scoped memory, voice, operational phone-call alerts, image generation, SMS, reminders, automation, diagnostics, and model routing. The public side gets KaizenAI Open: a limited, queued, rate-limited chat lane with strict boundaries.
Design Goals
How Kaizen Is Organized
The platform is not a single chatbot. It is a set of cooperating services with clear boundaries.
Private Runtime
The private Max experience handles owner context, memory, long-running jobs, iOS continuity, voice, and authenticated model routing.
Public Runtime
KaizenAI Open is a deliberately smaller assistant with a public prompt, request queue, rate limits, abuse monitoring, and no personal memory.
Orchestrator
The dashboard manages service health, prompts, memory editing, public chat controls, diagnostics, and stack operations.
Communications
SMS, email, Pushover, operational voice calls, delivery receipts, reminders, and contact relay run through profile-aware services instead of ad hoc scripts.
Reliability Agents
Beacon workflows perform scheduled checks, local transcription, classification, notification, and audit logging for time-sensitive private use cases.
Proof Layer
Service status, release notes, dashboard screenshots, delivery logs, and audit reports are the public-safe receipts for what the platform actually does.
Shawn's Journey
The project grew out of decades of owning the full stack instead of renting pieces of it.
From IRC to infrastructure
INTRAC.NET began in 1998, when Shawn McCalla was running internet services, IRC networks, dedicated servers, and security-focused infrastructure at a time when uptime and hardening were learned in real production conditions.
That work shaped the pattern that still drives Kaizen: understand the layers, own the stack, keep systems observable, and build tools that can be trusted when they are used every day.
From homelab to AI
The same mindset later moved through home automation, networking, secure tunnels, private dashboards, and local services. When local Apple Silicon and Apple Metal made private AI practical, Project Kaizen became the next step.
The goal is not novelty. It is a personal intelligence system that feels useful, private, inspectable, and controlled: Max for the protected owner workflow, and KaizenAI Open as a carefully bounded public lane.
Evolution Roadmap
A clean public view of how Kaizen has evolved and where it is going.
INTRAC.NET foundation
Cybersecurity, internet services, server operations, IRC networks, and the early discipline of owning production systems end to end.
Private infrastructure layer
Networking, DNS, tunnels, dashboards, storage, monitoring, and home automation became one integrated operating environment.
Local AI runtime
Private assistant workflows, current search, voice, memory, model routing, diagnostics, and service orchestration moved into one local-first stack.
Native iOS experience
The private assistant became portable through a native app with chat, advanced voice, runtime status messaging, background continuity, and diagnostics.
Professional tooling and a stack that watches itself
KaizenAI Open is live and the dashboard manages the stack. July 2026 added Communications Hub, Max SMS, operational call alerts, reminders, delivery receipts, and Beacon watchdogs early in the month — then live real-time voice conversations with Max, local-first search routing, and a health lane the assistant reads to report on its own platform.
More polish, safer access, deeper automation
Planned work focuses on clearer public receipts, stronger benchmark pages, sanitized code excerpts, better public queue visibility, and carefully gated multi-user access.
July 2026 Platform Additions
The newest work is less about prose and more about operational surfaces that can be inspected.
Communications Hub
Telnyx SMS inbox, delivery receipts, STOP/START/HELP handling, allowlisted contacts, and an append-only SMS audit log.
Max SMS
A hidden short-form Max lane for SMS conversations, routed with the sender's canonical user profile and memory boundary.
Reminders
Natural-language reminders for known Kaizen users, delivered through SMS, email, Pushover, and future call paths without arbitrary recipient expansion.
Voice Calls
Operational call alerts exist today, and the real-time path is live: approved users call the Kaizen number and talk to Max with local speech processing and an operator kill switch.
Local-First Search
Search routing moved local-first: a self-hosted meta-search engine answers most queries, with paid search APIs reserved for the moments they add real value.
Self-Aware Health
A standing health collector and live probe lane let the assistant answer questions about its own stack, backed by a dashboard health panel.
Beacon Watchdogs
Private scheduled-check agents now verify delivery receipts and escalate when a critical notification is accepted locally but not delivered downstream.
Release Naming
Public terminology is product-facing, not model-facing.
Why It Exists
Project Kaizen is about ownership, observability, and safer AI operations.
Private by default
Personal memory, account context, device control, home automation, and operational diagnostics remain on the protected side of the system.
Public by design
Visitors can try a controlled public assistant without gaining access to the private assistant, model inventory, prompts, runtime internals, or home systems.
