INTRAC.NET · Project Kaizen

Own the intelligence. Control the entire system.

Project Kaizen is not another AI wrapper or rented chatbot. It is a private, local-first intelligence platform built from the runtime up, with custom orchestration, model routing, persistent memory, voice, agents, automation, search, tools, and infrastructure operating as one system.

Your hardware. Your models. Your data. Your rules.

Public Lane

KaizenAI Open

A constrained public assistant backed by the Kaizen stack, separated from private memory, tools, contacts, automations, and operator controls.

What's the weather looking like in Houston this weekend?
Hot and humid — highs near 96 with afternoon storms possible Saturday. That's from the National Weather Service, pulled a few minutes ago.
What are you?
I'm KaizenAI Open, the public lane of Project Kaizen. The private assistant, memory, and tools stay behind the boundary.
Fair-use queue Daily limits Public-safe prompt
30+
Services in the stack
Open
Public chat lane
Max
Private assistant
iOS
Native KaizenAI app
Public voice & SMS lane

Max answers a real phone line.

Call it or text it and you are talking to a locally run AI in real time — no app, no signup, no account. It is a bounded public demo: no stored memory, a short daily limit, and STOP honored instantly.

Call +1 (504) 605-1092 Send a text Msg & data rates may apply. Reply STOP to opt out, HELP for help.

What Kaizen does

One private stack, many lanes. The public lane is deliberately narrow; everything personal, operational, and automated stays behind the boundary.

Public AI lane

KaizenAI Open is a live public chat backed by the same runtime as the private stack — with queueing, daily limits, current web context when needed, and private tools blocked.

LivePublic-safe

Talk to Max on the phone

Call or text Max on a real Kaizen number and hold a spoken conversation. Speech is processed locally. Approved users get the full assistant; anyone else reaches a bounded public demo with no stored memory.

LivePublic demoApproved users

Native iOS app

KaizenAI lives in a native iOS app — chat, live voice conversation, transcription, and image workflows, wired to the private stack from anywhere.

PrivateiOS native

Communications Hub

Max handles SMS with approved users — delivery receipts, STOP/START/HELP, reminders, and message history — plus voice-call alerts when something needs a human.

PrivateSMS + voice alerts

Memory, search, and creation

Private semantic memory with editing controls and per-user separation, live web search, and image generation with edits, upscaling, and replayable jobs.

PrivateOperator-governed

Self-aware operations

A protected dashboard runs the stack — health, logs, users, prompts, model routing — while Beacon reliability agents watch scheduled sources and keep an audit trail. Ask Max "how is the stack?" and it answers from live probes.

PrivateAudited

Public boundary

The public endpoint is deliberately smaller than the private Kaizen system.

What is enabled

Model identityKaizenAI Open
QueueingEnabled
Daily limitsEnabled

What is blocked

Private memoryBlocked
SMS and callsBlocked
Home controlsBlocked

Why it exists

KaizenAI Open gives visitors a small, safe way to experience the platform while the private assistant, dashboard, memory, and automation layers remain protected.

Public request logs are metadata-only for health, abuse monitoring, and capacity planning.

From the lab

Engineering write-ups from the same stack that serves this page.

The origin story

I Built a Private AI Platform in My Closet — how a Mac Studio in a hallway closet became a full AI operating layer.

Build log

The architecture

Agent Citadel — the local-first agent ecosystem, voice pipeline, and home automation that Kaizen grew out of.

Deep dive

The stubborn fix

The Internet Said It Was Unfixable. We Fixed It Anyway. — chasing an ARM64 transcoding bug everyone else had given up on.

War story

Try it yourself

KaizenAI Open is the public lane of this stack — no account, no app, no tracking profile. Just a rate-limited chat backed by the same runtime Max runs on. Prefer to talk? The same assistant answers a real phone line.