thisisthegoodlife@juliegood.comSan Antonio, Texas

Build · AI & Automation · San Antonio

AI, in production.

Plenty of people have opinions about AI. I have a production system. The IMC Machine is the media operations platform I designed and built at Good Creative Media, and the automation practice around it runs every working day: publishing, transcribing, indexing, generating, verifying. This page is the machine room.

Everything below is work I built and operate myself

Products on one platform

Working sectors: legal and live events

AI models in consensus on court-bound documents

30+

Distribution channels in The Creative Studio

The Flagship

The IMC Machine.

The flagship product of Good Creative Media: an AI-driven media operations platform with one engine, one canonical record, and two working sectors. The cultural side runs venues, theaters, galleries, festivals, and touring artists. The legal side runs evidence preparation and motion practice for criminal defense offices working through multi-format discovery. A concert and a captured incident are the same shape: a live event, recorded from several angles, that has to become an organized record. The platform reads both, end to end.

The Headline Capability

The AI reads the MP4.

Most platforms treat a video file as opaque. The IMC Machine treats it as readable. Google’s Gemini transcribes the audio, captions what is on screen, and indexes the file by what is actually in it: body cameras, dash cameras, interview audio, CCTV. Silent files get descriptive captions, because silence is part of the record too.

The Discipline

Three AIs, one signature.

Before any court-bound document is treated as final, three models read it: OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini. All three have to agree. Disagreement comes back to the attorney as a question, never as a silent override. The human signs; the machines check each other first.

Vite React Supabase WordPress by API

A small, disciplined stack, chosen so a one-person studio can build it, ship it, and maintain it.

Visit imcmachine.com How I teach with it

The Four Products

Two for the stage. Two for the courtroom.

Four products on one spine. Each one ships real artifacts: campaigns, run-of-show books, evidence packets, motions. Nothing publishes or files without a human approval.

Product 01

The Creative Studio.

One approved event, release, or record in. A full campaign out: press release, press page, social set, formatted images, and calendar copy in English and Spanish. It drafts; you approve. It never posts on its own.

Live events · Arts & culture

Product 02

The Live Event Production Hub.

The operational backbone for a show: run of show, crew portal, safety, day-of-show operations, and the wrap, with venue and artist profiles you build once and reuse all season.

Live events · Venues & producers

Product 03

The Evidence Orchestrator.

Court-ready evidence preparation. AI reads body camera, dash camera, interview audio, and CCTV footage, then returns indexed, captioned packets an attorney can stand behind in a courtroom.

Legal · Criminal defense

Product 04

The Dismissal Engine.

Motion practice for criminal defense: theory mapping, a canonical motion format, Article 39.14 tooling, and the three-AI consensus check before the attorney signs anything.

Legal · Motion practice

The Voice Layer

Buddy Chat.

A website that can only be read leaves people out. Buddy Chat is the concierge layer of The IMC Machine: text chat for the people who type, real-time voice for the people who talk, built on OpenAI’s models and grounded in each organization’s own published content.

It is deployed on client sites today, including Forrest Good PLLC, where it answers questions about the practice in plain language and hands anything consequential to a human. That handoff is the design, not a limitation. An AI concierge that makes promises is a liability; one that listens well and routes correctly is a front desk.

Each deployment is scoped to its own organization. The agent on one site cannot answer for another, and the answers stay inside the content the organization has approved. I build the widget, wire the models, write the guardrails, and maintain every tenant myself.

TextStreamed chat, grounded in the site’s own published content
VoiceReal-time spoken conversation, not a phone tree and not a recording
ScopeTenant-scoped: one organization per deployment, no crossover
ControlEscalates to a human; it answers questions, it does not make promises

The Daily Practice

The automation that runs the studio.

WordPress automation.

Client sites are published, audited, and repaired through the WordPress REST API. Pages, metadata, and full-site audits run as code, with a human review before anything goes live.

Evidence and media pipelines.

Transcription, captioning, and indexing pipelines that turn raw footage and audio into searchable records, processed as one consistent, controlled run per matter.

AI image generation.

Production image workflows on OpenAI’s image models, sized and shipped for web, social, and print. One rule holds across all of it: nothing fabricated is presented as real.

The Classroom

How I teach with this.

I have taught in the Alamo Colleges District for more than twelve years, and the method has never changed: students learn tools by shipping real work with them. AI did not change that method. It raised the stakes for keeping it.

Method

Project-based, always.

Students leave my courses with published artifacts: a campaign, a site, a record that exists in public. The IMC Machine runs on the same logic, which is what makes it an honest teaching instrument rather than a demo.

Currency

Current tools, real stakes.

We work with the platforms and models running in my studio this semester, not the ones in a three-year-old textbook. When the tools move, the course moves, because I am using them every day and I notice.

Judgment

Verification is the skill.

My own platform requires three AI models to agree before a court-bound document counts. That is the habit I teach: AI output is a draft until a person, or a better process, has checked it. Prompting is easy. Judgment is the curriculum.

Built by Hand

I built the entire IMC Machine app myself.

Not a team. Not an agency. One person, on purpose. I designed the architecture, wrote the code, shipped the products, and I maintain every line. The way I did it is the work: I built the whole platform using three AIs in concert, OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini, treating each one as a collaborator I direct, correct, and verify rather than a tool I trust on faith. That is the practitioner skill I bring into the room, and into the classroom: not prompting, but conducting a system to a finished, shippable result.

OpenAI’s ChatGPT

Drafting, reasoning, and the image and copy generation that fills The Creative Studio.

Anthropic’s Claude

Code, system design, and long-document review across the build and the legal side.

Google’s Gemini

Reading the MP4: transcription, on-screen captioning, and indexing of recorded evidence.

Inside Two Products

What the live-events side actually does.

Two products carry the cultural side of the platform. Here is what is under the hood, in plain terms.

The Creative Studio

AI prompting that makes the campaign.

Approve one event, and a directed chain of AI prompting produces the assets a launch needs, ready for a human to review before anything ships.

Ad copy drafted for the release, the page, and the social set
Graphics generated and formatted for web, social, and print
MP4 videos generated from the same approved brief

The Live Event Production Hub

AI-directed capture, then dual distribution.

The operational backbone for a show, with AI directing the cameras and handling the audio so a small crew can run a big night.

AI-video-directed capture across the cameras during the show
Podcast distribution of the recorded audio to the feeds
An auto-equalizer that balances the sound without a dedicated engineer
Dual distribution: live streaming the event and building a media library for the event’s clients

Who And What

One machine, two columns of work.

The platform was built for a specific set of operators and the events they run. Left is who it serves. Right is what it captures and organizes.

User types

Venues and presenting houses
Festivals and multi-day events
Theaters and seated venues
Art shows and galleries
Bands and musicians on tour or in residency
Criminal defense attorneys and their offices
Private investigators and investigative agencies

Event types

Live concerts and music nights
Theater runs and staged performances
Art openings and exhibitions
Conferences and presented programs
Recorded depositions and interview audio
Evidence capture: body camera, dash camera, and CCTV footage

The Licensed Edition

The defense version lives at evidefense.com.

The same software the cultural side runs on is licensed to criminal defense attorneys as its own product, built for their work and their standards. One engine, two front doors. The legal edition is at evidefense.com.

Visit evidefense.com

evidefense.com

The Stack