Michael Grippo.

AI product and workflow strategist  /  Long Island, New York

AI writes the code.I decide what ships.

I'm Michael Grippo. I direct AI-assisted software development from the first requirement to the installer build. I define the product, set the rules the models cannot break, verify what comes back and make the release call.

On authorship: AI coding agents wrote most of the code in these projects. The requirements, the rules, the verification and the release decisions are mine.

Gate 01 / human approval Michael Grippo in a burgundy velvet blazer over a black turtleneck, hands on his lapels, looking at the camera

Release gate4 / 4

  • Change stays inside its scope
  • Source fingerprints match
  • Tests reported, pass or fail
  • Approved by a human
Claude CodeChatGPTGrokGeminiPythonPySide6PyMuPDFSQLiteElectronReactTypeScriptWindows CIREAPER

The record so far

2Windows desktop applications taken from requirements to installer buildsTwo technology stacks
13Installer builds of Takeoff Artist, through version 1.4.8Local repository audit
43 / 50Pull requests merged on BandBlaster, with Windows CI passingGitHub audit
32ChatGPT-to-Claude UI rounds, with source verification and Windows test checkpointsWorkflow record
4AI systems coordinated: Claude Code, ChatGPT, Grok and GeminiHandoff procedures

Figures come from project audits dated November 8, 2026. Both applications are pre-release.

Selected work / 04

Software that knows what it is not allowed to do.

Independent projects, directed by me and built with specialist AI coding agents under explicit requirements, scoped changes, source verification and human approval.

01

Takeoff Artist

Construction estimating for Windows

  • Python
  • PySide6
  • PyMuPDF
  • SQLite
Pre-release / v1.4.8

Estimating software that refuses to guess.

A Windows desktop application for construction takeoff, pricing and proposals. I directed its development with AI coding agents through 13 versioned local installers, and wrote the rules the AI is never allowed to break. In software that puts dollar figures in front of a client, a confident wrong number is worse than no number.

  • NeverThe AI invents a measurement, a scale or a fee.
  • NeverA missing price reads as $0. It stays blank.
  • BlockedA proposal publishes without an explicit, current pricing source.
Line itemQtyUnit priceLine total
Interior doors14$385.00$5,390.00
Floor finish2,140 sq ft——
Paint, two coats6,300 sq ft$1.85$11,655.00
Estimate≥ $17,045.00
Illustration with made-up numbers. A price nobody entered shows as a dash, and any total that includes one is marked as a minimum.

Status: a real application with real installer builds. QA and release validation are still in progress, and it has not had a public commercial launch.

02

BandBlaster

Local-first artist management for desktop

  • Electron
  • React
  • TypeScript
  • SQLite
Internal / pre-release

An artist's back office that lives on their own machine.

A local-first desktop platform for managing an artist's operations. I ran Claude Code, ChatGPT, Grok and Gemini on scoped implementation, audit and integration work, one branch and one pull request per task. Agents propose. I merge.

50 pull requests opened, 43 merged. The Windows CI workflow passed on November 8, 2026.

Status: installs locally on Windows and is used internally. It has not been released to customers, and its live connectors are still being verified.

03

The handoff system

Human-governed multi-agent development

  • Claude Code
  • ChatGPT
  • Grok
  • Gemini
In daily use

Four AI systems, one set of rules for handing work between them.

The handoff and acceptance procedures I created and run across both projects. Each agent owns a named scope. Source is fingerprinted before a handoff and checked after it. Changes land on review branches, test results are reported as they are, third-party data stays restricted, and nothing integrates until I approve it.

32 documented UI rounds between ChatGPT and Claude, covering theme consistency, accessibility and keyboard focus, with source verification and Windows testing checkpoints. Three custom agent skills now run the verification and Windows checks.
04

Audio technology and media recovery

AI-assisted analysis, separation and REAPER automation

  • REAPER
  • EEL2 / JSFX
  • Source separation
Prototype / workflow

A broken session conversion, treated like a forensics job.

I directed and operated AI-assisted audio measurement, source separation and REAPER iteration workflows. When a Logic-to-REAPER conversion came back with silent media, I recovered the source references from the defective project. I also built an AI-assisted EEL2/JSFX effect prototype.

Illustration of source separation. Much of this work integrates third-party software that I operate. The effect is a prototype, and the session recovery is partial.

Method

How a change gets from idea to merged.

The same five steps every time, whichever model is doing the typing.

  1. 01

    Specify

    Write the requirement and the hard rules before any model touches the code.

  2. 02

    Scope

    Hand one agent one small, named task and the files it is allowed to change.

  3. 03

    Build

    The agent implements, audits or integrates inside that scope and reports back.

  4. 04

    Verify

    Check the source fingerprints, read the test report, and test on a real Windows machine.

  5. 05

    Approve

    I make the call. Nothing merges or ships on a model's say-so.

My part

  • Product concept and requirements
  • The rules the AI cannot break
  • The scope of every handoff
  • Hands-on testing on Windows
  • Merge and release decisions

The agents' part

  • Most of the production code
  • Test suites
  • Read-only audits
  • Integration work
  • First drafts of nearly everything

I don't claim to have hand-written this software. I claim the direction, the rules and the verification, and I keep the records that show it.

Background

I learned to ship in front of an audience.

Michael Grippo sitting on a railing under a venue marquee that lists Spineshank, Hed PE, Primer 55 and Kissing Candice
On tour. The marquee is the visible part of the job.

I started programming in BASIC at 13 and taught myself C before college. Then music took over. I spent more than a decade as the bassist of a nationally touring band and ran its digital and business side: social, paid campaigns, the website and store, merchandise manufacturing, partnerships and licensing.

That work is where the verification habit comes from. A release date does not move because a store page is broken, so you learn to test before you ship. AI brought the programming and the operations together.

2026
Independent AI product developmentTakeoff Artist and BandBlaster, directed with multi-agent workflows. Also contract data labeling for AI model training.
From 2020
Washed Up, owner and operatorA clothing line run solo, from design and sourcing through orders and fulfillment.
From 2018
Independent social media strategy and consultingFor artists, brands, creators and small businesses.
2013 – 2022
YAI, Direct Support ProfessionalSupported young adults with intellectual and developmental disabilities.
From 2012
Kissing Candice, digital and business operationsBuilt brand and personal accounts to tens of thousands of followers, with platform verification.
2012
Stony Brook University, B.A. SociologyStudied computer science, journalism and environmental studies along the way.

Capabilities

What I bring to a team.

AI delivery

  • AI product strategy
  • Multi-agent orchestration
  • Human-in-the-loop systems
  • AI workflow design
  • AI integration governance
  • Automated research

Product and quality

  • Requirements engineering
  • Technical program management
  • Software QA and evaluation
  • Release readiness
  • UX and accessibility review
  • Source control workflows

Platforms

  • Windows desktop applications
  • Python and PySide6
  • Electron, React and TypeScript
  • SQLite
  • Git and GitHub
  • Claude Code, ChatGPT, Grok, Gemini
The stacks my projects run on and the agents I direct.

Creative and growth

  • Digital marketing strategy
  • Paid and organic social
  • E-commerce and merchandising
  • Brand and creative direction
  • Audio technology
  • Adobe Creative Cloud

Contact

Have a repetitive process, or an idea that should be a tool?

thisis@michaelgrippo.com
LinkedIn

Open to employment and consulting in AI product operations, technical program leadership, workflow automation, applied AI solutions and quality-focused product delivery.

Michael Grippo in black, leaning against a wall beside a blue door numbered 104
The door is open.