Michael Ramos
Michael Ramos is an individual software engineer, AI builder, and open-source creator rather than an incorporated company. His public work focuses on practical AI-assisted software development: giving developers better context, review, feedback, and control over coding agents, while his professional background centers on complex AI systems and public-sector deployment.
At a Glance
- Software engineers and teams using Claude Code, OpenCode, Pi, Codex, Gemini, Droid, Kiro, Copilot, Amp, or similar coding agents
- AI-assisted development and code-review teams
- Developers who need local-first, human-in-the-loop control over agent output
- Machine-learning and enterprise AI teams concerned with governance, provenance, and verifiable computing
AI Tools by Michael Ramos
(2)
pstack
Agent Skills for Coding Agents
Plannotator
AI Agent Plan Review UI
Discussions
No discussions yet
Be the first to start a discussion about Michael Ramos
Latest News
Plannotator continued shipping responsive review and inbox improvements for iPad, narrow windows, and mobile layouts.
bro added the ladder skill for rendering a lane of work as a vertical progress ladder.
Published Context Monorepos, describing a meta-operating layer for organizing research, decisions, and multiple product repositories for longer-running agent work.
Published Please Do Not A/B Test My Workflow, arguing for transparency and configurability in AI coding tools.
Products & Services
Open-source, local browser-based review surface for AI coding agents. It lets users annotate plans, specifications, Markdown, HTML artifacts, agent messages, local changes, and remote pull/merge requests, then send feedback back to the agent.
VS Code extension that turns an entire codebase, selected files, directory structure, GitHub issues, and PR diffs into AI-ready context; includes visual file selection, token counting, structured packaging, and .towerignore controls.
A collection of practical quality-of-life skills for AI-assisted development, including bro (plain-language restatement), readback, facts, recap, clean-room, status, and ladder. Distributed as Claude Code and Codex plugins/skills.
Shell aliases and an install script for common Git worktree operations: creating feature worktrees, reviewing remote branches, switching between worktrees, and cleanup.
Market Position
Ramos's projects compete primarily as open-source, developer-controlled alternatives to opaque or hosted AI-coding workflow tools. Plannotator differentiates through local/browser review, direct agent feedback, and broad harness/version-control integrations; Prompt Tower differentiates through explicit repository context selection and packaging rather than manual copy/paste. The broader positioning is human-controlled, inspectable agent workflows rather than autonomous black-box coding.
Leadership
Founders
Michael Ramos
Software engineer and AI practitioner with more than a decade of public-sector experience. A 2024 AI Conference bio says he implemented AI for bank marketing from 2017, led major U.S. government AI projects, and grew an AI program to more than $120 million in value; his public profiles describe him as a veteran and cofounder of EQTY Lab, with current work involving AI systems and governance.
Executive Team
Michael Ramos
Independent software engineer, AI builder, and creator
Public profiles identify Ramos as the owner of the backnotprop GitHub account and website, a veteran and AI systems practitioner, and a cofounder of EQTY Lab. His conference biography describes public-sector AI deployment, bank-marketing AI work, and leadership of a government AI program valued at more than $120 million.
Board of Directors
Founding Story
This is a personal creator practice, not a separately described company. Ramos's site and GitHub show a progression from technical experiments and educational writing (stable matching, TensorFlow/CoreML style transfer, Kubernetes) into open-source tools for AI coding workflows. The recurring motivation in his writing is to keep humans in control of agent planning, context, review, and verification rather than hiding those mechanisms behind opaque memory or automation.
Business Model
Revenue Model
The principal personal product, Plannotator, is free and open source and runs locally. A hosted Workspaces layer for teams is shown as a waitlist product; no public paid pricing or revenue figures were found.
Pricing Tiers
Free and open source; local operation with no usage telemetry or analytics according to the product site.
Hosted layer for teams to turn plans and code reviews into shared team knowledge; the site does not publish a price.
Target Markets
- Software engineers and teams using Claude Code, OpenCode, Pi, Codex, Gemini, Droid, Kiro, Copilot, Amp, or similar coding agents
- AI-assisted development and code-review teams
- Developers who need local-first, human-in-the-loop control over agent output
- Machine-learning and enterprise AI teams concerned with governance, provenance, and verifiable computing
- Human-in-the-loop planning and specification review before an AI coding agent implements work
- Reviewing AI-generated code changes and pull requests
- Annotating HTML artifacts and agent messages
- Building focused prompts from large repositories for Claude, ChatGPT, Cursor, Gemini, and other coding assistants
- Managing parallel agent work with Git worktrees
- Searching prior Claude Code sessions and project history
- Plannotator is publicly presented as an open tool rather than a customer case-study business; the product site names integrations and community contributions but does not identify paying customers.