I’m not a software developer. I’m a lifelong geek and gamer who somehow uses AI to shape messy ideas, workflows, and personal problems into practical tools and protocols.
I usually have too many ideas in my head. When I can, I try to turn some of them into something real and share the result freely. Most of my projects start as solutions to problems I actually run into, which is why they can sometimes be useful to other people too.
I don’t build things just to have a portfolio. I build when an idea feels real, when it solves something for me, and when it seems worth shaping into something other people can try, inspect, or reuse.
Most of what I create falls into personal tools, AI protocols, and practical systems shaped by real use.
- ideas shaped into usable tools
- practical systems for real problems
- experiments that try to become tools
- projects made first for my own needs, then shared for others
Try DeadPunk OS — it’s not broken, it’s "avant-garde".
A few projects I use, test, or keep shaping:
| Project | What it is |
|---|---|
🧮 A.D.A.M. - Adaptive Depth and Mode |
A.D.A.M. helps keep AI chats at the right depth, so simple questions stay simple and complex work doesn’t become unclear or overloaded. |
🧰 ai-protocol-kit |
A working kit of AI behavior contracts and operational protocols for shaping ideas, reviewing work, preparing pages and repositories, writing for real readers, and keeping AI-assisted workflows from breaking in stupid ways. |
🧠 Memory-Assisted Shaping for GPT Project |
Memory-Assisted Shaping helps keep long ChatGPT and GPT Project idea-shaping sessions coherent, so decisions, gates, discarded paths, and source boundaries do not get lost before the final artifact is ready. |
⚖️ PA-PVP |
PA-PVP helps turn vague plans, artifacts, or stuck decisions into a clear verdict, one next action, and a continuation point instead of another open-ended discussion. |
🔎 PA-PVP mini |
Lightweight adversarial review protocol for stress-testing ideas, plans, prompts, procedures, code, and AI outputs. It finds gaps, weak logic, missing constraints, and shaky fixes, then lets another AI challenge both the original work and the previous review. |
🛤️ Signal Rail |
Signal Rail helps keep live project material in the right place, so decisions, open work, ideas, constraints, and handoff notes do not collapse into one messy document. |
☁️ Cloudflare R2 Remote MCP Worker |
Cloudflare R2 Remote MCP Worker gives remote MCP-based hosts, such as ChatGPT, a personal bridge to files mirrored on Cloudflare R2, so they can work with your storage without needing a local MCP server or a native R2 connector. |
🧯 Canon Boundary Guard for Codex |
Codex skill that helps keep project evidence, chat context, operator instructions, working hypotheses, and model assumptions in separate cognitive layers. |
🧬 Canon Boundary Guard for GPT Project |
Helps ChatGPT keep uploaded files, chat history, instructions, hypotheses, drafts, and model assumptions separate in long Projects or chats before they silently become canon. |
🪞 Shaping Frame for Claude |
Shaping Frame helps keep long Claude sessions clear by separating hypotheses, decisions, Claude-generated material, and operator-approved direction. |
📚 CSBP - Codex Shared Best Practice |
CSBP is a small companion layer for Codex practices that are worth reusing, but not strong enough to become project rules. |
🛠️ SensecraftXStudio |
SensecraftXStudio helps keep AI-assisted technical work grounded before the assistant acts, so the target, context, assumptions, and final state stay visible for the operator and clear for the agent. |
🚀 GPT-PF / ChatGPT Project Forge |
Project Forge prepares the starting files for a ChatGPT Project, so scope, sources, constraints, and starting basis are clear before the real work begins. |
🧩 ChatGPT-SKILL-SYSTEM |
GPT Project Skill System gives ChatGPT Projects a controlled session workflow for using skill-like packages: boot the core, load named skills, and check external skills before putting them into context. |
Most of these projects are meant to be used with an AI assistant, not just read as static documentation.
If a repo is unclear, paste the repository link into your AI chat and ask:
What is this project for, how does it work, and what could I use it for in my own workflow?
Then ask for practical examples, limits, which files to read first, and whether it actually fits your problem.
That is often the fastest way to understand these projects, because many of them are not just documents. They are practical frameworks, protocols, or small systems shaped by real use.
These projects were created with AI assistance.
The ideas, documentation, repository materials, and practical structure were shaped through human-directed work supported by AI tools during drafting, structuring, review, testing, and refinement.
AI assistance does not make any project automatically correct, complete, or suitable for every use case. Read it, test it, and adapt it to your own context.
Because if something solved a real problem for me, there is a good chance it can help someone else too.
Some projects are rough. Some are weird. Some are more serious than they look at first glance.
But they are real, and they come from use, not from trying to look impressive.
