Scratch archive
Experiments
A record of the projects that did not become the main thing, but still mattered. Some were scratched. Some were suspended. All of them sharpened the next decision.
Cut cleanly
Scratched experiments
Projects that were explored, built, or validated far enough to teach something useful, then deliberately stopped.
Hyperliquid perpetuals trading bot explored as a systematic crypto execution project. It reached strategy and implementation work, but never went live.
Automated trading needs evidence, monitoring, and risk controls before ambition.
iPhone arbitrage deal finder built to spot resale opportunities and surface profitable listings. The MVP worked, but was not worth pushing into deployment.
It clarified the difference between a clever scraper and a durable product with repeatable edge.
Meeting and productivity idea investigated before build momentum took over. Existing tools already solved enough of the problem.
The best engineering decision was not to build. Research beat sunk cost.
Early SAP Business One MCP-first direction that helped shape thinking around ERP-connected agents before ARGUS became the stronger path.
The idea did not disappear. It got absorbed into stronger patterns around support, retrieval, and business-system context.
Human-AI collaboration portfolio platform that explored how to make process, iteration, and agent collaboration visible as proof of work.
Parts of the thesis survived here: the portfolio should show judgment and collaboration, not just finished screenshots.
Held deliberately
Suspended or paused projects
Projects that were real enough to preserve, but not right to keep active right now.
Agent-native recruiting MCP server and portal concept for candidates, recruiters, companies, and academic institutions. Built around structured profiles, verification, and AI-assisted matching.
The matching and verification ideas were strong, but the hackathon push was cancelled and focus moved back to higher-priority work.
Polymarket BTC 15-minute Up/Down trading bot that ran live for roughly a week before being paused with data preserved.
Live trading systems need stronger strategy evidence, monitoring, and operational confidence before they deserve more capital or attention.
Polymarket BTC 5-minute trading bot explored as a faster market variant after CISCO. Services were stopped and the data was preserved.
Faster markets do not automatically mean better opportunity. The shorter loop increased noise and reinforced the need for stricter validation.