| What it is | The rule it never breaks | |
|---|---|---|
| Orbit macOS |
A menu-bar app that turns what you work on into searchable memory, so "where did I leave that project?" gets a real answer: actual file names, URLs and projects. Rust · Tauri · React · FastAPI · SQLite FTS5 · Qdrant · Groq |
Never reads a password field or a keystroke, redacts secrets before anything reaches disk, and routes no data through a server of mine. Bring your own key, zero telemetry. |
| baseline‑skills Agent Skills |
Three skills that give any AI coding agent (Claude Code, Codex, Cursor, Antigravity, Windsurf) a secure-by-default baseline: OWASP, CIS and NIST-aligned review, project scaffolding, and pre-merge secret and dependency scans. Python stdlib · Agent Skills spec |
No company, cloud or framework baked in, and every script is short enough to read before you trust it. |
| Pakistan petrol prices Web |
Daily OGRA-notified fuel prices with full history, plus the Platts and exchange-rate inputs behind every change. Next.js · Supabase · Cloudflare Workers · Recharts |
Never shows a price it can't source. A failed sync keeps the last verified price, and the test suite fails if that ever changes. |
| saadaan.dev Web |
My site, on a CMS I built. Its chat agent answers from published work through tool calls, and AI assistants can draft posts into the CMS over MCP. Next.js · Supabase · AI SDK · Groq · MCP |
The chat never invents a client, a metric or a price, and the drafting credential can create drafts but never publish one. |
Client work from DotCode, including a laboratory information management system and a performance-based hiring platform, is written up as case studies.
- Start with how people actually work. A requirement tells you what someone wants built. Watching how they actually work tells you what they need, so that is where I start.
- The "never" rule comes first. Every project above has one, and it shaped the architecture more than any feature did: Orbit's bring-your-own-key design, the petrol tracker's sync tests, the CMS's draft-only credential.
- Agents write a lot of my code. I own the decisions. I review an agent's architecture the way I'd review a teammate's PR, and I built baseline-skills so the security basics don't depend on remembering to ask.
- What Is MCP? How Model Context Protocol Lets AI Agents Use Your Software
- How Much Does It Cost to Build an AI SaaS MVP in 2026?
- The Skill That Matters More When AI Writes Your Code
| Product | |
| Backend & data | |
| Systems & infra | |
| AI | Groq · OpenAI · Gemini · Voyage AI · RAG on pgvector and Qdrant · MCP · Agent Skills |






