Selected writing.
A curated list — not a CMS firehose. Order is by impact, not date. Talks live on /speaking.
1337 Use Cases for ChatGPT & Other Chatbots in the AI-Driven Era
An operator’s field guide to LLM workflows, expanded from a Medium essay read by 10,000+ people. Cited at least twice in academic literature on AI-in-education and graph databases.
ACTFOIA — Freedom of Information, automated
The methodology behind 15,000 FOIA requests to 5,000 Romanian public bodies. Templates, parsers, escalation timers, the legal scaffolding. The talk version landed at HOPE XV.
ADR-driven development — the practice, not the artifact
Why a 443-record decision log is the cheapest source of context for the LLM agent already in your engineering pipeline. The longer treatment runs at the masterclass.
Rootless computing — a manifesto for self-hosting in 2025
On running your own stack on Hetzner / Proxmox / Kubernetes, agent-aware. Drafted in public on Medium; the formal version lands when ServerClaw 1.0 ships.
GIJC17 — presenting ACTFOIA, a tool for journalists and civic activists fighting corruption
The first public write-up of the FOIA methodology, delivered at the Global Investigative Journalism Conference.
Analyzing protest trends in Romania 2012–2017 using Facebook data
POINT Conference, Sarajevo. The dataset, the methodology, and the limits of Facebook-as-civic-infrastructure.
sched.com — disclosing a path-traversal flaw across 4,467 events
How a double-slash URL bypass exposed 1.7M user records, the responsible-disclosure timeline (vendor patched within 24h), and what Facebook’s Associate General Counsel said next.
ADR-Driven Development Masterclass · Cohort 4
A live cohort course for engineering leads who already have an LLM agent in their pipeline and have noticed the agent makes bad calls. The fix is upstream of the agent.
MrData on YouTube — OSINT explainers and data storytelling
Short, evidence-led explainers about open-source intelligence and structured public data.
Rootless Computing — a methodology for self-hosted, AI-native operations
The longer treatment of the practice behind ServerClaw: how to run a serious stack on Hetzner / Proxmox / Kubernetes, agent-aware, with every decision logged in an ADR. The methodology has its own canonical site.
github.com/baditaflorin — 100+ public repositories
A decade of open-source work across data infrastructure, civic tooling, OSINT, MCP servers, and Go services. 293 public repos, ~1,100 stars across the network. The repos below are the ones worth starting with.
0exec — agent execution catalog
A catalog of safe, composable execution primitives for LLM agents.
AI & LLM operator courses
The expanded teaching surface from 1337 Use Cases for ChatGPT. Cohorts for engineering teams and civic-sector operators who want the practitioner’s map, not the marketing one.