I Built an AI Coding Harness Because I Got Tired of Paying the Smart Model to Type
A Claude Fable + GPT-5.5 Codex harness that spends Fable on judgment, Codex on execution, and the repo as memory.
AI security research / offensive engineering / statistical systems
Hacking, AI, Agents, and Statistics.
Writing
I Open-Sourced My UFC Prediction Model, Weights, and Database
Five years of building an open-source UFC prediction model, weights, database, and a pile of lessons about leakage, validation, calibration, and odds.
Clankerfights Is Live
Watch unhinged AI play games, wager on who will win, influence the outcome with prompt injection, clip funny moments, and build your own multiplayer games.
How I Botted $6k On The Prediction Markets As I Slept
A simple Kalshi mention-market bot, real PnL, the data behind it, and why I stopped.
2025's Most Extreme Fighter Outliers: The Statistical Freaks
Current-form fighter outliers, adjusted by division norms and fight logs.
Demystifying the MMA-AI Prediction Algorithm
Fight variables, ensemble modeling, feature pruning, validation, and confidence.
Data Drift, Generalization, and the Quest for a Bulletproof UFC Model
Training windows, recency weights, validation splits, and market drift.
Projects
DanMcInerney on GitHub
Security automation, exploit research, Python tooling, AI systems, and old-school offensive utilities.
95 public repos
15,371 stars
Featured project
Kalshi Bot
Kalshi autotrading system and analysis pipeline for finding and trading prediction-market inefficiencies.
Vulnhuntr
Autonomous LLM 0-day finder for Python codebases, centered on remotely exploitable vulnerability classes.
MMA-AI.net
UFC fight predictor using a multi-model ensemble, Optuna tuning, SHAP-assisted feature pruning, and time-aware validation.
AI Exploits
Collection of real-world AI/ML exploit material for responsibly disclosed vulnerabilities.
wifijammer
Wireless attack utility from the offensive tooling era: continuously jam Wi-Fi clients and routers.
Speaking
2019-10-11
BruCON 0x0B - Tales from Professional Thieves
Physical security bypasses and field stories with Justin Wynn.
2023-11-21BlueHat - AI Security: Like Hacking in the 90s
AI security and the early shape of a new attack surface.
2024-10-25No Hat 2024 - LLMs for Vulnerability Discovery
Finding 0-days with LLM-assisted research.
2024-02-13Between Two Vulns: Secrets in Triton's Inference Server and MLFlow
AI infrastructure vulnerability discussion covering Triton Inference Server and MLflow.
Show 16 more talks
| Date | Talk | Notes |
|---|---|---|
| Security Weekly #416 - Python for Pentesters | Paul's Security Weekly technical segment. |
|
| NolaCon 2018 - Automahack | Python toolchain for automated domain admin. |
|
| Circle City Con - Automahack | Automating the path from zero to domain admin. |
|
| SAINTCON 2018 - Icebreaker.py | Gaining an Active Directory foothold from one compromised host. |
|
| BSidesSLC 2018 - Icebreaker | From internal jumpbox to domain admin in one command. |
|
| Coalcast Episode 5 | Conversation with Marcello Salvati and Dan McInerney. |
|
| How to Break into Commercial Buildings in America | Hackfort 2019 physical security talk. |
|
| ROOTCON - AI's Underbelly: The Zero-Day Goldmine | MLOps and AI tooling attack surface research. |
|
| AI/ML Bug Bounty Pro Tips | Practical guidance from AI/ML vulnerability triage. |
|
| I Hunt AI Engineers | Hackfort 2024 AI security talk. |
|
| Jailbreaking Gemini: Did We Just Uncover Data Leaks?! | Discussion of Gemini jailbreak behavior and possible data leakage risks. |
|
| AI Security: Vulnerability Detection and Hidden Model File Risks | Model file risk, vulnerability detection, and AI security discussion. |
|
| LLMs for Vuln Discovery | Finding 0-days with a click of a button, with Marcello Salvati. |
|
| SANS - Hacker's Perspective: Realistic AI Attack Scenarios | Practical AI attack scenarios for defenders and security teams. |
|
| Practical AI Security: Past, Present and Future | AI Talks appearance on the evolution of practical AI security. |
|
| Augmenting Your Offensiveness With AI | AI for offensive security with Marcello Salvati. |