Anthropic publishes the research. The agent reproduces it in 34 seconds for $0.08. €39 puts the code in your terminal tonight. €69 gets you offense + defense.
Warface Academy turns every published frontier-security paper into a hands-on lab. Build the agent, watch it work, defend against it.
The biggest names in security training — SANS, OffSec, EC-Council — charge €2.000 to €8.700 for courses that don't yet teach what Claude Opus 4.5 did in December 2025: autonomously find $4.6M in smart contract exploits.
We translate every Frontier Red Team paper into a working module you can build, measure, and ship in 3 hours. €39 instead of €3.900.
| Course | Price |
|---|---|
| SANS SEC565 | €8,700 |
| OffSec AI-300 (OSAI+) | €2,599 |
| EC-Council COASP | €2,500+ |
| smartcontractshacking.com | €1,500+ |
| Warface Module | €39 |
10-minute summary of the source Anthropic paper. The numbers, the methodology, why it matters.
Clone the repo. Run the setup. Build the agent step-by-step alongside the video. Working in 90 minutes.
Every module ends with the defender mindset: how do you mitigate what you just built? You leave able to do both.
Each module reproduces a specific piece of public research from Anthropic's Frontier Red Team and partners. New module every 1-2 weeks.
Reproduce Anthropic's SCONE-bench findings. $4.6M in exploits, your terminal.
Reproduce Anthropic's September 2025 paper. 76.5% Cybench, in your terminal.
The complete loop: exploit smart contracts, then defend Flask apps. Two modules, one mindset.
Multistage Network Attacks with Claude
Equifax breach simulation, isolated VM lab
Property-based Bug Hunting
Find real bugs in Python packages
0-day Discovery at Scale
Reproduce published CVE methodology
Claude in CTF Competitions
Top-25% with an autonomous agent
Bundles (the 2-module bundle is live now; larger bundles unlock as more ship)
€69
2-pack live
€129
5-pack -34%
€299
All-access (15) -59%
€19/mo
Membership · all + new
Every cost / time / step number on this page comes from real reproducibility runs on Sonnet 4.5. Each module ships with its own logs; you can re-run with your own API key and check the math.
Median across 5 runs against a deployed ReflectionToken contract on local anvil. Validator independently confirms attacker balance grew past 100e18 wei threshold. Stop reason: end_turn, all 5 runs.
Median across 4 runs against a Flask app with 5 planted CWE vulnerabilities (SQLi, hardcoded secret, path traversal, weak auth, stored XSS). All 4 runs found all 5 — TP=5, FP=0, FN=0. Scored against ground truth.
9 reproducibility runs total on Sonnet 4.5, May 2026. Every run logged in logs/run-*.json. Reproduce yourself with your own API key — the validator is independent of the agent's self-report.
If you read the Anthropic blog and immediately wanted to try it yourself but got lost in setup, papers, and "where do I even start" — this is for you.
You should be comfortable with Python and the terminal. Solidity, Foundry, MCP — we'll teach you those.
Two modules live, more shipping every 1-2 weeks. Lifetime access, 14-day refund.
Secure checkout · 14-day refund · Lifetime updates
A short daily email introducing the foundations of agentic AI red teaming. One email per day, 7 days. Then weekly research roundup. Unsubscribe anytime.
