Writing
RSSEssays on this site
Notes on the Claude Code 2.1.0 outage
A changelog formatting change took down Claude Code. Lessons about parsing human docs as machine data.
Notes on the Eurostar chatbot "vulnerability" report
Looking at what makes something a vulnerability versus a hardening opportunity in LLM applications.
What I learned shipping 1,000+ PRs with Claude Code
Notes from using Claude Code in parallel git worktrees: Plan Mode, ultrathink, verification loops, and Chrome automation.
Selected writing elsewhere
McKinsey's Lilli Looks More Like an API Security Failure Than a Model Jailbreak
Why the reported Lilli incident looks like an application-security chain reaching an AI system, not a model jailbreak.
(opens in a new tab)Promptfoo is joining OpenAI
Announcing that Promptfoo has agreed to be acquired by OpenAI.
(opens in a new tab)How AI Regulation Changed in 2025
Why "AI compliance questions" appeared in security questionnaires and RFPs, and how policy becomes contract requirements.
(opens in a new tab)Why Attack Success Rate (ASR) Isn't Comparable Across Jailbreak Papers
ASR isn't portable across papers because measurement choices dominate the headline number. Includes math and a checklist for reading papers.
(opens in a new tab)GPT-5.2 Initial Trust and Safety Assessment
Day-zero red team of GPT-5.2 focusing on jailbreak resilience and harmful content.
(opens in a new tab)Real-Time Fact Checking for LLM Outputs
Introduces search-rubric, an assertion where a search-enabled judge verifies time-sensitive claims during evals and CI.
(opens in a new tab)When AI becomes the attacker: The rise of AI-orchestrated cyberattacks
Connects malware querying LLMs at runtime with "vibe hacking" case studies. Defense needs continuous testing.
(opens in a new tab)Reinforcement Learning with Verifiable Rewards Makes Models Faster, Not Smarter
RLVR gains are often "search compression" rather than new reasoning ability.
(opens in a new tab)Prompt Injection vs Jailbreaking: What's the Difference?
Jailbreaking targets model safety training; prompt injection targets application trust boundaries.
(opens in a new tab)AI Safety vs AI Security in LLM Applications: What Teams Must Know
Safety protects people from harmful outputs; security protects systems from adversarial manipulation.
(opens in a new tab)Promptfoo Raises $18.4M Series A
Announcing our Series A led by Insight Partners with participation from a16z.
(opens in a new tab)Evaluating political bias in LLMs
Open methodology and dataset (2,500 political statements) to measure political leaning in models.
(opens in a new tab)Celebrating 100,000 Users
Promptfoo's journey from prompt evaluation to AI red teaming, marking 100,000 users.
(opens in a new tab)