Failure Briefs
Postmortem-style analysis of AI system failures, fragility, and production risk.
Guides and explainers
Detailed guides and practical technical analysis.
Latest analysis
Recent research, benchmark reviews and technical updates.
Practical tools
Templates for budgets and project planning
BoredTools offers practical spreadsheets for budgets, freelance work and small projects.
AI Agents in Legal: What Works, What Fails, and What the Sanctions Data Actually Shows
In June 2023, attorneys Steven Schwartz and Peter LoDuca submitted a brief in a federal case citing six cases that did not exist. ChatGPT had invented them. When the opposing party asked for copies, the attorneys submitted fabricated pages. A judge sanctioned them $5,000 and required them to pers...
Why AI Agent Deployments Fail — And What the Survivors Do Differently
Agent deployments fail for recurring reasons: weak problem framing, brittle long-horizon performance, poor observability, and missing human-in-the-loop controls.
Enterprise AI Pilots Have a 70% Failure Rate
S&P Global found 42% of companies abandoned most AI initiatives. MIT reports 95% of GenAI pilots deliver no measurable return. The technology works. The organizational machinery that carries pilots to production doesn't.
RAG Pipelines Are Silently Dropping Context
Your RAG pipeline retrieves the right documents. The LLM ignores half of them. The RAG-E framework found generators skip the top-ranked passage in 47-67% of cases. The retrieval-utilization gap is the real bottleneck.
Red Teams Found Agents Leak More Than Models
Red teams found agents are far more vulnerable than standalone models. Mixed attack strategies hit 84.3% success rates. Memory poisoning persists across sessions. Every tool is a potential exfiltration path.
OpenAI Agents SDK in Production: Traces, Tooling, and Hand-offs That Don’t Break
Build reliable agent workflows with OpenAI Agents SDK: traces, tool-call guardrails, handoffs, retries, and deployment checks.
Multi-Agent AI Has a Security Architecture Problem That Better Models Won't Fix
193 documented threats. Agent defection. Reverse SSH tunnels. Why better models won't fix multi-agent AI security — and what actually helps.
Your Agent's System Prompt Is Fighting Itself
A framework called Arbiter treats agent system prompts as auditable code. Applied to Claude Code, Codex CLI, and Gemini CLI, it found 152 interference patterns — including critical contradictions and a structural data loss bug — for a total cost of $0.27.
Alignment Works in English. In Japanese, It Backfires.
A new study shows the same alignment intervention that produces strong safety effects in English reverses direction in Japanese, increasing harmful outputs. Tested across 1,584 simulations, 16 languages, and three model families.
Most AI Agents Don't Know When They're Wrong
A 4B parameter model just matched GPT-4o on tool-use tasks by learning to verify its own actions. The CoVe paper shows verification-first training beats the retry-and-pray approach plaguing production