Topic hub
AI Agents
Build autonomous AI agents that reason, use tools, and accomplish complex tasks.
All posts · 42
AI AGENTS01
AI Agent Says Done But Did Nothing: Detect False SuccessRoughly 45 to 48 percent of agent failures close with a confident completion claim. LLM judges catch almost none of them. Here is what actually does.
AI AGENTS02
AI Due Diligence Data Room Review: From Index to Request ListPreregistered testing put hallucination at 17 to 33 percent in retrieval-backed legal AI. Here is the four-pass data room review that survives a dispute.
AI AGENTS03
AI for AML Alert Triage: Cutting a 95% False Positive RateRule-based AML monitoring runs 90 to 95% false positives at 30 minutes an alert. Here is the triage pipeline, the evidence contract, and the validation path.
AI AGENTS04
MCP Error -32000: Connection Closed (6 Causes and Fixes)MCP error -32000 is not a timeout and not an MCP protocol code. One public debug log fails at 6544ms inside a 30,000ms budget. Six causes, in triage order.
AI AGENTS05
HANDBOOK.md Benchmark: Why Agents Ignore Your Policy DocThe best of 30 model configs passes just 36.2% of HANDBOOK.md trials under strict grading. Why long policy docs fail, and what to compile instead.
AI AGENTS06
Migrating AI Agents Between Models: 4 Silent BreaksFour things break silently when a production agent crosses model families: tool schemas, prompt caching, reasoning state, prompt style. Plus the fix for each.
AI AGENTS07
Agentforce vs Copilot Studio vs Vertex Agent BuilderAgentforce vs Copilot Studio vs Vertex Agent Builder decoded by data gravity: $2/conversation vs $200/25K credits vs per-token, and which fits your stack.
AI AGENTS08
Single Agents Beat Multi-Agent Swarms at Equal CostStanford held the thinking-token budget constant: single agents matched or beat multi-agent swarms at every level. When a swarm is worth its 15x token bill.
AI AGENTS09
Stop AI Agents Looping on the Same Failed Tool CallRecursion and iteration limits (25, 15) only cap total steps. A no-progress guard that hashes repeated (tool, args, error) tuples stops stuck agents early.