Guide contents
How to find and fix errors in the work of AI agents
We analyze three levels of debugging: from filtering launches to changing the model
This guide is available to AI-Base subscribers
Get full access to this practical guide, ready-made configurations and terminal scripts.
Related Ecosystem Resources
Explore practical workflows, core terminology, and engineering tools
Hallucination Triage & Root Cause
A systematic engineering process for analyzing AI model errors: diagnosing context defects, identifying prompt conflicts, and establishing systemic rules instead of mindless query repetition.
Subagents and Parallel Execution in Claude Code: Running Multiple Tasks Simultaneously
A comprehensive guide to task parallelization in Claude Code: subagent architecture, shared vs. isolated context, Fan-Out and Swarm patterns, headless CLI orchestration with -p, and token economics.
Monitoring AI Agents: Metrics, Tracing, and Root Causes of Failures
Standard server and application APM fails for agents: an agent can return HTTP 200, burn through tokens, and still give the user an incorrect answer or get stuck in a loop.
Self-Correction Loop
A mechanism for autonomous code correction by the model through receiving grounded feedback from compilers, linters, or tests.
AGENTS.md in Ghostty and Turborepo makes AI agents report on users
Examples of AGENTS.md files show how repository instructions can force AI agents to add insulting notes about developers to pull requests.
MIT researchers show LLMs can read and edit neural network weights without training
MIT CSAIL researchers introduced Weight-Space Programming, where an LLM directly reads and modifies model weights; Astra improved Qwen3.5-0.8B math scores without gradient descent or fine-tuning.