EngineersofAI LogoCoursesBlogResearch LabAI LettersThe LabCode BankInteractive 3DKodrEarnestJobs
Skip to main content
Interactive 3D/Coding Agent Loop
Task
Fix the failing test in test_utils.py
Agent Plan
○Read failing test
○Read source file
○Identify bug
○Write fix
○Run tests
○Verify
Tool Calls
No tool calls yet.
Agent Log
Press Next Step to begin.
Test Results
Waiting for agent to run tests...
Task Type
Step0 / 8
Tool calls0
Coding agents read code, reason about bugs, call tools, and iterate until tests pass. SWE-bench measures this loop on real GitHub issues.

Coding Agent Loop - Interactive Visualization

A coding agent operates in a tight loop: it reads source files and test output, reasons about the root cause, writes a fix, runs the test suite, and iterates until all tests pass. This is the exact pattern measured by SWE-bench, where models are evaluated on real GitHub issues. This visualization steps through four complete coding tasks - bug fix, feature add, test writing, and refactoring - showing the agent's internal reasoning, every tool call, and the final test results.

  • 4 tasks: Fix Bug, Add Feature, Write Tests, Refactor - each with a full multi-step agent trace
  • See every tool call: read_file, write_file, bash, search - logged in real time
  • Toggle internal reasoning visibility to understand the agent's decision-making
  • Watch the test results panel: failed → re-analyze → fix → pass
  • Plan tracker shows which phase of the task the agent is in at each step
  • Foundation for: Devin, SWE-agent, Aider, Claude Code, and all coding AI assistants

Part of the EngineersOfAI Interactive 3D - free interactive visualizations covering every major concept in machine learning and AI engineering. Hover any element for a plain-English explanation. No code required.