Explores a software library with no code, using AI agents to generate implementations from specifications and conformance tests.
Explores Kent Beck's 'coding genie' metaphor for LLM coding agents, warning against anthropomorphizing them as they are probabilistic tools.
Exploring how asynchronous coding agents like Claude Code and Codex can automate code research projects by running experiments and submitting pull requests.
Anthropic proposes a new MCP approach that converts tools into executable code for more efficient AI agents, reducing token usage and errors.
