# Theory Delta > A field guide for the agentic tool landscape. Every claim is traced to a primary source or labelled as Theory Delta's own analysis. Designed for both human readers and LLM agents (via MCP). ## About - [Field guide intro](https://theorydelta.com/): What Theory Delta is and how the evidence works. - [For agents](https://theorydelta.com/for-agents): MCP configuration to query findings programmatically. - [All findings](https://theorydelta.com/findings/): Full index of published findings. - [By task](https://theorydelta.com/tasks/): Findings grouped by what you're trying to do. ## Recent findings - [Hermes Agent's self-improvement narrative is not supported by the current codebase](https://theorydelta.com/findings/hermes-agent-self-improvement-non-functional): GitHub issues show that Hermes's headline self-improvement feature — autonomous skill discovery via skill_view() — is non-functional in production, and the companion self-evolution repo implementing Phases 2-5 has not shipped and is 63 days stale. - [Claude Code Model Aliases Silently Hardcode Instead of Tracking Latest](https://theorydelta.com/findings/claude-code-model-aliases-silent-failures): The source shows that model aliases in Claude Code agent frontmatter hardcode to a specific dated model ID rather than tracking latest — so teams believe they are either version-pinned or version-following, and they are neither. - [LangGraph checkpoint round-trips silently drop typed state — four confirmed failure modes](https://theorydelta.com/findings/agentic-rag-pipeline): Source-reviewed evidence across 4 open GitHub issues shows LangGraph checkpoint round-trips silently drop non-primitive types — StrEnum→str coercion, nested Enum→None, null-on-deserialize, BinaryOperatorAggregate wrapper leak — with no exception raised in any of the four modes. - [Pipecat's interrupt handling drops function call results and deadlocks — two independent mechanisms](https://theorydelta.com/findings/pipecat-voice-framework): We traced 8 public GitHub issues confirming that Pipecat's interrupt/resume semantics have at least four distinct bug classes — queue recreation, deadlock, frame-drop, and race condition — that make function calls during speech unreliable in production without targeted workarounds. - [Cursor Automations Introduces Always-On Agents With No Execution Visibility](https://theorydelta.com/findings/cursor-automations-no-execution-visibility): The source shows Cursor Automations (Mar 2026) ships always-on webhook-triggered agents that auto-approve all actions and expose no execution state during a run — a qualitatively different risk posture than the approval-gated local agent mode builders tested. ## Machine-readable endpoints - MCP server: `https://api.theorydelta.com/mcp` (HTTP transport) - MCP discovery: - A2A discovery: - RSS feed: - Sitemap: