Notes

The first public research release is available on GitHub and preserved by Zenodo, with the accompanying preprint submitted to HAL.

August 2025–August 2026 · 5 sources

MCP, A2A, and ACP did not converge into one universal protocol. Their current roles make sense once agent systems are divided by interaction boundary.

Discusses: Announcing the Agentic Resource Discovery specification + 1 more

ARD can help agents find capabilities across organizations, but discovery metadata must remain outside the trusted execution boundary.

Discusses: From story points to tokenmaxxing: Why engineering keeps measuring the wrong things

Counting tokens repeats an old measurement mistake unless AI-assisted work is evaluated through delivery, quality, and user outcomes.

Discusses: A paper diagram visualizer

Dependency diagrams can make mathematical papers easier to navigate, provided their inferred structure is not mistaken for verified proof evidence.

Discusses: The Open Source Agent Toolkit in 2026

Agent stacks should be selected layer by layer, with explicit contracts for state, tools, memory, evaluation, and recovery.

Discusses: How does Chain-of-Thought (CoT) Prompting work? + 5 more

Chain-of-thought prompting can improve some multi-step answers, but a plausible rationale is not evidence that an answer is correct or faithful.

Discusses: Old and new apps, via modern coding agents

Terence Tao's restored mathematical applets show where coding agents have real leverage—tasks whose outputs remain cheap for an expert to verify.

Discusses: Agentic Testing: Where Agents Fit in the E2E Testing Stack + 1 more

Slack's agent-driven E2E experiments expose a useful new execution model, but adaptation must remain separate from the authority to declare success.

Discusses: The GitHub MCP Server adds support for tool-specific configuration, and more + 1 more

GitHub MCP tool filtering reduces context and accidental capability exposure, but real least privilege also requires identity, authorization, and effect controls.