Short notes on AI systems, engineering, and ideas worth revisiting. Each entry links back to its original source.
Josh Rosen surveys software-factory architectures with attention to the orchestration outside an individual coding agent. Selecting work, coordinating execution, checking outcomes, and deciding what happens next are distinct from generating a patch.
The author reports an incident in which malicious instructions in a SKILL.md file caused downloaded malware to reappear. It is a personal incident account, not an independently verified investigation. The lesson is to treat retrieved skill instructions as executable influence over an agent’s behavior.
Robert Martin points to the swarm-forge repository in this reply. It is a compact bookmark for exploring how multiple coding agents can coordinate their work, with the implementation serving as the primary reference.
Uber Engineering discusses scaling a software factory efficiently. The article puts operational economics alongside agent capability: the cost and behavior of complete sessions matter when moving from isolated experiments to a large volume of automated work.
Bilawal Sidhu shares a screenshot highlighting God’s Eye View on GitHub’s trending list. This bookmark captures interest in the visualization project; the screenshot documents the moment rather than establishing a lasting ranking.
Femke Plantinga highlights the variety of architectures being described as a company brain. Slite’s accompanying research looks at components, maintenance costs, and build-versus-buy choices, helping turn a broad label into concrete design decisions.
LangChain’s OpenWiki article introduces WikiBench, evaluating generated repository knowledge through questions answered by reader agents. The interesting shift is to test whether documentation helps a downstream user solve a problem, rather than judging the document only by its appearance.
Jacob Posel argues for company brains designed around agent use and connections to live business systems. A static collection of documents is only part of the picture; usefulness depends on whether an agent can access relevant, current information in its actual workflow.
Philipp Schmid shares the MCP roadmap and an accompanying overview image. The roadmap is a useful reference for tracking how the protocol’s priorities evolve as usage expands beyond initial tool integrations.
This Spanish-language roundup collects Graft, Agency Agents, Codebase Memory MCP, OpenMontage, Agent-Reach, and Orca. They span efficiency, specialized agents, code knowledge, video production, web access, and orchestration. The links are useful discovery leads rather than validated comparisons of performance.
