Category: reading
-
Talk with Shunyu Yao：AI 研究的反馈、系统与长期方向
TLDR: The conversation frames agent research as moving from raw model scaling toward long-horizon tool use, memory, personalization, science, and grounded reliability.
-
Anthropic Blogs：Harness Engineering 与 Context Engineering
TLDR: Long agentic tasks fail when context, tool use, and coordination drift; the useful lesson is to treat context engineering as runtime design.
-
Building a C Compiler with Agent Teams
TLDR: A practical multi-agent software pipeline can stay simple: split compiler work across coding agents, isolate tasks, and judge progress with integration tests.
-
Claude Code Auto Mode：用 Classifier 替代手动权限审批
TLDR: Claude Code auto mode is a permissions problem: useful autonomy needs explicit trust boundaries around repos, commands, domains, and data access.
-
Claude Code Source：把 Agent 看成一个操作系统进程
TLDR: Looking at Claude Code as an OS process exposes the practical substrate of agents: files, permissions, plugins, subprocesses, and tool calls.