2月5日 周四
实验室前沿Lab Notes
全部Anthropic2 · Anthropic
Anthropic · 工程
Infrastructure noise in agentic coding evals
Agentic coding benchmarks are treated as precise measures of model capability, but infrastructure configuration alone can meaningfully shift…
Anthropic · 工程
Building a C compiler with parallel Claudes
How a team of 16 parallel Claude agents wrote a Rust-based C compiler from scratch, capable of compiling the Linux kernel, and what it taugh…
X 风向标Signal
全部脱敏 mock：真实样本截断 + 少量演示卡；不含 raw_pages / 完整抓取包。
21:00–00:00 CST
18:00–21:00 CST
15:00–18:00 CST