Weekly Paper Notes — three entries per week: 1 seminal + 2 top picks.
🔁 Seminal Paper of the Week
- The Byzantine Generals Problem — Lamport, Shostak & Pease (1982). The paper that named the problem, proved $n \geq 3f + 1$ for oral messages, and showed how signatures lift the bound.
⭐ Top Papers of the Week
- OpenForgeRL: Train Harness-native Agents in Any Environment — Xiao Yu, Baolin Peng, Ruize Xu et al. An open-source RL framework that trains agents inside real inference harnesses (Claude Code, Codex, OpenClaw) via a lightweight proxy + Kubernetes rollouts.
- Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context — Alagappan Valliappan. A training-free, lossless fix for speculative decoding at 1M-token context: sliding-window the MTP draft head only. 28–44% per-step speedup.
The full Slack digest covered roughly a dozen more papers across AI/ML, NLP, distributed computing, databases, OS, networking, and PL that were deliberately not expanded into blog notes. Only the highest-impact picks — plus one rotating seminal classic — become long-form posts.