15 pages

entities

People, organizations, projects, products, places. anthropic 5 AI lab; vendor of Claude + Claude Code; first-party loop/effort guidance. chroma 1 vector-DB company; Context-1 trains `prune_chunks` as a learned retrieval-side policy. claude-code 5 Anthropic's CLI / agent harness; loop primitives + model/effort dials. codex 3 OpenAI's coding agent; "no manual code" experiment; compact() internals; Windows sandbox. cursor 2 AI-first IDE; Composer 2 real-time RL + summary-as-graded-subtask; file-backed MCP A/B test. deepseek 1 Chinese AI lab; V3 stability + FP8, R1 four-stage post-training, GRPO origin, R1-Distill series. kaku 1 Tw93's open-source terminal (Rust + Lua); recurring case study for Skills and Hooks. lilian-weng 2 author of *Lil'Log*; wrote the 2023 [[llm-agent]] survey and the Jul 2026 RSI survey. moonshot-ai 1 Chinese AI lab (Kimi); Kimi K2.5 PARL — orchestrator-only Agent training, annealed `r_parallel`. openai 7 AI lab; InstructGPT, o1, RFT, Deliberative Alignment, Codex harness engineering. openclaw 1 Peter Steinberger's self-hosted personal-AI-assistant daemon; case study in Tw93's Agent essay. peter-steinberger 1 PSPDFKit founder; creator of [[openclaw]]. ryan-lopopolo 1 OpenAI engineer; author of [[2026-04-27-harness-engineering-codex-agent-first]]. shunyu-yao 2 OpenAI researcher; CALM → ReAct → tau-bench → *The Second Half*; named the utility problem. tw93 3 Alibaba/Fliggy engineer; author of all three seed essays and of [[kaku]].
esc