Today's take
Agent engineering toolchain matures with morluto/rea leading at 2,613 stars/day; model side sees 13% growth in DeepSeek V4.1 Flash 1017, 92% surge in Qwen3.8 27B; application side has Claude Code at 1.86T tokens.
Open source
Agent engineering toolchain continues to mature, with morluto/rea leading at 2,613 stars/day. Code analysis toolchain sees GoogleChrome/modern-web-guidance maintaining technical leadership through latest web practice updates. Model access tools show ebpala/sh-Go improving usability through local execution.
#1 morluto/rea https://github.com/morluto/rea 2,613 stars in 1 day. Reverse engineering tool enables agents to deconstruct app behavior to native code, driving toolchain engineering.
#10 mattpocock/skills https://github.com/mattpocock/skills 1,023 stars in 1 day. Engineer skill library codifies agent behaviors from real development scenarios, accelerating engineering adoption.
#19 h4ckf0r0day/obscura https://github.com/h4ckf0r0day/obscura 707 stars in 1 day. Brainstorming browser tool enhances agent web interaction capabilities, reflecting growing demand for engineering tools.
Models
Top 5 models show 12% overall growth, with DeepSeek V4.1 Flash 1017 leading at 9.13T tokens. Qwen3.8 27B sees 92% surge in usage, indicating rapid adoption of emerging models, while Nemotron 3.5 Super VL shows 10% decline reflecting user preference for more efficient models.
- DeepSeek V4.1 Flash 1017 9.13T tokens (+13%)
- Qwen3.8 27B 170B tokens (+92%)
- GPT-6 Luna 1.03T tokens (+18%)
- GLM 5.3 Flash 2.48T tokens (+30%)
Apps
CLI agents still dominate with Claude Code at 1.86T tokens. Portkey AI sees 15.8M requests growth indicating rising developer tool demand.
- Claude Code 1.86T tokens (CLI agent)
- Hermes Agent 1.75T tokens (Personal agent)
- Codex 838B tokens (+10.8M requests)
- Portkey AI 334B tokens (Programming app)
What to watch
The maturing agent engineering toolchain and surging model usage form a positive feedback loop. Watch how engineering tools further lower model accessibility barriers. Emerging models' rapid adoption may trigger new technological iterations.
