Today's take
Agent engineering frameworks continue to strengthen, with knowledge infrastructure and CLI toolchains growing in parallel. Top 5 model usage apps are all code-related tools. Open source projects show breakthroughs in skill frameworks and testing tools, corroborating the growth of CLI toolchains.
Open source
Agent engineering frameworks enter a fast development phase, with modular skills and improved testing systems as core trends. Open source projects significantly enhance agent development efficiency through the construction of reusable skill libraries and standardized testing processes.
#1 obra/superpowers https://github.com/obra/superpowers 321,627 stars. Signal: The agentic skills framework jumps to the top, reflecting the demand for engineering capability accumulation.
#2 morluto/rea https://github.com/morluto/rea 469,077 stars. Signal: Reverse engineering tools continue to grow, showing the trend of agent toolchains penetrating deeper.
#15 firecrawl/firecrawl https://github.com/firecrawl/firecrawl 203,498 total stars. Signal: Knowledge infrastructure gains recognition, providing structured data support for agents.
Models
Top 5 models show DeepSeek V4.1 Flash leading, with Kimi K3.1 Pro usage increasing by 50%. Emerging model GPT-6.2 Sol sees a 456% increase in usage, indicating accelerated model iteration.
- DeepSeek V4.1 Flash 1017 (batch): 9.44T tokens, +6% daily growth
- Kimi K3.1 Pro (batch): 664B tokens, +50% daily growth
- GPT-6.2 Sol (batch): 78.3B tokens, +456% daily growth
- Qwen3.6 35B A3B: 80.2B tokens, +74% daily growth
Apps
CLI toolchains occupy the top 5, with Hermes Agent leading at 2.29T tokens. Portkey AI shows significant usage growth, indicating the vitality of the development tool ecosystem.
- Hermes Agent: 2.29T tokens, continuously growing personal agent
- Claude Code: 1.83T tokens, representative code-related CLI tool
- Strix: 697B tokens, security vulnerability detection tool
- Portkey AI: 428B tokens, programming application control panel
What to watch
The synchronized improvement of agent engineering and model iteration speed indicates that AI development is entering a new stage of standardization and toolization. It is recommended to pay attention to the combination development of skill modular frameworks and knowledge infrastructure.
