OpenEnv in Practice: Evaluating Tool-Using Agents in Real-World Environments

kept by eddie feb 13

OpenEnv framework from Meta and Hugging Face reveals tool-using agents fail at multi-step reasoning, ambiguity resolution, and execution quality in realistic calendar environments.

The Code-Only Agent
An agent limited to only executing code can still accomplish complex tasks by creating and improving its own CLI tools.
also on Hugging Face, MCP, #ai-agents, #open-source, #developer-tools
8 plots that explain the state of open models
Qwen dominates open-model downloads and finetunes globally, while DeepSeek leads in large-scale models and GPT-OSS is the only Western contender gaining adoption.
also on Hugging Face, #benchmarks, Meta, #open-source, #developer-tools
How OpenAI Codex Works - ByteByteGo Newsletter
OpenAI's Codex agent relies on an orchestration layer handling prompt assembly, context management, and a custom protocol for multi-surface support.
also on MCP, #ai-agents, #open-source, #software-engineering, #developer-tools
Posts tagged "Agents Week" — Cloudflare Blog
Cloudflare's Agents Week highlights tools like AI Search, Kitesurf browser, and MCP v2, advancing agent-first infrastructure with continuous trust evaluation and open platforms for AI-driven work and data exploration.
also on MCP, #ai-agents, #open-source, #developer-tools