Fetching from the wire…
Agents2026-07-26 · source-backed
The 0.7.0b1 (July 23) and 0.7.0b2 (July 24) rollup makes TodoListMiddleware opt-in, strips deprecated backend shims, adds filesystem delete and write_file overwrite, and trims tool descriptions for a roughly 43% drop in total description tokens on the default agent's schemas (GitHub). In parallel, deepagents-code 0.1.46 and 0.1.47 took Auto mode to GA with configurable goal-criteria acceptance, fewer redundant approval prompts, a "yolo" approval mode, and the recursion limit raised to 2000 and made adjustable. The framework is now competing on tokens-per-task instead of features, which is exactly what you'd expect given story two.
Each link below shares sources, entities, or timing with this story.
MIT-licensed, built on LangChain and LangGraph, v2.0 expanding from research-only to sandboxed filesystem, persistent memory, a skills system, and sub-agent orchestration with Slack/Telegram/Feishu integrations (GitHub). The pattern across every framework today is identical: m...
Per-node tracing declared in the graph rather than configured globally or wrapped by hand (GitHub), useful when one node handles PII or generates trace volume you don't want to store. Also a checkpoint fix for writes at plain-value seeds in delta channel history, and a cryptog...
Fundamental, sentiment, and technical analysts plus bull/bear researchers and risk managers over LangGraph, v0.2.5 adding GPT-5.5 coverage. Backtests claim 24.9% annual returns at a 5.60 Sharpe on AAPL/GOOGL/AMZN (GitHub). I'd take backtested Sharpe of 5.60 with heavy skeptici...
Sydney Runkle and Harrison Chase's July 22 retrospective draws a usable line: use a graph when you can encode real world-knowledge about stage order (classify → search → synthesize), don't when planning and delegation should emerge at runtime (LangChain). They cite GPT Researc...
Release notes almost never say "we removed a feature because the evals said it doesn't work." Deep Agents v0.7, landed July 29, says it twice. LangChain removed the base system prompt entirely. Trimmed tool descriptions by 43%. Net effect: base input tokens dropped roughly 65%...
LangChain, LangGraph, CrewAI, AutoGen, Microsoft Agent Framework, and Google ADK, presented at Black Hat (The Register). Insecure deserialization, SSRF, path traversal, use-after-free. That's the point: prompt-controlled content crosses into trusted framework logic, and then i...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.