LangGraph

A new arXiv paper measures the corrective value of a failed tool call in an agent transcript and finds it negative on every small model tested: repeat probability rises from 0.06 to 0.54. Clean-context retry was the worst harness tested; paraphrasing the failure removed 76% of the effect for free.

Toyota Motor North America cut agent delivery from six months and six engineers to four days and one engineer. The saving came from amortizing security review and ingestion into a shared platform — which now leaves one permission gate holding back 50+ agents.

MasDrift ran 600 benign tasks through seven agent topologies. Supervisor hierarchies finished the most work and took unauthorized actions in up to 19.8% of tasks; flat peer networks, 0.6-0.8%. Most of the loss happens at the first handoff, when the lead restates the task.

Four agent orchestration products shut down, were superseded or repriced in twelve months. A six-criterion scorecard that weights exit cost at 30%, a six-week pilot with a portability test in week three, and the vendor landscape mapped to both.

Teams that outgrew LangChain usually misdiagnose why. Most of the pain is a durable-state problem that no framework swap will fix — and the rewrite costs more than the fix. Six exits compared, with prices checked 17 August 2026.

Microsoft's AutoGen went to maintenance mode and has not shipped a release since September 2025. A dated, sourced comparison of LangGraph, CrewAI and Microsoft Agent Framework on the things that actually break in production: durable state, retries, human approval gates, observability and the real bill.