Grok 4.7 and the reliability of the agents around it

The poster praises Grok 4.7 but argues that the reliability of the systems around it is now the bottleneck. They say commenters wanted agents that finish work without stopping halfway, losing the browser, forgetting tasks or stalling unnoticed. They point to tools, memory, state, retries and errors as ways an agent can fail, and say they have been working on some of those issues.