Unfortunately, it’s just not possible to keep control.
Already in the software industry, most devs don’t know what code is being written. They’re trusting their agents to explain it to them.
People should always be in control of the future.
OpenAI is against the notion that it would be acceptable for people to be relegated to a role of doing things that “AI systems think makes sense to reward in humans”.
People should not cede control of the future to
Every agent harness will soon run in the cloud and have access to any machine. The agent loop won’t run locally anymore.
Claude Code will migrate to this architecture in 2027, built on top of Claude Managed Agents. No more jsonl files on your disk, they will own the agent state
