Two weeks ago, I wrote that falling AI costs are making agents more economical. This week, we are seeing what that efficiency has unlocked. Claude Code (and the user-friendly desktop application) introduced a new “loop” feature that lets an AI keep checking on a task and come back to it later without being manually prompted each time.
That is very similar to what the viral open-source project OpenClaw was trying to do with its cron feature. In both cases, the idea is simple: Let the AI wake itself back up later and keep going. This simple shift has unlocked a lot of creative use cases. And the more AI can keep working without constant human nudging, the more useful it becomes for real business workflows.
For now, Claude’s “loop” capabilities are limited compared to OpenClaw, likely to avoid a spike in inference costs on Anthropic’s end. But it is a step in the same direction as the open-source project, and it’s not too hard to see how we will soon have decent agents running around the clock.
Here’s a real-world example of how lower token costs and stronger agentic capabilities are making a difference. One of Applied.AI’s clients, a specialist news organization, regularly reviews a large volume of text from a few state-run newspapers. An analyst manually extracts data from this text and inputs it into the client’s backend. But with a very large context window (and strong retrieval capabilities), an entire week’s worth of text can be processed in a single pass, allowing a language model to piece together additional datapoints that a one-by-one analysis would miss.
And since tokens are getting cheaper, Applied.AI could rapidly backtests on four months of historical data and compare the process’ performance against the human baseline, iterating the prompts and other parameters with little worry about costs spiraling out of control.
It’s not hard to imagine how a Claude “loop”, or a well-designed agentic system on a regular schedule, could also review a set of primary sources and write spot news on the fly, too.
Put together, these trends point in the same direction. AI is getting cheaper, easier to keep running, and better at handling more complex work. That combination is what will drive the next wave of adoption. The important story is not that each individual token costs less. It is that businesses will find more and more reasons to use a lot more of them.
If you found this newsletter useful, forward it to someone who would too. Click here to learn how Applied.AI can help your business make AI work for you.
