This is a classic case of a simple workflow being held hostage by a rigid tool. The user's instinct to sort by date and grab the top file is exactly right, but the spreadsheet software then undermines that logic by hardcoding the file name in the next step. That tension, between a dynamic filter and a static reference, is the real problem, and it's one that a modern, AI-native spreadsheet should solve without requiring a workaround.
What this user needs is a system that treats the folder as a live source, not a one-time import. If the connector can dynamically resolve the latest file based on metadata like date added or timestamp, the subsequent query should inherit that reference automatically. The user has already done the hard part: they identified the correct subfolder and established a sorting rule. The failure point is the tool's inability to carry that logic forward. A smarter spreadsheet would let them define a "latest file" variable once and use it across all downstream steps, so the timestamp in the filename becomes irrelevant.
The naming convention itself, "export yyyy-mm-dd hh-mm-ss", is a gift. It's machine-readable, deterministic, and perfectly suited for pattern matching. Instead of fighting it, the solution should embrace it. The approach is straightforward: use the folder connector to list all files, filter by the subfolder, then extract the timestamp from the filename to sort and select the most recent. The file name should never appear as a hardcoded string; it should be a reference to the top row of that filtered list. Any query that breaks when a new file arrives is not a pipeline, it's a fragile script waiting to fail.
The practical takeaway here is that data pipelines don't have to be brittle. The user's scenario, twice-daily snapshots with a clear naming convention, is exactly the kind of repeatable process that should run on autopilot. The barrier is not the complexity of the task but the inflexibility of the tool. If your spreadsheet can't handle a simple "always use the most recent file" instruction, it's time to explore a platform that treats data as a living stream, not a static import. The solution is not a more complex workaround; it's a tool that respects the logic you already wrote.