A lab meeting ends the way too many do: three senior researchers pushing three directions, each sure theirs is the obvious next project. Nobody has the numbers to settle it, so the argument loops until someone gives in. WisPaper, an AI academic agent, can surface what the literature supports, but the harder problem comes first.
Choosing a Direction Is the Hard Part
Most research groups don't lack ideas; they lack a way to rank them. The whiteboard fills with candidate problems, each with a champion and a plausible story. Deciding means estimating how crowded a subfield is and whether the methods exist. Tools like WisPaper can supply part of that estimate, but most of it is made on feel, which a lab meeting argues about.
Following the Crowd
When a group can't decide, the default is to drift toward whatever is fashionable: a trendy benchmark, a hot architecture, a dataset everyone else is using. This feels safe because reviewers know the name, but it lands the work in the middle of a crowd. Junior researchers feel it most; they inherit a project chosen by inertia and spend two years discovering the field moved on.
A Real Gap or a Dead End
The distinction that matters is between a gap open because the method only recently became possible, and a gap empty because nobody cared. Both look identical from a distance: no papers, no citations, silence. One is an opportunity; the other is a hole with a reason. Telling them apart means reading what people tried and abandoned, rarely written down.
Mapping What's Already Been Done
Some AI tools now do the legwork of that distinction. WisPaper can scan indexed work to flag where a subfield is thin and where past attempts stalled. It can also map which questions were answered and which were tried then dropped, turning that into a survey with references that trace to real papers. That is AI academic agent work: mapping the terrain before anyone commits to digging.
A Question Worth Answering
When a group can see the difference between an open gap and a dead end, the lab meeting changes. The argument becomes about which question has real, unmapped room, and a direction gets chosen on the evidence.
The best research gap is the one that was invisible until someone could see the whole field at once. As tools get better at drawing that map, more energy goes into answering questions instead of guessing which are real.