How to Find an Open Research Problem (Not a Crowded One)
The best research problems are already written down — in the limitations sections nobody reads. How to find an open gap instead of a crowded one.
Ask most early researchers where problems come from and they'll say "inspiration." Ask productive ones and they'll tell you something more boring and more useful: the best open problems are already written down, usually in the parts of papers nobody reads.
Here's how to find a gap worth working on instead of piling onto a crowded one.
Crowded vs. open
A crowded problem is one where the hot result just landed and five hundred people are now racing to extend it. It feels exciting and safe — everyone agrees it matters. It is also the worst place to be a newcomer: you're competing with established labs who saw it first, and by the time you publish, the frontier has moved.
An open problem is one where a real limitation exists, a specific question is genuinely unanswered, and not many people are looking yet. Less validation, far better odds. The trick is that open problems rarely announce themselves — you have to go find where they're quietly stated.
Where gaps actually hide
Not in abstracts (those sell the result) and not in introductions (those motivate it). Gaps live in three places most people skim past:
- Limitations sections. "Our method assumes X," "we do not handle Y," "performance degrades when Z." Every one of those is a stated open problem, written by the person closest to it.
- Future-work paragraphs. "A natural next step is…", "it remains open whether…", "we leave the general case to future work." These are invitations, published on purpose.
- The gap between two adjacent papers. Paper A solves a problem in one setting; paper B solves a related one elsewhere; nobody has connected them. The bridge is often the thesis.
Train yourself to read *for* these. A paper's most valuable sentence for you is usually the one where the authors admit what they couldn't do.
Read the frontier, not the textbook
Textbooks and highly-cited classics tell you what's *settled*. To find what's open you need the last few years of work in a narrow area — the papers actively wrestling with the current edge. That means reading recent, not canonical: the 2024–2025 papers where limitations are fresh and future-work sections point at problems no one has taken yet.
This is also why "find an open problem" and "find the right advisor" are the same search. The person whose recent paper states the gap you want to work on is, very often, exactly the person you should be working with.
Ground your idea in a stated gap, not a guess
The failure mode here is inventing a problem that *sounds* open but is actually solved, ill-posed, or a known dead end. The defense is grounding: anchor your candidate problem to a specific, real paper that explicitly states the gap — a verbatim "this remains open" you can point to — not to a vague sense that "surely no one has done this." If you can't find a real source that states the gap, treat that as a warning, not a green light.
This is the idea behind the "where it goes" view in Siddhanta: it takes your work, finds the adjacent research frontier and the real researchers already on it, and points you at where your work could extend — grounded in their actual recent papers, never a made-up direction. It won't hand you a fabricated "novel problem" (the most dangerous thing an AI can do here); it surfaces the gaps real papers have already stated, near your work.
The habit to build: when you read a paper, don't stop at the result. Find the sentence where the authors tell you what they left undone — and ask whether that undone thing is yours.