# Dalton + Michael: How To Get Unique AI Startup Ideas Status: deepened milieu note ## Source - Date captured: 2026-06-18 - Source published date: 2026-03-16 - Source type: `video` - Source title: `How To Get Unique AI Startup Ideas` - Source URL: - Source show / channel / publication: Dalton + Michael - Platform: YouTube - Local source file: `external_material/archive/processed/https__youtube.com_watch_v=Eht6XjI-p64&is=aJ6iaLJ6vGgF0Dxw.txt` - Local transcript: `external_material/transcripts/20260316-Eht6XjI-p64.en.txt` ## People / Organizations - Primary speakers: Dalton Caldwell, Michael Seibel - Organization / context: YC / startup ideation - Retrieval names: Dalton Caldwell, Michael Seibel, YC, unique startup ideas, AI startup ideas, non-consensus ideas ## Neutral Summary Dalton and Michael argue that in an AI era where building has become easier, the bottleneck shifts toward finding differentiated ideas. Their core warning is that many founders are copying what recently raised money, what appears in YC batches, or what is currently popular in VC/startup media. That process creates derivative ideas. They recommend moving away from consensus: - do not treat recently funded startups as validation; - look in the "discard bin" for ideas others saw and rejected; - be willing to work on things that sound strange, small, or bad to friends and investors; - solve your own problem or build something you would have wanted; - recognize that weird or non-conforming people often have an easier time producing weird, original ideas; - understand that not everyone needs to be the person who originates the idea; some people should join or work with the person who has the strange idea. The strongest thread is that unique startup ideas often begin as non-consensus, personally felt, or discarded ideas, not as extrapolations from current investor fashion. ## Why This Caught Attention This source fits the project's "what to build" side. It directly addresses the post-agentic-coding problem: when more people can build, idea selection, judgment, conviction, and taste become more important. ## How Theme Theory Relates Theme Theory can add structure to the problem of finding ideas. The video says founders should avoid consensus and find ideas they can believe in. Theme Theory can ask what kind of audience-side object makes an idea worth building around. A Theme Theory version of the question would be: ```text What desired real-life state is underserved, meaningful, and supportable enough that a builder can organize product, creative, and audience around it? ``` The "discard bin" idea is especially relevant. Some objects of interest may look too specific, too weird, too non-consensus, or too hard to explain at the surface level. But if they represent a meaningful higher-order state that some audience strongly wants, they may be fertile places to build. This also connects to the role of taste. The founder who is solving their own problem may have direct access to the audience's desired state because they are part of the audience. That can make them better at judging what matters than a founder triangulating from investor signals. ## Deep Corpus Comparison This source is directly aligned with the WTB track. The video says differentiated ideas are often weird, non-consensus, personally felt, or sitting in the discard bin. Theme Theory can explain one reason: ```text a strong object of interest may be illegible to outsiders before it is framed as a meaningful desired state. ``` That is different from contrarianism for its own sake. A weird idea is not good because it is weird. It is good if the weirdness comes from access to an underserved, under-described, or hard-to-see audience-side state. This gives a future `identify your theme` / WTB heuristic: - start from lived problems or unusual taste; - ask what desired state is missing or fragile; - test whether the state is meaningful, supportable, and durable; - then decide whether software, media, services, or AI can support it. The source also validates the user's instinct that TT should help judge concepts, not merely list them. The discard bin contains both overlooked gold and bad ideas. The object-of-interest frame is a way to separate them. ## Core Links - [What This Is](../../core/what-this-is.md) - [Creators, Builders, And Audience](../../core/creators-builders-and-audience.md) - [Object Of Interest](../../core/object-of-interest.md) ## Candidate Concepts / Edges - unique startup idea -> non-consensus object of interest - solve your own problem -> builder is also audience member - discard bin -> overlooked desired state or hard-to-serve state - weird founder -> unusual access to unusual objects of interest - AI makes building easier -> idea selection and taste become bottlenecks ## Promotion Judgment - Promote to core? `maybe` - Reason: strong fit for the future `identify your theme` / `what to build` docs. This should probably be revisited when those core docs are drafted. ## Open Questions - How should Theme Theory distinguish a merely weird idea from a strong non-consensus object of interest? - Does "solve your own problem" become one path into identifying a theme? - Can the theory help evaluate discarded ideas without sanding off what makes them non-consensus?