Skip to content
Dalton + MichaelDalton + Michael

How To Get Unique AI Startup Ideas

One of the biggest problems founders (and builders in general) have right now is not "how do I build my MVP?" but instead "what idea should I be building?" It feels like all of the good ideas are taken, and the competition for those ideas is ruthless. In this episode of Dalton + Michael, the two discuss some of the starting conditions for coming up with unique startup ideas, and why it can feel so hard to come up with something that is both: 1) something people want and 2) original. They suggest some tactics you consider trying that are more likely to yield differentiated startup ideas than derivations of whatever The Current Thing is. Dalton + Michael is brought to you by @Standard_Cap Dalton Caldwell on X: https://x.com/daltonc Michael Seibel on X: https://x.com/mwseibel

Dalton CaldwellhostMichael Seibelhost
Mar 15, 202616mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

How to find non-consensus AI startup ideas that actually win

  1. They argue that today’s AI tools make building easy, so the real bottleneck is generating genuinely non-derivative ideas rather than copying what’s trending on Twitter or recently funded.
  2. They warn that using YC batches, VC content, or recent funding rounds as ‘market validation’ is misleading because funding decisions often reflect teams and pivots—not a pre-validated idea.
  3. They recommend looking in the ‘discard bin’: pursue ideas many people have already considered and rejected, since uniqueness often comes from embracing what others filtered out.
  4. They critique the obsession with $100B+ outcomes as an ideation filter that eliminates quirky but ultimately massive companies like Twitch or Whatnot would have looked early on.
  5. They emphasize that “weird” nonconformists generate novel ideas more naturally, and founders should be ready to endure years of skepticism—or partner with someone who has that ideation strength.

IDEAS WORTH REMEMBERING

5 ideas

Treat idea generation—not building—as the core constraint now.

With AI and modern tooling, execution speed has increased, so the differentiator shifts to having a strong, unusual thesis that isn’t instantly replicable by everyone watching the same trends.

Don’t pick startup ideas by copying what just got funded.

They argue this is a broken proxy: companies often get funded for founders, timing, or a direction that changes via pivot, so copying the visible ‘idea’ is chasing noise, not signal.

Use the ‘discard bin’ to find uniqueness.

Instead of searching for something no one has thought of, look for ideas many people have considered and rejected—then ask what assumption made them discard it and whether that assumption is now wrong.

Actively move away from consensus—friends, investors, and founder-Twitter included.

If everyone around you thinks an idea is obviously good, it’s likely already crowded; controversy or concerned reactions can be a sign you’re exploring a non-consensus space with asymmetric upside.

Consider ideas that take longer than two years to work.

A simple way to ‘cheat’ toward non-consensus is to pursue projects with longer timelines; many founders self-censor to short horizons, which compresses them into the same safe, derivative ideas.

WORDS WORTH SAVING

5 quotes

Instead of trying to find something no one's ever thought of, find something that everyone's thought of and thinks is bad.

— Dalton Caldwell

You're gonna decide on what you're gonna work on for maybe the next 10 to 20 years of your life by looking at a list of what other people are working on?

— Michael Seibel

If you're making all of your life decisions based on podcasts, that is the definition of coming up with ideas that are not unique.

— Dalton Caldwell

Many successful YC companies, every one of their friends and family members thought the idea was bad.

— Michael Seibel

Don't be approval seeking from authority figures if you wanna actually have unique ideas.

— Dalton Caldwell

The ideation bottleneck in the AI eraCloning trends vs original conviction‘Discard bin’ idea selectionFalse signals from funding and YC batchesConsensus filters and VC content effectsSolving your own problem vs TAM obsessionWeirdness, nonconformity, and long time horizons

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome