The video begins with Dalton and Michael discussing feedback they've received since releasing their original "tarpit ideas" video about a year ago. Many viewers have approached Dalton asking if their specific startup idea is a tarpit.

"When some of the people ask me, 'Oh is my idea tarpit?' I'll be like, 'Hey, well have you talked to any users?' and they're like, 'No, I just... no. I thought you would tell me though.'" - Dalton"

Dalton notes that many people misinterpreted the concept, thinking a tarpit is simply "an idea that I don't like" rather than understanding the specific definition they provided in the original video.

Michael and Dalton revisit their original definition of a tarpit idea - ideas that consistently attract many founders but consistently fail to work. The term comes from how tar pits would trap dinosaurs: appearing like a water source from afar, dinosaurs would approach to drink, get stuck in the tar, and die.

"That's why we have all these dead founders in these tarpit pools, all these dinosaurs in tarpit pools." - Michael"

They clarify some key misconceptions:

• Tarpit ideas are not simply "bad ideas" - they're a specific subset
• Tarpits aren't a static list that never changes
• The idea space is constantly evolving with technology

Tarpit ideas share several defining characteristics that make them particularly dangerous for founders:

1. They're extraordinarily attractive and common - "If other founders don't come up with them all the time constantly, it is not a tarpit idea."

2. They receive unusually positive feedback from friends and peers.

"Tarpit ideas get like a lot of praise. Like the second your friend's like, 'That's a great idea!' Yeah, like a food discovery... You know that problem where you can't decide what restaurant to go to? That's a great idea!" - Dalton"

3. Many similar attempts have failed without any fundamental change to the underlying conditions.

"It's been done before a lot, and nothing has changed. Like there's no... I would argue that there are a number of ideas that at any given time in the last hundred years might have been tarpit, and then technology changed, and suddenly these are not only possible, they're great businesses." - Michael"

Since recording their original tarpit video, Dalton and Michael acknowledge that the emergence of LLMs and AI has significantly changed the landscape of viable startup ideas.

"I am open to even classic tarpit ideas no longer being tarpit. I'm open. What I love to see is when a founder can explain to me why the new technology might make something accomplishable that wasn't accomplishable before." - Michael"

What they dislike is when founders completely ignore the history of failed attempts:

"What I hate to see is a founder that's not even acknowledging that like many people have tried and failed before. It's like, 'Oh, history that happened before I'm alive didn't happen.'" - Michael"

The partners discuss a common pattern they're seeing with AI startups - the "X for Y" formula, particularly "Co-pilot for X" variations. While not necessarily tarpits (since not enough people have tried them yet), these ideas are becoming increasingly common:

"Everyone says co-pilot for X, and then like we could play Mad Libs. Co-pilot for real estate agents, co-pilot for investment bankers, co-pilot for lawyers, co-pilot for pilots, co-pilot for surgeons... co-pilot for dog walkers. Who would not want a co-pilot?" - Michael"

They note that while these aren't tarpits by definition, they are extremely common first-principle approaches to AI startup ideas.

Michael emphasizes that the current technology landscape determines which problems can be addressed at any given time:

"Whatever current set of technology is available can address only a current set of problems, and like some problems are going to fall outside of that set, and every time there's a new chunk of technology, we have to re-explore what problems are available." - Michael"

He shares that he's now funding ideas he would have rejected in the "pre-LLM world," showing how technology shifts can transform tarpits into viable opportunities.

Dalton discusses investing in AI-generated podcast content, which wouldn't have been viable before recent advances in LLMs and voice generation:

"I funded a company that's basically doing completely AI-generated podcasts, and it's like without LLMs and the ability to do voice generation, I would have never considered funding it." - Dalton"

They note that YC has been funding generative AI companies since 2016-2017, but the quality is now dramatically improving. They share an amusing anecdote about how many AI voice models are trained on their YouTube videos:

"They always train them on YouTube video, and guess who has a lot of YouTube video of them? And so all these new demos that we see, it's us in the videos. Quite embarrassing." - Michael"

Dalton relates how his wife mistook an AI-generated podcast featuring his voice for a real interview he had done.

Michael identifies two primary categories of tarpit ideas:

1. Utopian Social Change: Ideas that require everyone to simultaneously adopt a new behavior.

"Just a belief that the world should work differently than it does. Better. And this is all coming from a good place, but it's basically 'Wow, the way X works is bad, we should do it Y' and everyone simultaneously needs to adopt the scheme." - Michael"

He gives the example of apps for coordinating social activities with friends - while appealing in theory, they consistently fail because changing human behavior at scale is extremely difficult.

2. Arbitrage/Get-Rich-Quick: Ideas based on exploiting temporary market conditions or trends.

"Some form of arbitrage or getting rich very quickly. There's like a limited window where I have some secret knowledge and I can get very rich very fast... I can kind of like fast forward through the entire startup thing and kill it." - Dalton"

They mention examples influenced by Wall Street Bets during the pandemic, where people wanted to quickly capitalize on market movements.

Concepts:

• Tarpit Ideas - Startup ideas that consistently attract founders but consistently fail
• Wall Street Bets - Reddit community mentioned as influencing startup ideas during the pandemic
• LLMs/AI - Large Language Models and Artificial Intelligence that are changing the viability of certain ideas
• Co-pilot for X - Common pattern for AI startup ideas following GitHub Copilot's success

Tools/Products:

• Pocket Pod - AI-generated personalized podcast mentioned by Dalton that his wife mistook for NPR

• Tarpit ideas are not just "bad ideas" but a specific subset that are highly attractive yet consistently fail
• Technology changes can transform former tarpits into viable businesses (AI/LLMs are doing this now)
• True tarpits share key characteristics: they're common, receive positive feedback, and have a history of failures
• The "X for Y" formula (especially "Co-pilot for X") is becoming increasingly common in AI startups
• Two main tarpit categories: ideas requiring mass behavior change and get-rich-quick arbitrage schemes
• Founders should research historical attempts at their idea and explain why technology changes make success possible now
• What sounds too good to be true and gets too much positive feedback often is a warning sign
• Actually talking to users and getting customers remains crucial, regardless of how appealing the idea sounds