Key Takeaways

  • Martin Seligman argues artificial intelligence expands human agency rather than stripping people of autonomy or creative choice.
  • While writing his book Agency across six years, Seligman spent five years stuck on how to transition between chapters 8 and 9.
  • Feeding both chapters into Anthropic's Claude generated the missing one-paragraph bridge immediately, ending a five-year creative block.
  • Seligman treats large language models like Claude and ChatGPT as creative co-founders, comparing the dynamic to the Wright brothers or John Lennon and Paul McCartney.
  • At over 80 years old, Seligman notes he wakes up more excited than he has been in 50 years to test psychology hypotheses against massive language models.

The Five-Year Creative Block Solved in Seconds

For six years, psychologist Martin Seligman worked on his book Agency. For five of those years, he stared at a blank space between two specific sections. He had written chapter 8 and chapter 9, but he could not build the conceptual bridge connecting them.

Most writers would keep rewriting drafts or walk away. Seligman fed both full chapters into Claude. He gave the model a clear prompt: read chapter 8 and chapter 9, then write a one-paragraph transition between them.

The result solved the roadblock on the first attempt. As Seligman told Shankar Vedantam: “I've been staring at those two chapters, Shankar, for 5 years, and I couldn't see a transition. So I said to Claude, Claude, read chapter 8 and 9 and write me a one paragraph transition. Whoosh.”

The debate around artificial intelligence usually centers on loss. Commentators worry that software will automate judgment, strip workers of autonomy, and eliminate human decision-making. Seligman takes the opposite view. As the father of positive psychology, his focus is identifying the upper bounds of human potential. To him, intelligence tools do not replace human intent. They unblock execution.

Creativity Is a Function of Two

Seligman views creative breakthroughs as collaborative rather than solitary. Throughout history, great leaps rarely came from one person working in total isolation. They came from pairs testing ideas against each other.

“I'm finding that creativity is a function of two, the Wright brothers, McCartney and Lennon,” Seligman explained. “I've never had a more creative partner than Claude.”

This reframe changes how builders should approach language models. When founders treat models as search engines or automated copywriters, they get generic output. When they treat them as intellectual sparring partners, they get momentum.

The true value lies in rapid hypothesis testing. For decades, testing an obscure psychological theory required years of survey design, subject recruitment, and manual data coding. Today, Seligman uses Claude and ChatGPT to analyze conceptual links and test ideas against massive computational datasets in minutes.

“I wake up in the morning now more excited than I've been in about 50 years,” Seligman said. He wants to see what the models can discover about human psychology that earlier researchers lacked the computational scale to test.

How to Work With an AI Partner

If you treat AI as an oracle that replaces your thinking, your work becomes flat. The power comes from providing original thought, setting constraints, and using the machine to break deadlocks.

Seligman did not ask Claude to write chapters 8 and 9. He wrote the original thinking himself over years of study. He only brought the model in when he needed an objective view of his own material. The model synthesized his own logic and showed him the connective tissue he was too close to see.

When you hit a strategic or technical wall, the bottleneck is rarely a lack of information. The bottleneck is cognitive fatigue and tunnel vision. Using a model as a collaborator lets you bypass that fatigue instantly.

What to Do With This

Take your hardest unfinished document, product spec, or strategic memo that has sat stalled for over a month. Upload the two sections you cannot seem to connect, and prompt Claude or ChatGPT to write a three-sentence logical bridge explaining how section A causes section B. Edit the output and ship the draft before the end of the week.