Fried says agents let him build a tool he never could have
The blocker he names is not money or talent on staff, it is the cost of asking a colleague to go build a side idea.
Jason Fried was asked a direct question by @fagnersales25, whether agents are simply the thing that made the cost of experimenting low enough to finally try. Fried's answer was that he could not have built the tool himself in a reasonable amount of time without learning a bunch of new stuff, and that he could not have asked anyone else to build it for him because they had more important stuff to do.
I've always had the vision. Never the time or talent. Now I have it all.
That second half is the interesting one. The first constraint is the familiar one, a person with ideas who does not have the specific skills and does not want to spend months acquiring them. The second is organisational. Inside a company with real work in flight, a side idea has to win an argument against whatever the team is already shipping, and most side ideas lose that argument before anyone writes a line of code. Agents do not make the idea better. They just mean nobody has to be reassigned to find out.
Fried does not put a number on any of this. No build time, no cost, no description of which agent or model he used, and no claim about what the tool earns or whether it ships to anyone else. Fagner Sales said he hoped Fried would open source it. There is no answer on that in the sources.
The read
This is one founder describing one build, and it is a feeling rather than a benchmark. But it sits alongside what Fried said earlier about being able to build and design software alone again, and it sharpens the claim. The unlock he is describing is not output speed. It is permission. The internal cost of trying something speculative used to be another person's week, and now it is not.
If you run a team, that reframes the backlog. The ideas you killed were often killed on opportunity cost rather than merit, and that filter just got a lot weaker.