The economy does not need a real AGI: CEO of Replit

Forget building god-level superintelligence. Replit CEO Amjad Masad said we don’t need this to change the economy and society – we just need “functional AGI”.

The vibe coding startup’s CEO said in an episode of the “a16z” podcast released Thursday that while Silicon Valley is obsessed with real AGI, the practical version is already within reach and is enough to automate huge swaths of the economy.

“We can like functional AGI,” Masad said. He defines functional AGI as AI that does not require human consciousness or reasoning, but simply systems that can learn from real-world data and perform verifiable tasks on their own.

“We will target every economic sector and so you can automate a lot of the work,” he said. “We’re on that path, that’s for sure.”

Masad said he is not convinced we will ever achieve true AGI, the type of artificial intelligence that can learn and adapt to all areas of knowledge, like the human mind.

While real AGI could “propel us to the next level of human civilization,” Masad says he is “pessimistic that AGI will truly advance because what we have built is very useful and economically valuable.”

Masad also said the industry could be stuck in a “maximum local trap,” meaning AI companies are optimizing what already works instead of reinventing the field. By looking for small, cost-effective improvements to today’s big models, they might miss the path to a real breakthrough.

“Maybe the general problem does not arise in our lifetime,” Masad said, referring to solving the problem of general intelligence itself. “Who knows?”

Masad did not respond to a request for comment from Business Insider.

The AGI dream is losing its luster

Masad’s comments come amid renewed debate over whether AGI is a meaningful goal.

Big AI labs still view AGI as the ultimate prize. OpenAI, Google, Meta and Microsoft have dedicated their best researchers to the same goal.

However, some experts question whether LLMs will ever evolve into true general intelligence.

“No one with integrity should believe that pure scaling will get us to AGI,” AI leader and best-selling author Gary Marcus wrote in a blog post in August. “Even some tech professionals are waking up to the fact that ‘AGI in 2027’ was marketing, not reality.”

OpenAI’s GPT-5 release also failed to live up to the AGI hype.

“It’s clearly a generally intelligent model, although I think in the way most of us define AGI, we’re still missing something pretty important, or a lot of things pretty important,” Sam Altman, CEO of OpenAI, told reporters on a press call in August before the release of GPT-5.

Other industry veterans echoed that caution. Meta’s chief AI scientist, Yann LeCun, said we could still be “decades away” from reaching AGI.

“Most interesting problems scale extremely poorly,” LeCun said at the National University of Singapore in April. “You can’t just assume that more data and more computing means smarter AI.”