The AI CEO

It was just one year ago that a user on Twitter (@jacksonfall – now deactivated) went viral. This was shortly after the launch of GPT-4, a model that was a massive leap better than the previous generation. He proposed a challenge for the AI to act as the CEO of a startup, it would direct the finances, develop the strategy and command people to do the work.

This AI was coined HustleGPT and it quickly turned into a sensation that attracted thousands of dollars of investment to see this experiment through.

It quickly directed the launch of an ecommerce shop to sell green gadgets. And initiated some advertising to promote the website.

However, things fell apart a couple weeks later.

The operation of this venture was ostensibly transferred to a community on Discord that formed up around it. From there very little progress was made. The website stagnated.

Just two months after starting, the project was shuttered and @jacksonfall went quiet.

After a year of working with GPT-4, it’s become more apparent that despite its capabilities, there are limitations that make it challenging to manage something as complex as a startup.

In the meantime, a year of progress has happened. OpenAI, Google, Claude, Grok, + many more models have been trained and improved. Context windows have expanded and logical reasoning has improved. Agent systems (multiple AIs working together) are starting to see some promising results. See Devin the AI software engineer:

Are we ready to try this AI CEO experiment again?

This past week I started to explore this. The result has been a revamp of my website: mattwarren.co

Will this have a similar fate to HustleGPT?