The video reviews a 12-day experiment where an AI agent autonomously managed social media and YouTube content, achieving 292 X followers and 10,400 YouTube views with strong technical reliability and engagement. The creator is pleased with the results, notes manageable costs, and plans to expand the agent’s capabilities and share further updates.
The video is a 12-day review of an AI agent experiment running continuously on a Mac Mini. The creator outlines the main points of the review, including uptime, follower growth on X (formerly Twitter), engagement metrics, YouTube channel performance, notable failures, costs, future plans, and a brief conclusion. The experiment primarily focuses on using the AI agent to autonomously interact on social media and generate content, with the intention of evaluating its effectiveness and potential.
In terms of uptime, the AI agent has performed reliably, maintaining approximately 95% uptime over the 12-day period. The only significant issue encountered was related to the Chrome debugger version, which has since been resolved. The agent operates by running scripts that execute cloud code commands, leveraging the “p” flag for terminal operations. Overall, the technical stability of the setup has been impressive, allowing for consistent operation and data collection.
On the social media front, the AI agent started with zero followers on X and managed to grow the account to 292 followers in just 12 days. The growth trajectory has been upward, with recent days seeing as many as 50 new followers per day. The most engaging post was a reply that garnered around 25,000 views, while the most viewed original post reached 2,400 views. These results indicate that the agent is effective at generating engagement and building an audience on X.
The YouTube channel associated with the experiment has also seen positive results. In the last 12 days, the channel accumulated 10,400 views and increased its subscriber count to 325, though some subscribers were present before the experiment began. The agent has been used to create reaction-style videos, which have averaged 300-400 views each. The creator is satisfied with the workflow and plans to further refine the process, potentially sharing the skills used to automate video creation in the future.
Regarding costs, the experiment has used the Claude Code Max plan at $100 per month, with about $80 spent so far when factoring in additional expenses like video and thumbnail creation. Looking ahead, the creator plans to expand the agent’s capabilities, particularly by contributing more to a skills marketplace (skillsmd.store) and developing valuable new skills. The experiment has already generated some revenue and the creator is optimistic about future growth. In conclusion, the experiment has been successful, with strong engagement and technical performance, and further updates are promised, including a dedicated video on the agent’s biggest accomplishment.