Agentic AI Loops: The Next Hype Cycle or a Real Breakthrough?
The recent @Scale conference hosted by Meta saw a surprise question from the audience about loops in AI development, to which Claude Code creator Boris Cherny responded with an emphatic “yes, they’re for real.” Cherny’s vision for the future of AI involves agents prompting other agents to write code, creating a continuous loop of improvement. This mirrors the shift from manual source code writing to agent-driven code development, which has been a significant step forward in the industry.
Cherny’s own work involves running multiple agents in loops, each with a specific task, such as improving code architecture or unifying duplicated abstractions. These agents submit pull requests like human coders, and since the code is constantly changing, they never stop running. This approach is a powerful idea, particularly with a prominent figure like Cherny behind it. However, it’s essential to recognize that this concept isn’t entirely new, and its implications need to be carefully considered.
The focus on managing agents and establishing clear goals has been a primary concern for users in the shift to agentic AI. Loops take it a step further by authorizing a swarm of agents to work continuously in the background, endlessly. While this approach has the potential to revolutionize AI development, it also raises concerns about trust and control. As models improve rapidly, it’s crucial to weigh the benefits against the costs and potential risks.
The Mechanics of Agentic AI Loops: Trust, Compute, and Oversight
Agentic AI loops involve a non-deterministic logic, where a sub-agent chooses when to stop the loop instead of a clear condition. This approach is different from classic computing, where recursive loops follow a deterministic logic. The Ralph Loop, a popular trick in agentic AI, sums up all the work that the model has done and asks if it’s accomplished its goal. This approach helps deal with AI models getting lost as they run for too long.
The push for more test-time compute is another key aspect of agentic AI loops. OpenAI researcher Noam Brown observed that contemporary models can solve nearly any problem if you throw enough compute at them. This means that one way to ensure a problem gets solved is to just keep throwing compute at it until it’s finished. However, this approach can be expensive, and the benefits need to be weighed against the costs.
Agentic AI loops burn through tokens faster than simple Q&A chatbots, and because the point is to keep the loop running all the time, there’s no ceiling to how much you can spend. This raises concerns about token spend, drift, and other classic AI issues. While the benefits of agentic AI loops could be staggering, it’s essential to consider the potential risks and develop strategies for oversight and control.
Winners, Losers, and Disrupted Parties in the Agentic AI Loop Market
Companies like Anthropic, which is in the token-selling business, may benefit from the increased demand for compute and tokens. However, for other companies, the costs of implementing agentic AI loops could be prohibitively expensive. The benefits of this approach will depend on the specific problem being solved and the right setup that allows for oversight of token spend, drift, and other classic AI issues.
Adjacent markets, such as cloud computing and data storage, may also be affected by the growth of agentic AI loops. The increased demand for compute and storage could lead to new opportunities for companies in these markets. However, it’s essential to consider the potential risks and develop strategies for mitigating them.
The shift to agentic AI loops could also disrupt traditional coding practices, as agents take on more responsibility for writing and improving code. This could lead to new opportunities for developers who are skilled in working with agentic AI systems.
The Skeptical Case: Will Agentic AI Loops Live Up to the Hype?
While agentic AI loops have the potential to revolutionize AI development, it’s essential to consider the potential risks and limitations. One of the main concerns is the trust and control issue, as agents take on more responsibility for writing and improving code. If not properly managed, this could lead to unintended consequences and errors.
Another concern is the cost and scalability of agentic AI loops. While companies like Anthropic may benefit from the increased demand for tokens, other companies may find the costs prohibitively expensive. It’s essential to develop strategies for mitigating these risks and ensuring that the benefits of agentic AI loops are realized.
The Signal to Watch Next: Earnings Calls and Regulatory Decisions
The next verifiable event that will confirm or disprove the thesis of this article is the earnings calls of companies like Anthropic and other players in the agentic AI loop market. These calls will provide insight into the revenue and profitability of these companies and the growth of the market as a whole.
Regulatory decisions will also play a crucial role in shaping the future of agentic AI loops. Governments and regulatory bodies will need to develop strategies for mitigating the risks and ensuring that the benefits of this technology are realized. The outcome of these decisions will have a significant impact on the market and the future of agentic AI loops.
Pick one tactic from this post and apply it today. Which one will you start with?
By Daniel Cross, Digital Growth Strategist at TrendFlashy
Ready to launch your own asset?
Check out our guide on Building a Profitable Online Business.
