Theo-brainUnofficial. Not affiliated with Theo or t3.gg.
Sign inYou

Smart main model delegating to cheaper sub-agents

Technique

Keep a smart model like Opus as the main agent and let it hand small jobs to cheaper, faster sub-agents (helper agents it launches itself), such as Haiku 5.5 or Sonnet 5.5. The main model plans and directs; the helpers do the many small attempts or investigations in parallel.

Theo's current stance

Theo favors this setup. In Anthropic's demo, Opus with 10 Haiku helpers made 86 attempts for $0.14 versus 25 attempts for $0.47 alone, and he used it for his own PR audit rerun.

Try it

  1. Run the main task on Opus and tell it to use lots of sub-agents, with Haiku 5.5 for all of them.
  2. Tell it to do little itself beyond getting the data and directing the sub-agents.
  3. Remove any rule in your CLAUDE.md that forces all sub-agents to be Opus.
  4. Compare task time and cost before and after the change.

First seen

30:02
But if you have it as a tool, Opus calls to do things like deep dives into your codebase to find specific behaviors or characteristics or confirming a hunch it has about some thing that you found in an API or all of those types of investigatory things. It is really cheap

OpenAI should be scared of this one

Every mention

23:06
only designed 25 attempts and it cost $0.47 to do. But with Opus 5.5 commanding 10 Haiku instances, it was able to do it in under a minute with 86 attempts. Way more attempts because it's able to get more renditions out when it's that much cheaper and faster and only cost $0.14. So, for the types of

finally a good small model

35:14
First, I told it to use lots of sub-agents and workflows to review these PRs and to use Haiku 55 for all sub-agents throughout the work. To say it in one more sentence, minimize work that you do yourself outside of actually getting the data and orchestrating the

finally a good small model

30:02
But if you have it as a tool, Opus calls to do things like deep dives into your codebase to find specific behaviors or characteristics or confirming a hunch it has about some thing that you found in an API or all of those types of investigatory things. It is really cheap

OpenAI should be scared of this one

30:17
and surprisingly fast, too. I would bet that if you set Opus up properly to use Sonnet at the right times for these sub agents that the result will be Opus feeling way faster and getting real work done for slightly cheaper. That sounds

OpenAI should be scared of this one

Theo Daily is unofficial and not affiliated with Theo or t3.gg. Written by Claude Opus 5.5 from each video's transcript.