Source details
- Original source
- Allie K. Miller
- Published
- 2026-07-08
- Primary topic
- AI Agents
Why it matters
Agent products, browser agents, autonomous workflows, operator systems, and orchestration tools. This item originated as a short-form social post, so the context blocks below help expand it into tools, models, and evaluation guides.
What happened
We need AI model selection to be MUCH easier ASAP. I want proactive flags from my AI systems suggesting models. I want my AI harness to say "hey allie, my girl, you keep asking for bar recommendations that serve dry cider, um, I don't think you need Opus 4.8 high for that" or even abstract it away if that would actually work. Unfortunately the attempts from OpenAI and Anthropic have not handled that well in the past. You just can't expect a business user who is swamped with a million other things to know this. Today, my AI chief of staff is Opus, subagents are Sonnet, temp agents are Haiku, CoS assistant is Sonnet. And now, my advisor is Fable 5. Anthropic shared that on SWE-bench Pro, Sonnet 5 with Fable 5 as an advisor (getting called only once) got 92% of Fable 5 alone's score at 63% of the cost. They also shared that on the BrowseComp benchmark, Fable 5 as an orchestrator for Sonnet 5 subagents achieved 96% of Fable 5's solo performance at 46% of the price. We need better model selection. Or we need to purchase an intelligence package. Or something. But I think we will look back at the "Opus 4.8 xhigh ultra mode mini max v2" era and realize we were insane to ask this of users.
What to do next
Move into automation and workflow tools next so you can evaluate whether the agent story is actionable or still mostly experimental.
We need AI model selection to be MUCH easier ASAP. I want proactive flags from my AI systems suggesting models. I want my AI harness to say "hey allie, my girl, you keep asking for bar recommendations that serve dry cider, um, I don't think you need Opus 4.8 high for that" or even abstract it away if that would actually work. Unfortunately the attempts from OpenAI and Anthropic have not handled that well in the past. You just can't expect a business user who is swamped with a million other things to know this. Today, my AI chief of staff is Opus, subagents are Sonnet, temp agents are Haiku, CoS assistant is Sonnet. And now, my advisor is Fable 5. Anthropic shared that on SWE-bench Pro, Sonnet 5 with Fable 5 as an advisor (getting called only once) got 92% of Fable 5 alone's score at 63% of the cost. They also shared that on the BrowseComp benchmark, Fable 5 as an orchestrator for Sonnet 5 subagents achieved 96% of Fable 5's solo performance at 46% of the price. We need better model selection. Or we need to purchase an intelligence package. Or something. But I think we will look back at the "Opus 4.8 xhigh ultra mode mini max v2" era and realize we were insane to ask this of users.
This AimostAll brief summarizes the linked source so readers can scan AI developments quickly and jump to the original reporting when needed.
