When we spoke on the call yesterday I gave the wrong cost for this model. I opted for the next level up so I could dispatch research agents in this mode: "Goal mode, where agents autonomously work until completion"

https://lnkd.in/p/gTnTXdY9

@Kari Davis quick one I thought the group might find useful. Now that Fable 5 is staying on as my assistant, I had it map out something I'd only ever kept in my head: how Claude in Chrome and headless Playwright actually divide the work in my setup, and how the system decides which one to reach for on any given job. The short version is that it stopped being a "which tool is better" question. They're two engines for two kinds of work. Anything hands-on — verifying a deploy, checking a page behind my own logins, a client walkthrough — runs through Claude in Chrome. Anything unattended and repeating — my scheduled monitoring, research pulls, bulk work — runs on Playwright, because Chrome can't sit in that seat at all. Build on one, ship on the other. The attached brief lays it out against my real workflows, plus a section on how the same split turns into client offerings. Thought it might be of interest.

https://youtu.be/5teG6ou8mWU

I don’t know if any of this works, but most of it sounds reasonable enough, and is easy to implement. To paraphrase Otto von Bismarck -” It is the wise pirate who learns from the mistakes of others.” The link is from X but you don’t need an account to read it. 🏴‍☠️🏴‍☠️🏴‍☠️ https://x.com/free_ai_guides/status/2073050543027638443?s=46