Fable 5 is also by far the best computer use model according to Stagehand Agent Evals it also costs less than half of GPT-5.5

I set up a new MacBook and only touched Settings once, to enable Codex Computer Use. Codex installed every app I needed from email client to Stream Deck, set up all my accounts (I just punched in my passwords), ssh keys, themed my Kitty with Starship, and stood up @code with great DX settings and a beautiful Catpuccin theme (even got the icons without me asking it to) What used to take me all day or a bunch of scripting was finished before my second cup of coffee Anything with programs, settings, anything administrative or fiddly — you should be chatting, not clicking Any company that's letting their employees set up a work laptop manually is wasting days of dev time

Parloa launched Agent Skills, built on MCP, letting CX teams connect AI agents to CRM, booking, ticketing and compliance workflows without middleware. Parloa says integration timelines can fall from 4-8 weeks to hours, with early results showing 67 sec lower AHT.

Here's a simple loop: Tell codex to maintain your repos, wake up every 5 minutes and direct work to threads. That makes it easy to parallelize steer work as needed. I use a orchestrator skill combined with my triage autoreview computer use skills, so some work can land autonomously. github.com/steipete/agent-sc… github.com/steipete/agent-sc…

Progress update on the physics-based vibe coding. Yesterday I got 2 days of work done in 30 minutes. I noticed the software, which simulates the loss of ice from lunar soil as it sits in an excavator bucket in the sunlight, was outputting a plot that seemed wrong. I pointed out the error to the AI, which told me immediately the cause of the problem in the code structure. In the first timestep the sublimation rate is high, and it extrapolates that rate for the entire time step, which predicts an unreasonably high amount of sublimation. I suggested to the AI that vapor pressure in lunar soil reaches equilibrium in a microsecond, which is short relative to the timestep, so we could restructure the code to always assume equilibrium conditions and use a Robin boundary condition for the outflow. After discussing with the AI how to make these changes in the least disruptive way — to minimize the risk of introducing bugs — I told the AI to make it so, and it gave me error-free code on the first try. For the past 24 hours, I have now been running simulations and getting perfect outputs. I feel confident that we now know how to design lunar ice mining to avoid loss of the resource. I’m working in python, and I don’t even know python. I’m working from my laptop, but there is no compiler installed on my laptop. Everything is running in the cloud. My productivity has been multiplied by a large factor. The people who say AI is not giving us practical benefits are simply not paying attention.