GitHub Copilot uses a request model too. The plan request creates the plan then the “go do that” agent request is another one. I probably iterate on the plan a couple of times (each a request). Times the model multiplier (e.g., 3x for Opus 4.6).

I sell AI dev tools. The ROI conversation starts with them telling me they need to measure the increase in developer productivity. I ask them, “how do you measure developer productivity today?”

🦗 crickets 🦗

I’ve tried explaining this to family. Two of them run IT businesses. One is embracing AI and experimenting how it can make his processes run smoother, and learning how to effectively build agents. The other still thinks of everything AI like it’s GPT-3.

Wooooooow

Nope.

Nice. I’ve down ranked some domains or blocked them but the report option can help others. AI generated isn’t always slop but I’m running into more and more pure slop.

Firin’ mah lazers!

+1 for @kagi.com. I have a family sub and enroll my older relatives as well.

Looks promising! openhamclock.com

Yeah it really makes me sad how much of it is out there. I stopped spinning the dial on 40m at night. Can’t imagine 80m. 😅

On Signal we’re all the same. 💙

Built myself a whole thing in surprisingly little time with Opus 4.6 in GitHub Copilot CLI yesterday. This is really impressive work.

I have to wonder how many people still have shortwave receivers.

HamStudy.org is your friend. And even if you don't deeply understand all of the topics (just enough to pass the test), lots of things make *a lot* more sense once you can start using the sklls. As another ham told me, "it's a license to learn."