My wife grew a pineapple in the back yard. Turned out pretty tasty!. And she says you can plant the top and do it again (takes a few years). Will have to survive the “winters”

Claude's annoying because it likes to do verbose summaries and tell you how awesome it is and how proud of its work it is, but at least it has some justification for its joy.

Claude sonnet 4 is still better than gpt-5 for coding ia cursor. Doing some real basic things (playwright tests for a vscode web extension). It's not just "did you arrive at something that works" (gpt5 struggled) but also "is the code nice?" . spent an hour+ flailing. nah.

A couple of salesforce/vscode extension announcements I wanted to publicize:.

Sun’s up in Texas. Time to burn less fossil fuel and recharge the batteries with cheap power.

if I had more time to screw around with ai video generation, I'd love to redo the IronMan "vibe engineering" scenes where Tony is working with Jarvis. Except that Jarvis lies to him and makes up stuff and his suit won't fly but then argues that it's actually flying.

interesting, it does ok reading one domino. But it can’t apply that multiple times, even though the response reads like that’s what it knows to do.

Someday the robots will be smarter than a 1st grader.

a clear breezy day here in Texas @ERCOT_ISO

Thanks, google. Glad we have AI now for amazing insights like this. And that off topic explanation…

My first time seeing the Gulf of America.

Introduced grown son to X-files Paused it and saw the best ad I’ve seen in a while. Nice @PNCBank

I see why you'd do that for love, but for anyone else, much less a robot? the code gen just isn't great to begin with.

or maybe a more charitable way to think of it is like the movie 50 First Dates (spoiler alert) where he ends up making a video to explain the world for the girl to watch each day.

like, I'd invest time in a person, knowing they'd gradually accumulate knowledge/wisdom. adding it to "the prompt" knowing that there's a limited number of characters is like working with a moron.

with LLMs, they can't remember any of those things unless you tell them each time.

even with someone who is learning for the first time, they'll start to adapt to feedback, pick up the patterns, etc and lean/improve.

Imagine you are pairing with someone, and they keep making the same mistakes repeatedly, and the only way to prevent them from doing is to write long docs about what not to do.

the reason I hate working with AI for coding is that it's like working with a moron.