It's an interesting phenomenon. If you know the domain, they can be really useful, within limits. If you don't, LLMs start to act like "psychics" who read what you want to hear, then spit it back out in official-sounding language. Like having a pocket "yes man" always hyping you up.
Interesting! I have 2 NAS boxes sitting there all bored. I was going to install this quick and be like "check it out!" with a screenshot. Then I saw Linux things and Docker Compose things, and it's after midnight. Maybe this weekend...
I really don't know and probably should, with my 2 y/o loving screens. I did build something for her, and UX testing with a toddler has been interesting. You should definitely meet Lauren Darcey, who apparently doesn't have a bsky, but can intro:
www.linkedin.com/in/perlgurl/
Actual pic from an apartment listing. It feels like AI, but it could also just be a terrible photoshop. Maybe a cheap 3d scan thing. Either way, is this helping the sale?
Weather went bananas for my birthday. Wind chill -29C. Flying off to warm weather in the morning, though. 2 weeks. Hopefully this winter nonsense will be behind us when we get back.
Hats off to the old school digital marketing teams who still think a birthday is a great excuse to reach out.
more doing significantly less on the API plan. The Clauge Agent SDK is amazing for putting something together quick without a lot of painful integration and debugging, and is on the subscription plan.
to help maintain project context docs by performing extended code analysis to find architecture, code, and pattern drift and update context. About to try multiple parallel and compare. 3 days into this week and I've barely hit 20%. Not that you should be wasting resources. But I was spending way
I finally tried Claude Max subscription. I had avoided it, assuming it was a bad deal. My first week was last week, and I barely hit 20%. It's changed my usage. I've written (today) deep project analysis tools that run the same analysis in parallel, then merge the results. Also wrote a tool
I do think open models will become quite capable and competitive, but to run a big one you need a big box. I have 96g of ram on a M2 Max, but there's a pretty hard limit to model size that isn't dogshit slow. Do coding work with those models? Nah.
I really like Goose, but this article is silly: venturebeat.com/technology/c.... I tried to do a "detailed" skim to see where it explained that a local model isn't going to be anywhere near as capable as a frontier model. Nope. The implication is that the $200 is for the tool. It's not.
Spent a lot of time with family this holiday season, exposing Eve to everything Christmas. If you actually count up the gifts of the "Twelve days of Christmas", and take the words literally, thats 184 birds, 140 people, and 40 rings. Your "true love" is an aristocratic stalker. Run.
SKIE 0.10.7 just released with support now for Kotlin 2.2.20/21 #KMP github.com/touchlab/SKI...
I lost two pair getting out of a single Uber once. Both fell out of my pocket (regular and pro). Right before vacation too.
I miss the days of the CVS receipt coupon scarf. At least that just wasted paper and ink.
My CVS, like all, has a crazy squeeze on staff. Yet, without fail, any time a prescription is sent by Dr, I'll get texts either asking or telling me my "auto-fill" is on. That means they refill as soon as possible. Whether I want it or not. Then it's a scavenger hunt through the overflow bins.
Pro tip for Android dev with AI agents. Don't let them edit Gradle. They'll rip out the wiring before checking the circuit breaker when trying to fix something. Hopefully the Android Studio agent has some better tooling than stock agents. www.youtube.com/live/0vMDXa2... (updates look really nice)
Curious to try with the updated Android Studio agent. I have some custom tools that will at least let the agent look up gradle versions, so the LLM doesn't pick something from it's cutoff date, but throw in KSP, etc, and it's a mess.
I generally don't let agents do Gradle now. They'll rip out the house wiring trying to figure out what's wrong before checking the circuit breaker.
Love AI agents, hate AI agents. They can be fun. (It did seem to think about that longer than I imagine it would have with "Yes", but we'll never know...)
Wow, sorry. That's terrible.
Take chicken wing bones. People toss them on the sidewalk constantly. The only time my dog actually fights me on anything is trying to remove a chicken bone from his mouth. We try, because they can be a problem, but it has to usually work out considering how many dogs and bones are on the UES.
I've heard of dogs eating chocolate and being fine. That's kind of what I'm saying I guess. Each of these things can be a problem, but if they were as bad as advertised for every dog, like how could they still be around?
Just looked up almonds, and they're not so bad. Might cause some upset stomach and GI issues. Of course, our dog as a puppy "ate" like 2/3rds of a bag of them. Salted. Drank lots of water after. Didn't chew, so they went down whole. Made for interesting walks over the next day or so.
Really?! I kind of vaguely remember that. So many things. Not that we're swimming in grapes, but there's no way he hasn't managed to steal some in his time.
Sometimes I'm amazed dogs have survived. Just found out onions are toxic to dogs. Onions, garlic, chocolate, almonds, various other nuts, on and on. Ours seems to be sensitive to chicken. Any time I'm about to break down and give him some of my food, quick check, and nope. Something's toxic.
Hell, if I was in college and did well in regionals (mecc.college), you're damn right I'd have some kind of terrible tattoo.
Funny you picked accountants. I used to joke that devs had hackathons, because we're really into code. Accountants don't have "Accountathons." I was corrected. Close your eyes, guess what the music in the video *should* be for, then watch it. www.youtube.com/watch?v=xnH0...
To briefly elaborate, you demonstrate the value of loyalty to the folks who feel the need to be protected (or whatever). It just seems like the people whose loyalty he wants to ensure, by carrot/stick, also hated George. Everybody hated George. But, hey, maybe not. I'm sure George is charming.
But, thus is why I don't post much about politics. We're collectively so far down the rabbit hole, posting online about it seems like it'll just get churned into some future LLM's training chaff. But, the GS thing stuck out as a wild potential own goal. Nobody cared, yet he did it anyway.