My wife was in her second trimester when the lockdowns started here. We only found out I could be present for the birth the week of. The pandemic took everything that made our walkable neighbourhood great, and WFH with a newborn in the next room with no activities or distraction sucked for everyone.
Ah, I didn’t realize that the fifth reply gets to claim ownership on the thread and overrule the context of the previous replies. Carry on, I wish you luck.
Did you… not realize you were jumping into a thread full of lawyers?
Missing or stray punctuation in code will cause a compiler or runtime error. Missing or stray punctuation in a contract can cost millions.
I think LLM boosters sometimes fail to appreciate what correctness and validation means in non-code use cases. Code runs or it doesn’t, and the LLM can iterate until it does. It’s the low hanging fruit of LLM use cases.
But correctness and verification in a legal context is not so easily done.
Bold move to attack a guy when your chief complaint is that he’s wielding his metal mallet too loudly.
Oof man. My sympathies. So much depends on the trajectory of their personality as they decline.
My FIL has mellowed out and is easy going.
My aunt had her mom move in and it culminated in feces being thrown, punches, three cops, two paramedics, a psych nurse, and finally a trip to the ER.
I think the customer service use case actually has legs and I expect improvements there in the near term (at least some examples of firms with impressive implementations).
I’m still looking for a good meeting notes tool. I hate taking notes.
I was having issues with Claude (quality and limit changes both), so I decided to try OpenCode, and I don’t really miss Claude. Obra/superpowers still works, and at least for my hobby projects I get way more value from their Go service. I can use a copilot subscription to get access to Opus.
His national profile seems to have juiced his royalties by 25%, but frankly his $1267 in royalties from 2024 is way more than I would have expected.
The usual options would be overcooked or undercooked, but that's just sad not funny. Did you find a secret third thing?
It’s definitely a sliding scale as you age. The half your age plus seven thing shouldn’t be taken literally, but it does contain a nugget of truth.
NIMBYs are a menace, but dealing with them is only necessary and not sufficient for fixing the mess housing has become in the anglosphere.
Housing is also almost unique in that its in the base of the hierarchy of needs, has been fetishized as the cornerstone of middle class wealth, has been financialized to hell and back, and new supply has slowed by rising costs, reduced government investment, and zoning/regulatory restrictions.
Geography is obviously a factor, but policy choices also play a role. 50% of Canada’s population lives in a straight line from Windsor to Montreal and there’s no high speed rail here. There’s some talk of building one recently, but it has all the early hallmarks of a failed boondoggle in the making.
And if I really want Opus back, there's a $10 dollar Copilot subscription that has access to it. I'm sure I'd burn my usage there pretty quick on Opus, but I wouldn't need it to do everything, just specific tasks that needed the bigger brain.
I've been playing with OpenCode and an OpenCode Go subscription, and at least for my needs there's not a huge difference between mid-April Opus and GLM-5.1, and I get way more usage from the OC-Go sub for half the price.
I noticed the changes to usage limits first, but recently noticed a definitely decline in service quality (read: smarts). I'd already canceled my CC Pro sub over the usage limit changes, but I'd planned to use it until my current month expires. Now I'm not even sure I'll bother...
Also, I feel I should point out that Anthropic seems to be having capacity issues and is dealing with it in part by making Opus dumber and making the system prompts in Claude Code lazier. Check this out: gist.github.com/roman01la/48...
I wasn't necessarily suggesting you leverage obra/superpowers directly, but the approach of leveraging skills and sub-agents to review the work is interesting and could be adapted to many problem domains.
If you’re using Claude code, look at the obra/superpowers plugin. It’s a set of skills and hooks that enforce a lot of (ai powered) verification. It’s not a substitute for human review, but I’ve watched the code reviewer agent reject work and send it back to the implementation agent to fix.
Not “like”, it *is* a phishing scam.
The same bad actor is trying multiple methods, and only a couple touch on the nature of ATProto (and account hijacking isn’t one). But they’re all just work to build the funnel into the scam.
Stand back everyone, we’re watching a true master - the man who literally wrote the book The Art of the Deal - work his magic. Let’s see what masterful gambit he’s using now…
66 years separate the first powered human flight at kitty hawk by the wright brothers and the moon landing. That’s the one that really boggles my mind. Like, there were a lot of folks alive to read about both events in the newspaper.
Alright, so the state of play is the Iranians can toll the strait if they share the proceeds with Oman, and they can enrich uranium as long as they share it with the Holy See. Is that correct?
I don’t need voice chat, but I’m testing Claude alternatives and I kinda miss the more conversational approach that Claude & Opus has, especially once I tuned it with an output style.
Don’t leave us hanging. How did the dog feel about the book?
The failure mode for space is better than bears because it’s probably going to be faster.
That the Easter bunny brings candy to our house that’s literally illegal in the US makes me wonder: is the Easter bunny a franchise operation?
One day, "It" will happen, by which I mean sudden and unexpected news that you want to celebrate. In such cases, you will want the right outfit. 🧵