Cheat codes are available, by the way: GPT 2, which couldn’t even summarize this thread, successfully caused data centers to be built, by making it incentive compatible for humans to build data centers. It didn’t want to do that. It probably wasn’t capable of wanting anything.
@patio11
-
Debate on whether LLMs can physically deliver a pizza
By
–
Is this argument “LLMs can’t make a pizza appear at my door?” Because there is a fairly straightforward way for them to make a pizza appear at your door. Or is it “Well they can’t use humans; that’s cheating”, in which case cheating is delicious if slightly fattening.
-
Dashboard scoring institutions against $20/month LLMs
By
–
I’m looking forward to the constantly updated dashboard scoring high status institutions against $20 a month LLMs, LessWrong top voted posts, and a hamster attached to a webcam. I think the experts will beat the hamster, but important to have a control group.
-
LLM handles first-pass chargeback resolution instead of BPO staff
By
–
If one is looking for an issuer-side adoption story it is “First-pass resolution for all chargebacks goes through an LLM with relevant context provided to it rather than a BPO employee repetitively clicking Chargeback Provisionally Approved, which sends the notification.”
-
Academia more scandalized by AI-assisted papers than false
By
–
This implies that academia would be more scandalized by papers being AI-assisted than them being merely false. … That is, regrettably, very plausible to me.
-
LLMs enable trivial click-through tasks via text messages
By
–
I think LLMs are such an interesting tool to build into very complicated apps because the first use cases people will do are the things they can already do ~trivially by clicking through screens, except now with a text message form factor.
-
Mercury’s Command feature tested with wire transfer, works on first try
By
–
Took Mercury's new Command feature out for a test spin with approximately the scariest thing you can do in a bank account, asking for a wire out, and it worked on the first try. Nice UX on upgrading from chatting-with-an-LLM interface to embedded this-is-fully-engineered flow.
-
Claude message replaces SQL for trivial text edits
By
–
Probably even better now that trivial edits to it are a message to Claude away versus having to write a SQL query to do equivalent of a one character/line insert/edit.
-
Handlebars templating can’t do string manipulation for LLM index
By
–
"What kind of programmable automated system can't do string manipulation!?" Handlebars templating, apparently. (I'm building an LLM-available index into some of my sites to have various tools able to read status of pieces w/o needing direct DB access.)
-
Dumb system apologizes to smart system in 2026 programming
By
–
I just caused an automated system which is dumb to apologize-in-advance to an automated system which is smart. (~ "The system which generates this file can't do string manipulation. You can, though, so to succeed you will want to… Sorry about extra work.") Programming in 2026.