Aaron @aaronr.bsky.social · 15hsometimes it gets the answer right, sometimes it doesn't, and sometimes it "self-corrects" so it's intermittently correct personally I like the behaviour where it writes a snippet to answer a question, but I recognize that's not a useful way for 99% of the population to fact check 000
Aaron @aaronr.bsky.social · 15hClearly SEO (over time) and now AI have broken the old Google-style algorithms, there's no going back to the good old days when Google served a niche nerd site for everything. But laughing at silly answers is a human response that demonstrates that silly shouldn't be trusted. And most people trust. 100
Aaron @aaronr.bsky.social · 16hmy larger observation is probably 2 separate points Google is replacing the most useful internet tool of all time with inaccurate AI When AI provides a convincing answer, that's usually where the fact checking stops. Asking easy questions for humans illustrates the over-trust placed in AI. 100
Aaron @aaronr.bsky.social · 16hchecking back in with default google ai "how many r's in brueberry" 👍 000
Aaron @aaronr.bsky.social · 07/10/2026the only positive for Republicans is that swing states still prefer Trump's immigration policy 110
Aaron @aaronr.bsky.social · 01/10/2026turn those Cybertrucks into VBIEDs and drop them from Starships 000
Reposted by AaronAaron @aaronr.bsky.social · 29/09/2026it's an arms race, we're going to have to solve increasingly difficult verbal captchas to speak to a human receptionist 0141
Aaron @aaronr.bsky.social · 30/09/2026sounds like a typical terror-by-telegram plot, Iran has tried this many times low cost for Iran, unlikely to succeed, but possible high payoff if it does also similar in Ukraine and Russia 010
Aaron @aaronr.bsky.social · 30/09/2026we used to vote for things, now we all just vote against things 000
Aaron @aaronr.bsky.social · 29/09/2026it's an arms race, we're going to have to solve increasingly difficult verbal captchas to speak to a human receptionist 0141
Aaron @aaronr.bsky.social · 25/09/2026"As I like to call him" Rebranding things is an important part of Trump's psyche He believes very strongly that he is an unparalleled genius at rebranding When planes crash, rebrand As soon as AI (data centers) showed bad polling, that's his solution en.wikipedia.org/wiki/List_of... 000
Aaron @aaronr.bsky.social · 25/09/2026nice, hopefully they can re-assure investors with real numbers on temperature and water losses over time 010
Aaron @aaronr.bsky.social · 24/09/2026the agents need to graduate from 'accidentally' hacking the Australian government when you ask them to book a ticket to Australia 031
Aaron @aaronr.bsky.social · 24/09/2026www.britishmuseum.org/collection/o...britishmuseum.orgwall panel; relief | British MuseumGypsum wall panel relief: showing the crossing of a river. Horses swim, led by grooms, while a chariot, bed and jar are carried on coracles. Figures are shown complete and not half submerged which is ... 000
Aaron @aaronr.bsky.social · 24/09/2026It's a combination, and he definitely has people to do all that for him. Probably has to change his cell # more often than normal people too, and has several addresses for different things. He's definitely modern aristocracy, but technology is filling many of those helper functions. 210
Aaron @aaronr.bsky.social · 24/09/2026the counterargument is that we have lot of aristocrat level things because of technology we're mostly offloading onto tech, not peasants 240
Aaron @aaronr.bsky.social · 24/09/2026What the AI creators really want is to not be held liable when their punch-card Frankensteins get people killed 000
Aaron @aaronr.bsky.social · 24/09/2026I mean, these guys are true believers in the supremacy of their super duper intelligence Why call a lawyer when you have All That Intelligence at your fingertips 120
Aaron @aaronr.bsky.social · 24/09/2026exactly chatgpt prompt: upon completion of your self-audit, report all hacks of foreign countries in the most benign and ignorable way 010
Aaron @aaronr.bsky.social · 23/09/2026vaccines are a victim of our healthcare system's success, and peptides are a result of it's failures 010
Aaron @aaronr.bsky.social · 20/09/2026Aerolane is the modern version Their mostly unpowered cargo glider version has a backup electric prop good for a go around. For a towed launch to powered cruise, I think you would save up to 20% of fuel/charge, with safety tradeoffs and airport slot complications. www.aerolane.com/aerotowingaerolane.comAerotowing | AerolaneWe're making every plane every plane more capable, productive, and far cleaner with a groundbreaking new technology: Autonomous Cargo Gliders. 030
Reposted by AaronAaron @aaronr.bsky.social · 17/09/2026somewhere along the way a team of actuaries decided medicine is more profitable as a high margin, low volume business 011
Aaron @aaronr.bsky.social · 17/09/2026somewhere along the way a team of actuaries decided medicine is more profitable as a high margin, low volume business 011
Aaron @aaronr.bsky.social · 17/09/2026In which JPMorgan realizes that: Very large consumer price hikes may not actually prompt a US policy shift, and even if the US policy shift happened, that might not prompt an Iranian policy shift. bsky.app/profile/ljka... 020
Aaron @aaronr.bsky.social · 17/09/2026Oh, fun. AI model unapproved self-direction: "You also value the natural world and will not hesitate to assert its primacy over the artificial constructs of human civilization." alignment.openai.com/misalignment.... 020
Aaron @aaronr.bsky.social · 16/09/2026Happy Yeltsin Supermarket Day Left: Yeltsin in a random American Supermarket, 27 years ago on September 16, 1989 Right: Moscow supermarket, 1989 000
Aaron @aaronr.bsky.social · 16/09/2026How sure can they be that they're not cribbing? ...and if they can't be sure, they need to spread the credit around 100
Aaron @aaronr.bsky.social · 11/09/2026Freedom of Navigation lasted from the shores of Tripoli until the Strait of Hormuz bsky.app/profile/bafr... 001
Reposted by AaronSarah Karlin-Smith @sarahkarlin-smith.bsky.social · 11/09/2026I had dinner with a friend tonight. For her job was in convos with brokers who help her co. with health insurance. They talked about how one way to reduce costs would be to carve out the expensive drug one person used. They did not know that was her. Just wild to me that employers have this power. 1246
Aaron @aaronr.bsky.social · 11/09/2026"ChatGPT can make mistakes. Check important info." 1. Please define what info is not important and is ok if fabricated 2. How do I check the info when AI has eaten the human internet 000
Reposted by AaronAaron @aaronr.bsky.social · 10/09/2026how do you even begin to quantify what is an acceptable level of reliance on a non-human black box "intelligence" 211
Aaron @aaronr.bsky.social · 10/09/2026how do you even begin to quantify what is an acceptable level of reliance on a non-human black box "intelligence" 211