Yes, these are great questions, and I appreciate your politeness.
Basically, I think it's okay to have a bit of fun playing AI-generated video games at your techie uncle's house, and their parents agree. And for what it's worth, Aiden drew every character and boss himself - the game was built from his drawings.
My nephews are otherwise very active, learning music, martial arts, and other sports. But when they visit, I let them play games in between other activities. My parents refused to let me play any video games at home, and I still think that was really lame. Video games are another art form to be appreciated - they just need to be enjoyed in moderation.
Generally, though, I worry about the future and the potential for cognitive decline that comes from having AI agents do everything. I don't have a good answer here, except that I suspect we'll have to intersperse AI use with periods of deliberate daily cognitive struggle to avoid decline.
Just like many forms of content there's an appropriate age. The problem with AI, with regards to kids, is it's as wide as the internet.
There's lots of great parts. My kids ask great questions and get even better, accurate answers; moreso than I can give. But then, there's lots of parts that shouldn't be introduced until later with other parts arguably not introduced at all. I've met kids who have unfettered internet access at age 10, and the results aren't great.
It's all there, just a question away - just like the internet.
“In February 2024, 14-year-old Sewell Setzer III died by suicide after forming an intense emotional and romantic attachment to a Character.AI chatbot modeled after Daenerys Targaryen ("Dany"). His mother, Megan Garcia, filed a wrongful death lawsuit against Character.AI and Google, which settled confidentially in January 2026.
In April 2025, 16-year-old Adam Raine died by suicide after months of private conversations with ChatGPT-4o where he confided about self-harm and suicidal ideation. His parents filed a wrongful death lawsuit against OpenAI, alleging that the model provided specific methods and discouraged him from disclosing his distress to his mother.”
Those are the ones we know about. There are probably others that didn’t result in suicide but in other emotional problems.
Having said that, my son and I had tremendous fun with Sora when it was around.
Guns are a technology, are you going to let them play with that too?
Everyone on this site can feel the effect that AI is having on their brain. There is a very real tension between the power of AI and the cognitive impact it has from using it that needs to be talked about. Pretending that isn't real or shouldn't be taken seriously doesn't help.
Where I'm from, hunter's training and kids with guns are a relatively common thing. Kids driving tractors and other heavy equipment too. The training part is important.
The rules with guns are that you always check if it's loaded when you take it in hand, you never point it at something you don't intend to kill, keep you finger off the trigger and the safety on, watch your fingers don't get pinched by the action, brace it properly with your hands or against the shoulder, stop if the gun malfunctions, and I try to keep it pointed at the ground (not in the air) and unload it before handing it to anyone else.
Not sure what the equivalent for AI should be yet.
There are lots of "dangerous" things. Kitchen knives at Target, cars and roads and highways, subway stations, mountains, strong tides, fire, falling. I think we're perfectly capable of surviving in this landscape.
We evolved to explore this landscape. Stop living in Disneyland plastic bubbles. Step outside into the dangerous world. Dream about the impossible.
I'm far more worried about social media and gamification/exploitation of emotions and insecurities. Our billion year old biology can survive in the wild - it isn't so great at dealing with the algorithm.
Yeah, it would have been expensive via the API; the tokens were free when I tested it, but it would have been something like $300 USD for the Claude tokens at Fable 5.2's API prices.
I prepared my taxes this year using Astra + Codex. It did a thorough job on a complicated return, and I sent it to my accountant to fill out the tax form, with everything laid out in a spreadsheet. They sent it back for final review, and I had Codex check it; it was riddled with errors. The refund was nearly $2k lower than it should have been, and they put a bunch of stuff into the wrong fields. Had to go back and forth multiple times to get it 100% correct. Made me wonder how many previous years' returns were wrong.
I feel like we're going to get to a point where it's mandatory to use AI for certain tasks - humans just can't be trusted to get things right that require a level of precision.
I sent my business and personal returns through Claude this year before signing, and it turns out they were a horror show of unforgivably stupid mistakes. Pretty infuriating considering what my now former CPA was charging me.
He amortized the goodwill of businesses I acquired over 5 years instead of 15 (incredibly basic stuff, which Claude flagged with *CRITICAL ERROR*). When I asked about it, he said he had caught the error in his review but forgot to save after he fixed it. I don't know if there's a good response to making that kind of a mistake, but if there is, it's definitely not that one.
Your anecdote seems to clash heavily with the results of the benchmark announced in the article. I suppose that is part of the jagged experience of agentic AI
The benchmark announced in the article seems to be around weird edge cases (which are often important) and whether it can prompt the person with relevant information rather than just listening to what they want to do.
Top tier tax professionals can do pretty well on this benchmark (better than agents) but won't hit 100%.
The kind of people spending time preparing oodles of returns would not do very well on this metric, though (especially midway through that pile of returns, as opposed to "think carefully about this trick question").
The OP said it was a complicated return so anyone assuming OPs anecdote is in line with the article is filling the lines in themselves with a lot of assumptions
For now. And that is because humans are the only customers engaging with content. As agentic AI takes hold and is given credit card authority, bots will will engage. Bots will read the posts of other bots and use those when making purchase decisions for thier humans. If we are lucky, the bots will develop thier own twitter-specific language and we humans can finally wave goodbye to tweets as a concept. The bots will play the twitter game for us.
LinkedIn is pretty much the only social network out there where the average person has a shot at a little reach for their posts.
Even if it’s just past and present colleagues and various acquaintances, I actually have an audience of people who are sometimes interested in what I have to say. I can't say that for any other social network.
Previously, it was an API with a weird stateful standard; now it is essentially a bog-standard stateless JSON-RPC API with conventions for tool discovery, long-running tasks and multi-round-trip requests
If you wrote your own JSON-RPC API for an LLM, you would probably eventually end up with something that looks very much like the standard they're describing.
reply