arjie

Sorry I missed your comment, Peter. I don't know that it's worth very much to describe. It's not sophisticated. Each component just has an attribute named data-ai-context and content that describes it. This needs no LLM. e.g. charts go into canvas, so their data-ai-context has a textual summary of the content (same as is fed to the charting library); buttons describe what they'd do. Then when you click the "Copy to AI", a piece of JS hits a backend to sign a 15 min token for the same user (as if they'd logged in but with a short token) and pulls all the data-ai-context attributes and gives you a prompt in your clipboard describing the site organization.

It just so happens that doing that tiny first bit allows even stupid models to navigate the site. After that, advanced models also sometimes query for the data-ai-context and they can mimic the Copy To AI to not have to nav the site.

e.g. content "Job Failure Rate: failed sync jobs as a share of jobs run, per connector per day, over 7 days. Day X: facebook=8 failed / 862 jobs; ga4=0 failed / 196 jobs;"

It's really just an alt text element and bizarrely worked the first time I tried it with old Opus 4.5 and I've kept it. Perhaps the biggest use is giving the agent the short-term token.

Wait. I remember reading about this. Doesn’t SBA 7a cover your use case? I’m sure there’s an SBA loan category for you. It’s backed by the machine. Recently they changed the valuation of things or something.

Yeah, I particularly like London’s implementation (I moved away just short of a decade ago - might have changed since) where if you just use the same card you don’t need a pass because there are daily/weekly/monthly maximums and you travel for free afterwards. Very good system. But it is slower that Suica/Pasmo.

That also happens with Clipper Card in the Bay Area. Happened to me a couple of weeks ago on BART. I wonder if all transit cards need to be loaded with money when used on iPhone. I remember having a Clipper Card with autoload and that would work just fine without balance.

Personally, I just can’t find it in me to be that self-interested. It’s how other people oppose buildings near them because it blocks their view. I really don’t want to stop other people from writing software. If they want to use AI to do it so be it.

That’ll be somewhat detrimental to me perhaps (which isn’t certain), but that’s okay. I’d rather attempt to adapt to a changing world.

My experience has been that responsiveness is value. For tasks where you need steering, responsiveness allows for better steering. For tasks which you want unattended, better models are just better.

There are still tasks that even Fable is bad at doing. And many are just mundane things. Because of the fact that you have to steer it on those tasks, you might as well steer an 80% model that is 5x faster. And those do exist.

Naturally there’s a bit of a gap because the faster models need steering on tasks the slower models don’t so there’s no smooth transition but I find it worth it. Especially if you want to stay in flow.

Ironically this sometimes means starting a plan with a great model, planning with a worse model, iterating, then submitting it to a better model for review, and then having the better model do the implementation.

It seems to be a self-instruction. Leak from reasoning token into output token or using output token for reasoning. Perhaps something accidental like constraining number of reasoning tokens leads to model using output tokens for reasoning or perhaps ensuring some minimum user-visible reasoning. Seems like an accidentally introduced artifact.

“I should not pretend other implementations don’t exist. I should name them plainly.” Unable to tell itself that it puts this in output so that it can tell itself there.

Same with the “not X but Y” style. It is self-steering introspection leaking into output for one reason or the other. Recent poor language use by model is probably an attempt to minimize this by being concise. But it really needs more thinking to solve problem and so it does some of it in output.

The GPTs are just as good and I can trust many tasks to open models. Autonomous coding is best with frontier class models but steered coding is fine with even DeepSeek V4 Flash.

I think it's pretty fun. Worker cooperatives no longer need different ranks of executive officers. They can all be peers with different functions. So if worker cooperatives are going to succeed in the US, they have pretty good conditions right now.

I’ve never had any trouble going to Taiwan. One of the easier places to go as an Indian who lives here in the US since you have an online form and you get 6 visits of 14 days each and you can do that while you wait for your visa application. My parents also found it quite easy and they live in India.

Also, even with NWOHR it’s not that hard. What trouble did you encounter? Might be useful for me to be aware.

Quite helpful, I’ll see if I can serve text/plain.

150 confirmed with 47 Americans? Surprising fraction. Mountaineers or that area is popular for some reason?

Everything is determined by cloudflare. If they supported Gemini on the same hostname I’d serve it everywhere. If they supported text/plain and text/html everywhere I’d serve that. But they don’t cache per content encoding so I just do what works for them.

They’re far too useful for me.

ATProto attempted to do this with their PDS based structure but it isn’t fully decentralized yet. The vision is similar to what you describe though.

Yeah, but they tried to change Adjustment of Status too.

https://www.uscis.gov/newsroom/news-releases/us-citizenship-...

I don't think the practice of the change is significant. Reddit has numerous stories of people who are still receiving it, so whatever that release says, things haven't changed. People seem to be receiving AoS based Green Cards the same way I did under the Biden Administration. It does seem to be a bit more uncertain which is always troubling.

Yeah, this is a pretty big risk. I'm similarly married to an American and we have an American child. Other country seems to have screwed the family, though. Mine (India) offers a fairly indefinite visa called an Overseas Citizen of India. And my wife's other nationality (Taiwan) also permits foreign spouses to live there after some straightforward paperwork. So, in theory, we could live in either of these other places together. I suppose I will only know when push comes to shove.

Very cool! Great resource dude.

If you're curious, the San Francisco Sanborn Maps are actually available directly from SF. Fun to look at https://sfplanninggis.org/pim/Sanborn.html?sanborn=V3P253.PD...

Everything is "chronically underfunded". There has never been a thing in the world that was "well funded" because people's objectives increase with the amount of money they have. No one has ever said "Wikipedia has enough now" or "The Internet Archive doesn't really need any more money" or "Firefox has collected enough money from Google that it can now operate independently forever".

Besides, this so-called goodwill is an illusion. Opposition to AI in the PR space is entirely dominated by motivated reasoning. Attempting to satisfy people with random things is simply going to shift complaints to "take the commons, chop it up, and give it back to us"; "Should AI companies be the arbiters of what is worth preserving?"; and other such things. These kind of public-relation gains are entirely fictive and neither the supposed beneficiary nor the supposed benefactor is fooled by the theory.

"Oh so they suck up all our water and burn our planet down and we're supposed to be happy that they chop up people's books (that humans spent years creating) just so they can give us digested slop back?". Everyone knows that this sentence is waiting in the wings, and that the entire effort is clearly bait so that someone is foolish enough to pull a Google Books and get sued into oblivion for it.

And that's saying nothing about this so-called "hoarding" of things in a world where we pulp tons of these hoarded things every day and no one cares to scan them.

The governments go where the people are. If they post on their own sites it’s the equivalent of “on display in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying ‘Beware of the Leopard’”.

But government information should definitely be first source on their site with federation elsewhere.

This is going to sound incredibly silly but I have every HTMX component expose an AI summary with actions possible and then I have a small copy to AI button on the page which creates a new short-lived access token and copies into a prompt the token, and all the components as text.

When I give this even to relatively small-model agents, they use this initial seed to browse the site and do things very well.

The initial prompt suffices. After that most agents just use the HTML to navigate very well. I suppose I could have that description in an aria label if I wanted but it’s the same.

For people familiar with the field, it would be like if you had every article about hardware read “Intel Central Processing Units (CPUs) with modern Double Data Rate 5 (DDR5) Random Access Memory (RAM) can be coupled with Nvidia Graphics Processing Units (GPUs) to run Large Language Models (LLMs) that are stored on Solid State Disks (SSDs)”. Just rapidly becomes unreadable.

The acronym constraint was valid in a pre-LLM world but now you are perhaps 3 clicks in a modern browser from learning.

If I read an article that spelled out Random Access Memory I would definitely treat that as a lay article.

One thing that happened is The Copenhagen Interpretation of Ethics, right? Mark Zuckerberg is panned for having 'defunded' a school. What happened was he paid for it for a while and then stopped in a planned manner. Those who didn't ever pay were better off. In general, at the margin you move some guy who was ambivalent to opposed. Jordan Henderson spoke for gay rights, and got panned when he moved to Saudi Arabia, while many footballers did not speak out about gay rights ever and received no censure over this.

The general rule with many of these causes is that unless you're willing to be absolutely committed, it's better not to be involved. If you cannot prove to yourself absolute fidelity perpetually into the future, it's better to not do it at all. The punishment for the apostate far outweighs the punishment for the infidel.

So your commitment has to be at least high enough to be willing to bear the resulting punishment for your apostasy. And if it is any less than that, you are strongly encouraged to just stay out of it. It's not that anyone is mean or anything. It's just that on the margin not-participating costs nothing and participating incurs a massive liability. That causes a shift in the window.

> The survey showed that 62% of U.S. adults self-report having little to no idea what statistics or statistical concepts like p-values are, but 90% of them would base decisions on reported statistics at least sometimes if they understood them better.

You're kidding, right? 38% self-report more than that? If their self-report were accurate it would imply an education system that has truly excelled.

Strangely, doesn't load for me with some Vercel/NextJS error. Disabling Javascript displays the whole post, though.

In an odd coincidence, Jolene by Dolly Parton played on YouTube Music as I drove home with my daughter. When we got home, my wife told me Dolly Parton had died. I don't listen to much of her though I'm fond of this song, and we get the Imagination Library books. Seems like she was good person, and a great singer and songwriter.

It would be pretty neat to actually have the thing funded through a revenue neutral ticketing scheme. i.e. you just auction off tickets for 4 week blocks. That will allow people who are time-sensitive to appropriately select, and if you make more than you need you can just rebate participants proportional to their spend. Might be a pretty good way to allocate the scarce resource.

Hahaha, I was surprised by one of those in Newark/EWR. I can't believe it's real haha. Who gets the tip? I just tipped the self-service worker in cash ;)

Thank you.

Interesting. What POS terminals are common there? I am curious about culture spreading through technology defaults and wonder if it's effectively driven by whatever Square chose and whomever else decided to ape the incumbent or whatever.