At some point when I was bored and playing around with LLMs, I had chatGPT keep making ascii art over and over. I'd ask for something like a duck holding an umbrella. No matter what blobbly garbage it produced (and it was all blobby garbage), I just kept encouraging it and telling it to make more. Add details, refine, making it even duckier, etc. And it just kept taking the input blob and adding more blobs to it, like an infinite ascii blob spiral.

What you're describing isn't an implementation of RAG, but supervised post-training like Directed Preference Optimization. In DPO, researchers compile sets of answers from LLMs to the same question with a human reviewer selecting which one is a better answer. This data is used to fine tune, which steers models towards the good answers and away from the bad ones, which is why it is such an effective means of changing a models' behavior to a certain style of answer.

It gets pushed way too far, which is why you can find tiny open-weights models that crush benchmarks, and tell you confidently that "there are 3 'r's' in strrawberrrry," because, as you point out, they can't count letters in token sequences.

No it isn't. It's like uploading a pic, and *not pasting the link anywhere*. It's not clear to some people that that image can now be indexed. It is still on them for not knowing that, but they don't have to have done anything to draw attention to the linked content.

... and the gate opens and lets you out. What happens if you can't find an employee? I don't know, every time I've encountered this at a safeway that's implemented it, there was an employee standing next to the gate machine to let me out. I've only seen this implemented in the self-checkout.

This is the kind of comment that passes a "common sense" sniff test, but is impossibly naive in the context of any legal system. This is only slightly tenable if the question before the court is one of pure statutory interpretation. If the *only* question before the court is, "what does this law say and do?" then in *some* circumstances, you would answer the question despite unfavorable policy implications, and even indicate that the legislature should address the found defects.

Conversely, in common law jurisdictions, which includes the United States, judges are relied on to shape the direction of the law through analysis of precedent, existing statutory schemes, and policy considerations. Further, legislators in common law jurisdictions draft laws knowing that the common law exists, and borrowing definitions and language from the common law with the intent that judges fill the gaps and interpret with an eye towards precedent and policy considerations. So, your position is plainly and openly incompatible with much of the US legal system, and many other common law jurisdictions. And even elsewhere, the lines are far less clear-cut.

There are conflicts of laws, superseding laws, open questions of a laws scope and breadth. Many laws explicitly call out policy effects as being a consideration in the application of said law. There is the law of Equity; an entire separate body of law that runs parallel within our system, which explicitly requires judges to balance the burdens (consequences) of a decision and certain notions of fairness resulting from the application of law to fact patterns and individuals, for the issuance of certain relief. These centuries-old frameworks often explicitly includes "policy considerations" within their well-established factor tests. We have statutes like ERISA that broadly govern all employment benefits plans (remember how healthcare is tied to employment in the USA?), and explicitly state that all remedies under this broad law, applying to all employed persons in the US, are to be in equity. I could go on and on and on.

did the exact same thing. So ChatGPT's going to be our next president?

If they have 5M downloads and $2M revenue in a single month, what they've succeeded at is marketing and activation, which is significantly harder than just vibecoding an app and getting it on an appstore. The app itself sounds like absolute nonsense, but they've managed to make a buck with it.

Unfortunately, Eucalyptus is not native to CA, and has become invasive growth in a lot of areas. If you've driven by giant Eucalyptus groves anywhere up and down the CA coast, then you'd understand why they get out of hand. But yes, getting rid of them would probably help. You left out that they shed and peel huge piles of kindling as they grow!

The Chicago fires would not have taught one how to address wildfire conditions in the ex-urbs and greater metro areas built into the hills surrounding Las Angeles. The actual urban areas of Las Angeles *did not burn*. The conditions that caused the severity of the 1871 fire have largely been addressed everywhere already.

In which municipality was Adam Carolla the fire chief, and for how many years? Oh right, none, so he's been talking out of his ass for over two decades. Got it.

..or extreme weather and Santa Ana winds. California doesn't have incompetence or underfunded fire depts (see https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fgavinnewsom.com%2Fcalifo...) that would increase wildfires, but yes, those could certainly be a problem elsewhere.

First off, please stop spreading fake news. Newsom has already addressed all of your misinformation/disinformation here: https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fgavinnewsom.com%2Fcalifo...

A lot of the property up in the hills is large private lots, so who is going to clear the brush? California already has the leading forestry maintenance in the U.S. in terms of spend and expertise, and that includes clearing brush. The right-wing media apparatus just heard of this term, "brush clearing", and decided California doesn't do it, and that that incompetence exacerbated the fires. In reality, winds were carrying cinders hundreds of meters in all directions due to hurricane-force winds, and it didn't matter what brush you cleared because the cinders were going to reach the next flammable thing, no matter what.

California had plenty of water and full reservoirs. Having the water is different than *delivering* the water. There is no human-made system on earth that can instantaneously move a reservoirs' amount of water from one place to another, just like there's no road system on earth that can move a full stadium's worth of cars across the city at the same time. You don't build a municipal water grid to handle 100x the volume of water that a community could ever use, because the cost to build and maintain it would be prohibitive. Similarly, when dozens of crews were fighting fires all across the LA hills, the rate of water exceeded what the water grid could deliver, because it was a circumstance never before encountered.

Another observation is that there was a correlation between companies that made political donations to Trump and co, and those who received exemptions from tariffs. So, stupefyingly, it appears that some of these broad tariffs are just a very basic protection racket. Trump raises the price of *everything* by 20%, then says to literally everyone, "hey, that's a very nice business you got there, shame if you had to pay 20% more in taxes. Why not kick us a few million, then you can "enjoy" the higher prices while we split the difference?"

How much would your cost to cook your own food change if you didn't have a car, didn't live near a grocery store, didn't have a kitchen, and didn't have time to cook? Additionally, at what point in time did your time become worthless? If you add in the cost of your time to shop and prepare and clean at an hourly rate of, say $25/hr. Now your meals are closer to $20-30 each, and the fast food (with the same time/opportunity costs baked in) is coming in around half that. Having the time and agency to prepare your own food is a luxury not everyone shares. So, first start with an honest accounting, then second, be thankful that *you can afford to save money* by translating your time and effort into an efficient dietary regime. For many people, your routine is enabled through luxuries they simply don't get to enjoy.

Does Van Tulleken have an actual definition of ultra-processed foods? I've read about a dozen books on food politics, food manufacturing, industrial food supplies, etc. by various nutritionists, food anthropologists, and investigative journalists, and I don't know what ultra-processed foods is supposed to mean. Food processing occurs on an infinite spectrum, and there is either some cut-off point after which food goes from 'processed' to 'ultra-processed', or it is a label with little hard-and-fast meaning. Or, maybe, there is just some concise definition, but I have yet to encounter it (it seems relatively recent in the food politics/industry popular literature).