tylervigen

Two years ago I re-vamped my "Spurious Correlations" side project, which is mostly just a bunch of charts. However, I couldn't find a charting software I liked that would display clean, simple visuals with the constraints I wanted. (I had used pCharts and HighCharts in the past, but didn't like charting in Javascript or PHP.)

I decided to "roll my own" and write Python scripts that outputted SVG markup. I was worried this would go about as well as every other "roll your own" project does, but was pleasantly surprised. It is surprisingly easy to output reliable, good-looking SVG graphics using Python. If you are making a chart, everything is just math.

The infinite scalability is almost just a happy upside to the simplicity of creating the visualizations, which is annoying in raster format. It made me like SVG even more.

> Google Maps Is Not a Directory. It’s a Market Maker.

I understand the author's meaning, but this isn't what the term "market maker" means. To "make a market" is to stand ready to buy and sell, usually a security, in order to create liquidity in a market. Usually this resolves the issue of timing, because it is unlikely that someone wants to buy at the exact moment someone else wants to sell.

So to "make a market" in London restaurants, Google could buy food during the day and sell it at night when the shops are closed but people are hungry. (This would be silly.)

Perhaps a more precise term is "algorithmic gatekeeper."

This is awesome. What a way to do a relaunch. I was a Pebble user in 2013 on first release and saddened when it all broke down. I was so turned off from that experience that I swore off smart watches (I wear a Casio now).

This could get me back, though I’ll admit the appeal has gone down since I’ve realized how nice it is to create separation between me and my notifications.

The researchers found a common ancestor of many different species that all conduct non-functional kissing. The surprising thing about the research is that it specifically implies kissing is not cultural!

I think it is @ given the context of the next paragraph, where they complain that @ doesn't work well in the grid.

The MVTA in Minnesota operates with 90% subsidies, so only 10% of revenue is from fares.

It feels like there could be some societal benefit to similarly reducing the number of busses and just making them free. (Today most busses are only at 10-30% capacity). This seems to support that idea.

Your experience is at odds with the statistical results of the linked studies covering nearly 100,000 participants.

Indeed, and I think this is a diredt result of OP's pipeline. Part of the workflow involves asking an LLM to prioritize articles that readers in 2025 will find interesting in hindsight.

It only requires that one parent has enough free hours to assign coursework. They don't have to exit the workforce, and don't necessarily need to directly supervise learning (but of course some of this is necessary for K-5).

I think a lot of how homeschooling can work, along with much of median/lower household income life in general, is misunderstood.

Source: Was homeschooled by a mom who worked.

I think Nano banana pro’s answer to the giraffe edit is far superior to the Seedream response, but you passed Seedream and failed NB pro.

Maybe that one is just not a good test?

Microsoft Teams does this: https://techcommunity.microsoft.com/discussions/microsofttea...

Yes, that complexity is what my acknowledgment was meant to refer to.

I disagree with your logic. Increased mean error 72 hours out (vs 24 hours out) is not an indication that GFS is getting worse over time. At that scale it’s obviously getting better over time; 24 hours out is further in the future than 72 hours out.

However, an increase in the mean error at the same time out year over year (or between 2005 and 2025) is an indication of an issue, and that’s what we see.

That is not entirely true. At least some of these tests (like HLE and ARC) take steps to keep the evaluation set private so that LLMs can’t just memorize the answers.

You could question how well this works, but it’s not like the answers are just hanging out on the public internet.

This comment was moved from another thread. The original thread included a benchmark chart with ARC performance: https://blog.google/products/gemini/gemini-3/#gemini-3

I agree that for any given test, you could build a specific pipeline to optimize for that test. I supposed that's why it is helpful to have many tests.

However, many people have worked hard to optimize tools specifically for ARC over many years, and it's proven to be a particularly hard test to optimize for. This is why I find it so interesting that LLMs can do it well at all, regardless of whether tests like it are included in training.

I am personally impressed by the continued improvement in ARC-AGI-2, where Gemini 3 got 31.1% (vs ChatGPT 5.1's 17.6%). To me this is the kind of problem that does not lend itself well to LLMs - many of the puzzles test the kind of thing that humans intuit because of millions of years of evolution, but these concepts do not necessarily appear in written form (or when they do, it's not clear how they connect to specific ARC puzzles).

The fact that these models can keep getting better at this task given the setup of training is mind-boggling to me.

The ARC puzzles in question: https://arcprize.org/arc-agi/2/

It's the subtitle of the article they linked to.

But to expand: the US flagship forecast model just had its worst year predicting hurricanes since 2005. The trend of errors over the last few years hasn't been great.

> I don't feel as if I get any more accurate data now than I did back when my parents turned to the daily weather channel for the forecast.

The accuracy improvement is provable. A four-day forecast today is as accurate as a one-day forecast 30 years ago. And this is supremely impressive, because the difficulty of predicting the weather grows exponentially, not linearly, with time.

You are welcome to your feelings - and to be fair, I'm not sure that our understanding of the weather has improved as much as our computational power to extend predictions has.

For lay-users they could have explained that better. I think they may not have completely uninformed users in mind for this page though.

Developing an ensemble of possible scenarios has been the central insight of weather forecasting since the 1960s when Edward Lorenz discovered that tiny differences in initial conditions can grow exponentially (the "butterfly effect"). Since they could really do it in the 90s, all competitive forecasts are based on these ensemble models.

When you hear "a 70% chance of rain," it more or less means "there was rain in 70 of the 100 scenarios we ran."[0] There is no "single accurate forecast scenario."

[0] Acknowledging this dramatically oversimplifies the models and the location where the rain could occur.

I’m not sure your point is warranted given there is a year between today and November 2026. Surely we can be more precise than a year?

> Inspired by the concept of “a piece of cloth”

I thought this was an April fools joke.

And for anyone who wants to be nerd sniped by this one comment chain, the Acquired podcast recently did an episode on Epic and it is great.

Title: "Epic Systems (MyChart)."

It's not separate. That would be included in manufacturing, transportation, and electricity.

I think the problem with this argument is that it’s like saying “don’t worry about nuclear weapons, worry about the people who will use nuclear weapons.”

Sure it’s true, in the absolute sense, but it’s more helpful to specify AI than it is to specify humans.

One thing we should be careful about regarding calculations related to the larger set of "all data centers" vs only "GenAI" is that the data centers include all the predictive algorithms for social media and advertising. I, for one, would not want to misdirect ire at ChatGPT that really belongs directed at ads.

I am not the author, but can speak as someone who has kept a "seemingly simple" semi-popular web service online for many years with no/low maintenance: it often comes down to the fact that issues tend to pop up just when you don't have time to work on them. It is usually not a lot of maintenance, but it feels like a lot when you have to dive back in to something that you haven't thought about in months. It ends up being a source of very low level chronic stress.

I can totally see why shifts in other priorities would make it attractive to decommission.

Probably farther than that, right? Plenty of natural disasters, including floods and wildfires, can affect an area larger than 1 km.

This site is a front for Miome.

The home page cites IBSprobiotics, which cites Bailey Hanna MS, who is employed by Miome according to LinkedIn. Thus, it is no coincidence that if you follow the link to the top probiotic and then look at which products contain it, all the ones available in the US are distributed by Miome.

This is wildly deceptive marketing.

This seems way above the clearing price for garage sale signs.