And if we were talking about “the underlying theory of machine learning”, you might have a point.

The key component of making good sneer club criticism is to never actually say out loud what your problem is.

I wrote 800 words explaining how TracingWoodgrains is a dishonest hack, when I could have been getting high instead.

But we don’t need to rely on my regrets to make this judgment, because we have a science-based system on this podcast instance. We can sort all the SneerClub comments by most rated. Nothing that the community has deemed an objective banger is vague.

An anti-recommendation from another thread:

Having now refreshed my vague memories of the Feynman Lectures on Computation, I wouldn’t recommend them as a first introduction to Turing machines and the halting problem. They’re overburdened with detail: You can tell that Feynman was gleeful over figuring out how to make a Turing machine that tests parentheses for balance, but for many readers, it’ll get in the way of the point. Comparing his discussion of the halting problem to the one in The Princeton Companion to Mathematics, for example, the latter is cleaner without losing anything that a first encounter would need. Feynman’s lecture is more like a lecture from the second week of a course, missing the first week.

Come to the Sneer Attractor, we have brownies

“… And litterin’.”

Every task you outsource to a machine is a task that you don’t learn how to do.

And school is THE PLACE WHERE YOU ARE SUPPOSED TO LEARN THINGS, JESUS H. FUCK

I’ve seen some outright blatant crank shit (as opposed to the crank shit that works hard to masquerade as more legitimate science) pretty highly upvoted and commented positively on lesswrong (GeneSmith’s wild genetic engineering fantasies come to mind).

Their fluffing Chris Langan is the example that comes to mind for me.

gwern: It’s not “AI slop” if I wasted hours dicking around with MidJourney to make it.

rsaarelm: People don’t appreciate the beauty of Substack’s built-in slop generator.

gwern: “I refuse to submit to the tyranny of the lowest common denominator and dumb down my writings or illustrations.” Have you appreciated the depth of my artist’s statement?

I’ve never tried a Pyrex roux before. I’ll have to give that a shot. Often, I use our Pyrexen to rehydrate textured vegetable protein. Scoop a couple cups from the giant box in the pantry, add a couple teaspoons of stock concentrate (e.g., the Better Than Bouillon veggie and roasted garlic flavors), add water until the granules start floating, stir, microwave 30 seconds, stir, microwave another 30 seconds. Then it’s ready for skillet-frying with whatever spices and other flavorings seem appropriate in the moment. Chili powder, red pepper flakes, cumin, oregano and a dash of cocoa powder makes for a good Tex-Mex flavor profile that can sub for ground beef in tacos, enchiladas, etc. Soy sauce, mirin and sugar or agave is a straightforward teriyaki. It’s pretty versatile stuff.

The Totole “Granulated Chicken Flavor Soup Base Mix” is another good flavor boost.

Blockquote glitch?

From the comments:

If Said returns, I’d like him to have something like a “you can only post things which Claude with this specific prompt says it expects to not cause <issues>” rule, and maybe a LLM would have the patience needed to show him some of the implications and consequences of how he presents himself.

And:

Couldn’t prediction markets solve this?

Ain’t enough lockers in the world, dammit

Of course, commenters on LessWrong are not dumb, and have read Scott Alexander,

It’s like sneering at fish in an aquarium

Here’s my recipe for blueberry bread that I make for parties and such.

And here’s my experimental recipe for yu hsiang eggplant (still in progress; this was my most recent attempt).

“They don’t need to develop protocols of communication that facilitate buying castles, fluffing our corporate overlords, or recruiting math pets. They share vegan recipes without even trying to build a murder cult.”

Longtime friends of the pod will recognize the trick of turning molehills into mountains. Creationists take a legitimate debate over a detail, like how many millions of years ago did species A and species B diverge, and they blow it up into “evolution is wrong”. Hossenfelder and her ilk do the same thing. They start with “pre-publication peer review has limited effectiveness” or “the allocation of funding is sometimes susceptible to fads”, and they blow it up into “physicists are a cabal out to suppress The Truth”.

One nugget of fact that Hossenfelder in particular exploits is that the specific way we have been investigating the corner of physics we like to call “fundamental” is, possibly, arguably, maybe tapped out. The same poster of sub-sub-atomic particles that you’d have put on your wall 30 or 40 years ago is still good today, with an edit or two in the corner. We found the top quark, we found the Higgs, and so, possibly, arguably, maybe, building an even bigger CERN machine isn’t a worthwhile priority right now. Does this spell doom for physics? No, having to reorganize how we do things in one corner of our subject after decades of astonishing success is not “doom”.

His group chats with Kevin Roose must be epic.

Just earlier this month, he was brushing off all the problems with GPT-5 and saying that “OpenAI is learning from its greatest success.” He wrapped up a whole story with the following:

At this stage of the AI boom, when every major chatbot is legitimately helpful in numerous ways, benchmarks, science, and rigor feel almost insignificant. What matters is how the chatbot feels—and, in the case of the Google integrations, that it can span your entire digital life. Before OpenAI builds artificial general intelligence—a model that can do basically any knowledge work as well as a human, and the first step, in the company’s narrative, toward overhauling the economy and curing all disease—it is aiming to build an artificial general assistant. This is a model that aims to do everything, fit for a company that wants to be everywhere.

Weaselly little promptfucker.

Louisiana has to build three new natural gas power plants to accommodate the “AI” data center that Meta just crammed through because said center will use Three Times as much electricity (and, thus, attendant resources) as the Entire City Of New Orleans, every year.

https://bsky.app/profile/wolvendamien.bsky.social/post/3lwyxhchxos2g

A bit of jank I have noticed on mobile: If I use the “view votes” feature of the mod menu, after I close the list I cannot scroll the page up or down. I have to close the tab and then undo, and then everything works again. (Am still on Firefox for mobile.)

I am not sure that quantum computing’s having been advertised as a code-breaking technology would actually give it a bad rep. Quantum information science has also been advertised as enabling key distribution, for even longer (1984 versus 1994). The quantum giveth and the quantum taketh away.