bcoates

I am doing the review in good faith though--by default, I scroll past to the first result, then if it seems unsuitable and I'm desperate enough I check the LLM thingy. If it were providing any value it would sometimes be both novel and correct.

Usually it's non-novel (correctly-harmelessly but unhelpfully, restating the web search results). When it's novel it's because it's wrong.

I would remember the situation where reading the LLM thingy added any value if it ever happened. The weird little UI thing they do where they only show the LLM result if you wait for it to render makes this very easy, I have to scroll up to even consider it.

Honestly that might be a mistake, when the consensus is greedy get scared and when it's scared get greedy

Every once in a while I bother not ignoring a Google AI overview, then I waste some time fact-checking it and find out it's wrong. Most recently about a python library (where it hallucinated a function that doesn't exist, complete with documentation and usage examples) and breaking news (where it authoritatively said [non-culture war, non-controversial, local] thing doesn't happen, above a dog-bites-man story from a conventional news source about how thing happened again)

So why does the little guy gallop forward when you turn the brain on?

Reminds me of the excellent sci-fi novel Semiosis

This article leaves me super unclear on the metallurgical process going on here--you fire gold ore on a bed of glass rubble and the impurities are adsorbed into the ceramic or ???

[flagged]

The only winning move is not to play. My grandfather tried his hand at chicken farming about 100 years ago (didn't go well). It was a bad business then and it’s a bad one now.

I use the local LLM-based autocomplete built into PyCharm and I'm pretty happy with it

If you're already built around Redis I'd just keep using it, but if you're doing new development there's not so much a single drop in replacement as a substantially better alternative for any given feature (and not particularly any advantage to having all your data in the "same Redis instance"). That said, 90+% of the time the answer is probably "transactional SQL database" or "message queue"

For Kafka, the answer is probably an object store, a message queue, a specialized logging system, an ordinary transactional database table, or whatever mechanism your chosen analytics DB uses for bulk input (probably S3 or equivalent these days). Or maybe just a REST interface in front of a filesystem. Unless of course you truly need to interface with a Kafka consumer/producer in which case you’re stuck with it (the actual reason I've seen for every Kafka deployment I've personally witnessed in recent history)

Oh God people are still using Mongo in production? Why?

Kafka exists but is deeply obsolete and mostly marginalized outside of things with dependencies on the weird way it works (Debezium, etc)

I've always liked Redis but choosing it as a core tech on a new product in the last, say, 6 years is basically malpractice? 10 if you're uncharitable.

The thing these all have in common is having their economics and ergonomics absolutely shattered by SSDs and cluster-virtualization-by-default (i.e. cloud and on-prem pseudo-cloud). They're just artifacts of a very narrow window of history where a rack of big-ram servers was a reasonable way of pairing storage IOPS to network bandwidth.

Dynamo is and always was niche. Thriving in its niche, but a specialized tool for specialized jobs.

The idea of a medically relevant neutrino dose is absolutely wild

"right, don't do that."

As a non-user of Monero, how do I find out what the security properties are and what information is leaked when various actions are taken? The OP's analysis is deeply lacking in this and the apparent rule against repeated transactions is non-obvious

Perfectly ordinary terrestrial mammals like elephants have much, much larger skulls at birth than humans, so it’s clearly a matter of tradeoffs not an absolute limit.

In the nosql era the idea that you could run even the basics for a >1m user SaaS platform on an ordinary, free, single-node transactional SQL database would have been considered nuts.

I think the underlying problem is that lots of orgs have siloed out security stuff into its own workflow and its own class of bugs.

It's basically Conway's law applied to the security/feature development split.

So even if they have a build/release procedure with a mature regression test suite it probably wouldn't have "security" issues like this in it just as a matter of internal organization

Either I'm wildly misunderstanding or that can't possibly be true--if you sample at high temperature and it chooses a very-low probability token, it continues consistent with the chosen token, not with the more likely ones

Marshall Brain's been peddling imminent overproduction-crisis-but-this-time-with-robots for more than 20 years now and in various forms it’s been confidently predicted as imminent since the 19th century

Maybe I'm nuts, but I absolutely love timesnapper (the non-LLM predecessor of Recall, but the same screenshot every few seconds concept).

I originally got it for it's main advertised function--making it easy to record hours for contract billing--but once I had it running I was hooked.

It's just incredibly useful to be able to pull up what you were doing at any given moment, or how you did a particular thing, a few months after the fact.

I haven't used Recall yet but hooking it up to a multimodal LLM seems like an obviously useful thing.

Lol at the loving description of the manufacturing process for plywood (I think technically pressure-treated oriented strand board, but, you know, plywood)

I don't think the adverse selection problem is anywhere near as serious as the author seems to think--insurance companies aren't selling superior knowledge of their client's situation--they're selling against their client's massively different risk appetite.

I expect home insurance to cost more than it pays out (both in median and mean terms) but I take the negative-value deal to protect against rare financially ruinous outcomes.

Quality underwriting and minimizing adverse selection gives an insurance company a massive advantage over competitor insurance companies but it doesn't make or break the market its own.

I'm also not sold on model provider diversity being the measure of risk diversity-surely most of the risk is coming from application errors and not failures of "safety" tuning of models (which are mostly about preventing LLMs from saying things you wouldn't want in the newspaper--I assume AI E&O isn't interested in ensuring reputation risk)

Wow, you're real proud of not understanding what a TN visa is or how to do the most basic of HR functions.

You’re doing anyone you don't give a job a favor.

100% agree with you that it's worth going back and reading The Books You Should Read as an adult who can just read them as books and not some sort of obligation. That said. some of them are still stinkers that only a weird boomer could love. I'm looking at you, A Separate Peace & fucking anything by Joseph Conrad (sorry).

Half the pleasure of reading these books as they were meant to be read (as books, and not at frogs to be dissected in class) is that you get to discover it for yourself--a mix of life-changing gems and I-guess-you-had-to-be-there meditations on being a failson in the twilight of British imperialism

From the article: It was also a matter of method. Education scholars often narrate the development of high-school-English pedagogy as a clash between two competing schools of thought. On one side is the “student-centered” approach typified by the education professor Louise M. Rosenblatt and her 1938 book, “Literature as Exploration,” which emphasized the resonances between the work and each reader’s individual experience."

I sure hope this is dead and buried. I couldn't imagine anything more dire than literature being reduced to a mirror reflecting back the (presumably young and intellectually deprived) readers sad little life back at them.

I was privileged enough to grow up in what I'll call the LeVar Burton school of literary interpretation: books are a window into a world entirely unlike your own where you can be Zhuang Zhou dreaming he is a butterfly. What’s more interesting: every book being about being a dull little high schooler, or any book being about anything: Farm animals reproducing the Russian revolution, European nobles murdering each other over random points of honor, being totally psyched for war and finding out you’re a giant pussy, navigating the world of being a mentally unstable prep school girl in the 1960s... entire universes of totally inaccessible experiences made possible through the magic of the novel.

Maybe I can rent other people's compute and prompt engineering, a sort of uber-for-openai-for-airbnb-for-x operation

Sure, but I'm gonna assume you're outnumbered by the absolutely deranged like 10:1

Source: look at online reviews of literally anything

A nugget in this that seemed significant:

Quote

“I don’t know if I want to call it a social network, because of the stigma associated with it,” says Ari Balogh, Airbnb’s CTO. So they employ a fuzzier term. “We think of it as a connection platform,” he says. “You’re going to see us build a lot more stuff on top of it, although we’re not an advertising system, thank goodness.” (My own observation is that any for-profit company that can host advertising will, but whatever.)

End quote

Launching a communications tool in 2025 that isn't one of the two overly trod spaces (the advertising-hyperengagement loop of instagram, etc. or the people-you-already-know of whatsapp) is a genuine moonshot in a way that "what if airbnb but for manicures" isn't, and it's something that an incumbent like Airbnb could do that would be impractical for anyone else.

I don't get airbnbs because they're cheaper, I get them because I want to rent something other than a hotel room. If best western had in-laws or 1br quadplexes I'd be happy to stay there, but most hotels I've stayed at are bad rooms in bad locations that almost always include at least one negative surprise every trip, and as you move up to more expensive ones they somehow get fancier but worse.

As someone who doesn't have whatever lead poisoning makes people post reviews on yelp the review system is worse than useless.

I'm not gonna leave a mixed or negative review because snitches get stitches and I can't imagine anyone else reviewing has any less pathological incentives.

1. Build atomic power plants sufficient to supply electricity needs for projected future AI megaprojects

2. Inevitable AI winter

3. Keep running the plants, clean energy achieved, stop burning coal, global warming solved