bomewish

Yeah it felt like claudeslop to me too — and such a strange choice. Maybe dude dictated it and said Claude to write it up. Save some time. Even so.

I also really dig the model and it’s new and super useful for another take. But google is so awful with basically everything else. They can’t get out of their own way.

Prospects look grim for these kind of light agent wrappers …?

Just a giant hetzner bare metal we’ve been on for years before they jacked up the prices - thank goodness. Avg record is maybe 1-2k words. Search is very fast, but it’s an internal product and so we’re not dealing with a large user population. Anyway parade all the way. You get a proper relational database but also proper fts without having to deal with elastic etc. Just a dream, so glad they came along.

Love to see this. Am a huge paradedb user (we running over 100m records in prod, index is 4tb or so) and I just loved this read.

I really like the idea of turning the search index into a public utility. It is one of those natural monopoly coordination problem things. Just quasi nationalise it for economic efficiency. Ofc google can still sell adds against their own ui (like everyone else). Hopefully this move will move that idea closer to reality.

It may also have been a social test — ie the polite thing to do in that circumstance is to say “oh wow, I thought I knew that one, weird. I’ll have to look that up again thank you” even if you know 100% that you’re correct. Weird social expectation thing.

This seems great tech but the obvious obvious LLM slop writing is just so incredibly off-putting. We read LLM slop everyday ourselves, but now we have to read it everywhere else too. Is this stuff so hard to just write oneself ??

Bit frustrating. any kind of benchmark or even a few examples showing that this produces better outputs would have been nice.

Plex + Tailscale = access it from “localhost” wherever you are. Just set up Tailscale serve to serve it in your tailnet.

Problem solved. Plex is free again

The “from scratch” is doing a lot of work here!

Forgive my ignorance. It sounded super yuppie.

Seriously ? These are not sympathetic! The first seems like a whole lot of pointless faff for nothing much at all — who needs an emulsion made out of salvaged organic chickpea bubbles — and the second complains about not being allowed to raise prices… yet she can! It’s just that you’d get fewer customers.

Also - this is presumably profit after all the wages are taken out including for owners of the place? If so it kinda depends on what their wages are to know if this whole situation seems super unfair.

And at least save us the pain of trying to scroll through the bs intro. Just put the 10 or whatever bullet points right there.

He was appalled and said, “but that’s a lot more work for us.”

This made me really, really angry.

Why would yc fund them given how obvious the risk is ? Esp since a Mac mini farm is capital intense.

I really think people should learn first and primarily from books written by people. Think about it: the LLMs are trained on that data. So they cannot actually be better than the best original texts. Get good quality original books then learn them in conjunction with AI, but don't try to learn using this. How to downvote?

So two different parts of the state use favoured journalists as methods to publicise information that they could not make public directly. When they do this they also get to shape the framing.

What a rather unpleasant business.

Guy couldn’t even bother to write his own damn post mortem. My goodness. No wonder they got owned by the ai.

If you cared so bad you could make your own evals.

Doh. I went in expecting a really cool thesis — because the idea seems somehow intuitive, or at least really intriguing. But I have no clue what I read. Just totally odd and unconvincing. Greenland? Dialectal substrate? The idea is still super intriguing to me though!

Isn’t that a completely bizarre metric though in this instance??! It is specifically the revenue generating arm of the government. If it wasn’t running at a “surplus” that would be very concerning indeed.

Kinda… funny? that one in this position could not just blurt out “they offered the most”?

Best thing about windows and biggest thing I miss. Have never been able to find equivalent for Mac — stuff that comes close but really not quite the magic of Everything. Same w Total Commander. Sad!

What are the much more effective and efficient ways — since you said it ?

Great read.

Article seems heavily written by Claude. Gets kinda annoying after a while.

So could this safely be used on Tailscale then ? I’m very curious though also a bit paranoid.

Gives you a randomised domain name for your service so it’s not exposed to the internet on the url that has already been publicly exposed.

Hey really recommend using a big long random string in that URL, because as you will have read above TAILNET NAMES ARE PUBLIC. You can find them here: https://crt.sh/?Identity=ts.net [warning, this will probably crash browser if you leave it open too long -- but you can see it's full of tailnet domains].

So anyway try it like:

tailscale funnel --set-path=/A8200B0F-6E0E-4FE2-9135-8A440DB9469D http://127.0.0.1:8001 or whatever

I use uuidgen and voila.