Hacker Newsnew | past | comments | ask | show | jobs | submit | oa335's commentslogin

> If you play all the tricks Google's Capacitor does to the data layout

can you please elaborate? does duckdb support querying custom storage formats?


> does duckdb support querying custom storage formats?

No, this is still parquet being loaded in.

The "trick" is running through sort order combinations which result in maximum packing in a Dremel style data format (like Parquet).

Here's the trick explained in enough detail to understand, but not enough to copy as-is

https://cloud.google.com/blog/products/bigquery/inside-capac...

Search for the section with

"Capacitor builds an approximation model that takes into account all relevant factors and comes up with a reasonable solution."

or a more detailed dive in the paper (by the favourite author of performance engineers, Daniel Lemire)

https://arxiv.org/abs/1207.2189


> I took a photo of the clock, sent it to hermes,

How precisely do you do this? I'm interested in setting up a similar system, and I keep getting tripped up on how best to share information with the agent in a way that doesn't compromise security but is still easy to use.


Tailscale on a phone, on a server, and then whatever you and your agent come up with. I had a web app I built to talk to it. One day I wanted to share files and I just told it to implement that feature.

I don't use much off the shelf anymore. It's too easy to make my own hyper customized versions.


what app do you use on your phone to communicate with yoir server? just a terminal emulator?

I built an orchestrator app with a nice voice interface so you can just talk to it about all your work in progress - which are sessions in herdr on your machines. That way you can do a lot just by talking to it through phone but when you're back at your desk you see it all in herdr. It's in beta tests with mobile webUI but iOS and Android should come very soon https://pitupitu.com

Just ask it to build you one. The era of hyper-focused personal software is just beginning.

I use the WhatsApp bridge

why?


Why would you consider a project which appears to be fully coded by Claude worthy of a Show HN post?


Why does Claude output being shown to humans mandate it be rewritten before before consumption? What school of thought is that from?

Snark humor: We are not in The Name Of The Rose and this code is not Aristotle's forbidden second volume of Poetics which is antithetical to the teachings of Christ and is deserving to being hidden in a labyrinth by Benedictine monks.

Absolutism as a philosophy sort of supports fanaticism. Absolute Certainly could be the religion which is being formed here during this onset of AI challenging humanity's utility.


because it is useful and interesting. why should they rewrite the project without LLM input?


what a horrible attitude, especially for a forum dedicated to "hackers".

the code is open source, why dont you show us what a quality test suite looks like?


Oh my. Don't you think that code generated by an LLM has nothing to do with "hackers"? And once/if you realize that, don't you immediately see that they're right? Maybe not in the way they say it, but otherwise - they're absolutely right.



"There's apparently only ~2000 Uzbek-Americans in all of northern IL"

where did you get this number? does that include chicago, because i live in Chicago area ... seems low.


https://www.neilsberg.com/insights/lists/uzbek-population-in...

Figuring some fudge factor too. It could be low - it's hard to find numbers for nationality counts for central Asians broken down by state.


that websites estimates are definitely too low - e.g. there are definitely more than 40 Uzbek people in Chicago... I've been to 3 uzbek restaurants in Chicago proper myself.


Broadly (in another comment I made) I think I've made sense of it. Regardless of the real population, it's probably not just the Uzbek population sustaining these places. That most likely wouldn't be viable.

My wife's restaurant for example targets a nationality that is wayyyy more numerous (not Mexicans; that'd be easy enough), and it's tough for us to stay profitable.


> "A company whose first demo was completely fraudulent"

Could you expand on this?


Agree 100%. Costco exemplifies american dream... recent immigrants perusing well-stocked aisles, friendly employees, ample parking, cheap tasty hot dogs, etc.


Perhaps this depends a lot on location. Parking is a nightmare at my local Costco. The employees are friendly enough most of the time. I truly admire the value and business model but Costco is pretty much the absolute worst shopping experience I can think of.


It's because they are so busy. Yeah, costco is the one store I go to that people seem to have 0 spatial awareness. Every time I go, someone will just stop moving in the middle of the aisle, creating backups, without any shame. But I bet if your other local grocery stores were as busy as costco, they would have the same issues.


That's interesting because my experience is the opposite. Costco can swallow up an irrational number of people and still be pretty smooth inside. The parking lot can be packed and when you walk in the door it turns out to be fine.

Walmart, on the other hand, is the absolute pinnacle of shitty shopper behavior. People being rude, pushy, completely oblivious standing in the middle of the aisle, etc. At least in my area, Costco attracts a different kind of clientele.


"the absolute worst shopping experience I can think of"

Indeed:

* Short opening hours

* checking the membership card at the entrance

* cell phone salesmen inside

* item locations change randomly

* items randomly disappear for weeks at a time, as if Leonid Brezhnev was in charge of restocking

* waiting at the exit for someone to check your receipt

* for a while, they'd check your receipt at the self-checkout and then check it again at the exit

* 1970s level IT, inaccurate inventory information

Microsoft used to give a cash bonus for patents; I imagine Costco has a bonus program for new ideas to make the shopping experience worse. Some ideas:

* throw lit firecrackers into the crowd

* pour oil on the floor so people slip

* complete darkness inside the store

* every 10th person at the checkout has their credit card cut into pieces


Agreed. The entire store is wired to encourage impulse buying and keep you from making rational decisions about whether you'll be able to finish that 3lb container of guacamole before it goes bad.


I don't see how Costco prevents you from making rational decisions. If anything the incredibly plain atmosphere (wide open spaces, no flashing lights, no loud music) feels much more calm than the typical retail experience.

My average food waste from a Costco trip is significantly lower than other grocery stores.


Also Costco has a really annoying "treasure hunt" strategy where goods are re-arranged around the store trying to get you to explore aisles (aka get lost and waste time) to impulse buy more stuff.

Other stores (Target, Walmart, etc) will let you look up the item's aisle in their app and be considerate of your time.


Costco's app will let you search inventory, but not let you see location or quantity.


I love Costco, but I hate that they constantly move shit!


I find the experience of shopping at Costco very uncomfortable. The parking lots are jammed packed, everyone is darting around with large shopping carts, the lines for the cashier are long, sales people are trying to pitch me on travel deals as I walk by - it almost feels like going through a busy airport. I am a Costco member, but I only go to the store when I really need to. The fact that I can shop Costco via Instacart was a gamechanger for me.


ample parking? Not at any locations in my area. Even on weekdays.


Go within an hour or so of opening.

I used to work right across the st from one and would spend most of my shift looking out at their parking lot and you could see it get more packed throughout the day, thin out a little bit in the early afternoon and then slowly drain towards closing.

It's always least crowded right at open and then an hour (? or maybe two?) later they open for the "regular" people and once that's the case, it fills quickly.


> Go within an hour or so of opening.

Hah. It’s such a PITA that Costco includes an hour earlier entry on the top tier membership.


I regularly shop at Costco and its usually one of five which ever is more convenient at the time. In all of my experience with Costco there is exactly one that has shitty parking, the rest are fine. That is if you have enough wrinkles in your brain to notice that in front of the store there 10 cars waiting for spot and parking a little bit in a back is easier.

I often see people cruising around still looking for parking while I already managed: to park, walk to the storefront, get myself a hot dog, eat my hot dog, grab a cart.


I park in the back any time I go to Costco. the locations nearest to me are always packed to the brim. even the sort of "secret" extra spots that are behind the building and could be considered employee spots.


My local locations are busy even during those times. it's a constant battle.


For lots and lots of Americans, endless consumerism, driving and junk food are not a part of the dream in any way.


The hotdogs. You know the world is ending when Costco raises the prices of their hotdog.


> friendly employees

Where are the employees? There are so few employees (other than cashier) on the ground in a football field sized warehouse. Good luck finding someone if you have a question


> it costs >1000% to run inference

do you have a source for this claim? i thought LLM providers earn high margins from inference (charged by token). is this no longer the case?


if a $6000000 cabinet can generate 10000/s tokens of Opus but only 1000/s tokens of Mythos, then Mythos costs 1000% to run no matter the markup.

no one has a source, because no one knows closed model parameter counts. we have only heuristics which strongly indicate that Mythos is simply a big fucking model that any other lab could make an equivalent of.


This was just theorised. The leaked OpenAI financials suggest otherwise (because of shady naming of losses)

The only ones who seem to profit are the ones running smaller Chinese models. Even NVIDIA seems to have to "reinvest" their profits into sponsoring companies to buy their cards now.


"Our most likely estimate is that replacement costs are approximately $91 billion. Military and strategic assets — nuclear facilities, missile and drone production infrastructure, air bases, naval vessels, and air defense systems — make up roughly half that total, at approximately $46 billion, which is equivalent to 4-6 years of Iran’s pre-war defense budget"

How exactly did they arrive at this estimate? I would like to see a building-by-building breakdown, or at least a methodology. Feels very vibe-researched tbh.


You could get to a pretty good figure by assuming the Iranians know they can ask for at least twice as much. Three times as much isn't even pushing it.


More is probably better for Trump if it's larger because it's more padding through which the deal can be engineered that "such and such" [friend/proxy of Trump] has to be destination company/supplier of reparations materials or other self-dealing benefits.


Exactly (I made just this point with a bit less clarity than yours, in another comment)


Consider applying for YC's Winter 2027 batch! Applications are open till November 2.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: