zzleeper

I was in that spot a few weeks ago. My requirements were not huge but a) I was on Windows and b), didn't want to waste too much time setting it up.

Tried a few DBs that didn't work well (e.g. I think it was ChromaDB that didn't support Python 3.12) and ended up picking LanceDB.

Very simple onboarding (just built on top of parquet) but there are a few rough edges.

Curious how it compares with qdrant for non-crazy problems

Can you do this easily? Without paying $$ in transaction costs due to shorting (plus margin)

I wish there would be a way to just buy sp500 minus a list of firms, so I could invest in a diversified way without going into Uber and friends (or any other firms that make profits on paper without the cash flow to back it)

I wanted to point this out, albeit slighty less cynically. NickB invested a lot of capital into becoming the expert on WFH (see e.g. https://wfhresearch.com/project-team/ ).

Whenever I see this in other economists, more likely than not their findings tend to go towards where their capital is invested.

> the Fed can correctly model how ...

I wish :(

Also around DC area, the leasing office folks told me that the price changed dynamically and could only be overruled by the HQ. They said that many companies use that software in order to obtain their "optimal" price for the area, which is of course observationally equivalent to colluding with extra steps.

BTW, more recently I saw the same thing with large contractor companies. One quoted me an insane price for something and I asked why, and they replied that it's the price that the software (ServiceTitan) tells them to charge, and even showed me their screen.

Again, they said ServiceTitan decides the price depending on "the average price of other companies in the area" which is code word for you-know-what.

That's nuts. I had to do 100-hour weeks a few times in my career, and felt that any time above that had zero or even negative productivity.

Is it really necessary to do so many slide decks in order for IB to be as productive?

You made me search for 2022-23 studies on HCQ, to see if anything changes.

Behold, still fucking useless: https://www.drugs.com/medical-answers/hydroxychloroquine-eff... (one of many meta studies). Where do you get that disinfo??

Random q but what car did you get? I'm looking at replacements for my 11-yr old Hyundai 5-door hatchback :)

I heard a common complaint was wipers

Where I live (~DC), daycare costs 2.2k per month. With two kids and a 35% marginal tax rate (including state taxes and Medicaid/etc) you would need to make $81k per year just to break even.

Add up extra food costs (vs cooking at home) and extra cleaning costs ($140 every two weeks so about $4k and the value prop of having one SAH parent doesn't look as bad (which is what several college-educated friends have done, sp the ones with 2+ kids)

When you said "at age 16" it made me feel the guy was 40 or so and it was long ago.

Not that the cheating took place only two year before the game against Carlsen, which is close enough to keep the same behaviors..

Revenue US$1.39 billion (2022)

Operating income US$–882 million (2022)

Net income US$–921 million (2022)

Total assets US$7.83 billion (2022)

Total equity US$3.53 billion (2022)

Number of employees 7,703 (2022)

(From Wikipedia)

Holy F these numbers are tough, sp in a high rate environment. Why would unity need almost 8k employees??

That federal program, even with high deductibles, completely subsidizes their customers at the cost of taxpayers. They massively underprice the hazards, at the taxpayers ends up being on the hook at the cost of people who decided to built (and rebuilt) "their house on the sand"

It seems you want something like a Fractal Design R5: https://www.fractal-design.com/products/cases/define/define-...

(or a larger alt like this one: https://www.fractal-design.com/products/cases/define/define-... )

That said, be aware that these things are heavy and bulky, as other commenters posted. I used to have one but replaced it with a SFFPC and have been quite happy with the change.

Do you have any suggestions about how to start implementing something like this in-house? I'm sitting on thousands of PDFs (that can be trivially turned into text) and it would be really useful to train an LLM on them for information retrieval.

But the dev and computing cost of this feels so huge that I'm not even sure where to start.

Stupid question from an outsider, but is it possible to grab a pretrained model (most likely one of the "camelids"), feed it your own data (I have about 1k documents that cannot leave my network) and use it for fast-but-sometimes-wrong information retrieval?

But then engaged parents will take their kids to Kumon, and rich parents will take their kids to private school.

Their kids will then mop the floor against all other kids, including engaged kids without engaged parents, that would have benefited from more advanced classes. Is that equality?

Are you sure? I've used Reddit for 15 years and it's the first time I don't really have a good way of using it, besides the desktop (on old.reddit).

On mobile (where I used to spend reddit most time) their own app is so crap I don't even have it installed, so I have no easy way to browse. On Chrome I'm not even logged-in, so I only use it whenever a google search leads me there, but I don't post or do anything meaningful.

All in all, it feels like a really stupid move. If you are going to kill the apps, then have a decent alternative. Otherwise, just charge them a profitable but not stupid amount. And if reddit is losing money, maybe just look at their headcount and start there...

I think it depends on the data.

If it's presorted by the join variable then rolling the loop is faster. Also, if the index is too big for memory, then it might be faster to loop.

I follow a few folks writing on war in Ukraine, but since maybe Nov-Dec they all disappeared from page, and I have to go all the way to the bottom of the barrel to see anything tangentially related. Only after the algo was released I understood that was a decision by Elon, for whatever reason...

I used to be addicted to Twitter, checking it 10+ times a day at the very least. Now the content I get is awful, and I only end up there whenever I follow someone else's link :/

Besides that error; the images are each above 1mb and very low quality. I believe if you just save them as pngs you should get much more quality at 20% the size.

> They gave me a mortgage rate much better than any other bank was offering

Isn't this why they got in trouble on the first place?

Vtech is pure evil, they are the most useless annoying toys; I would prefer giving kids YouTube than let them play with those unless it's with a hammer

Can you provide an example of what prompts would you use?

Agree it's QE. For those that say this isn't QE, it's understandable given the "original sin of QE" as Ricardo Reis put it: https://twitter.com/R2Rsquared/status/1576164361898708998?t=... (this episode is QE type 3 in his taxonomy)

Amazing; thanks a lot!

Hi Mark,

One question out of academic curiosity: I'm exploring ways to use these tools for research projects in econ and am struggling to see a good angle.

For instance, suppose I have lots of PDF reports on how firms have evolved on each quarter (10,000 reports or any other number beyond what I can read).

Can LLMs be used to spit out variables based on these reports? EG: indicator variables (optimistic-vs-pesimistic), categories, etc. that I can then use as data to test economic models?

I tried simple cases with ChatGPT3.5 and got the feeling it's great for outputting narrative text, but felt mediocre for when the output was more narrowly defined into categories (which was surprising).

> which is the largest city in Tanzania. (It was the capital until 1974, when it was replaced by Dodoma.)

Find it funny that even the humans got it wrong. Dodoma only became the official capital in 1996.