zzleeper

For me the issue is why there's not a new mini since 5-mini in August.

I have now switched web-related and data-related queries to Gemini, coding to Claude, and will probably try QWEN for less critical data queries. So where does OpenAI fits now?

Are there good open models out there that beat gemini 2.5 flash on price? I often run data extraction queries ("here is this article, tell me xyz") with structured output (pydantic) and wasn't aware of any feasible (= supports pydantic) cheap enough soln :/

If you are in early career (i.e. graduated your PhD within the last 5 years) you are extremely unlikely to get the gifted people visa. The standard approach is to just get the H1B (not the lottery stuff for tech companies but the non-lottery one for hiring faculty at universities). Ask any foreign MIT professor hired early in his career and they went through H1B (and later on, they are more reluctant to move into a place like Florida..)

Wow. The amount of quasi-xenophobic comments in this tread is nuts. They are also a bit misguided.

You don't hire a professor at a R1 school just to teach math101. You hire them they can build a research lab or otherwise help to advance the frontier of the field (cancer, stats, etc.). The talent pool in several of these fields is very very small for Americans, because the brightest just go (used to go?) to work to finance or tech. So if you say you can't bring any bright foreigners, you are constraining yourself to a lower talent pool than other countries, and thus will pay a price (in less research, in no foreign students applying and thus no $$ from them, etc etc)

Thanks for the insights! I'll try Mistral as well.. Gemini worked well for me so far but which model is SOTA is changing quite frequently these days

That's amazing!

I'm working on a kinda similar project (documenting bank runs from historical newspapers) and also opted for Claude to build a static website. Crazy that the two sites have a very similar look and feel: https://www.finhist.com/bank-runs/index.html . The only big difference is that mine lacks a map, which I should hopefully fix soon (I already have lat and lon and am linking to google maps).

PS: Do you know if mistral works better at OCRing handwritten text than gemini 3? Was planning on going the gemini3 for another project

That's actually amazing, and might give me a way to use all the cores I have lying around. 2s per page is an insane 600 pages per minute at 20 cores!

Please do open source it, even if you don't do much around it (worst case I can just spend a few million tokens trying to get opus 4.6 to get it to work)

Surprisingly, I have a few hundred gigs of old newspaper scans so am very curious.

How fast was it per page? Do you recall if it's CPU or GPU based? TY!

I just spent the last 20 minutes trying to debug why my non-www URL wasn't working and my www was. Oh well now I know.

I only see one redaction on page 128 (122 as in top-right of page), and it's just a URL. So there's a rule to redact URLs.

Do any of these models do well with information retrieval and reasoning from text?

I'm reading newspaper articles through a MoE of gemini3flash and gpt5mini, and what made it hard to use open models (at the time) was a lack of support for pydantic.

Don't worry once Central Time wakes up and gets their coffee, this will get <flagged>

The problem is that a relatively small group of people (flaggers) just veto what we see and don't see. This made sense when we relied on flagging to just remove spam, useless posts, etc. but its now being used to remove anything that goes against MAGA.

I'm pretty sure that if you sqldump the list of flaggers of this and other posts (like the MN posts) you will find it's not a uniformly distributed list of users.

That gun looks like the extra gun one of the ICE folks have. Would you be surprised if you find out they planted it out as evidence?

People say there are users who don't want politics on their frontpage, but I think it's just simpler: we have a nontrivial minority of users who are pro-MAGA and like what's going on, so they downvote anything that makes Trump's action look bad. :/

Because that editor was installed in order to be able to censor CBS. They also bought Tiktok, and are trying to buy other media companies in order to be able to better control the flow of information.

Oh god, I'm sure I wouldn't come close to 50%; that's so hard to read

Same! I want to do some data stuff from documents and 2.0 pricing was amazing, but the constant increases go the wrong way for this task :/

Wait so quantum is going to actually deliver something useful within the next 10-20 years??

Hope they don't confuse her boat and blow her up :/

Love this! Would have liked to see something like textract for a pre-LLM benchmark (but of course that's expensive), and also a distinction between handwritten text and printed one.

But still, this is incredibly useful!

A lot of comments here about how diversity is great etc etc. Yes, but also a lot of people value consistency. Now that Win10 is gone and people are reluctant to use Win11, it would be a great opportunity to switch. Yet even I am still reluctant to switch entirely (after using Linux as my daily driver some years ago) due to the lack of UI/UX consistency (I have to use Ubuntu at work; have Mint in my jellyfin pc at home, and even between those there are a lot of annoying differences)

Same here. My three-year old loved maps and we always played with them (making map of her room, etc etc)

We enrolled her at the local Montessori and she rushed to the map section but was told she is forbidden from using it until she takes the lesson on that or whatever is called. That lesson was 2-3 months away, and meanwhile all other kids were able to play with the maps.

This, combined with other rigidities and a crazy schedule totally unsuited for working parents (9-1pm wtf) made it impossible. After struggling a lot for two months, she went back to her old daycare and was very happy there, and is now at her elementary school now

That kinda misses the point for me. I use LLMs a lot and have become an invaluable part of my work (coding, collecting data, proof-reading text) and even personal time (amazing Japanese menu translator!).

But I still think they can "fail" in the sense that current investments need an absurd amount of profits in order to have a positive NPV. Data centers depreciate fast, so if you just did $500bn in capex, you better make sure you will get $50bn/yr out of it from the get go or you will fail.

And that's the problem; even if sales are high, all vendors are bleeding money and so far the offerings are commoditized enough that I can't see how they will recoup their investment.

Quick q what does "yield" mean here?

Also, in which currency are these returns? Because forex shifts (which for sure happened on oct10) mess up the observed returns quite a bit

There's a lot of ways to siphon money out of firms. People declare their car as a "work vehicle", CEOs fly to the beach on the company's jet, etc.

So I guess you want some level of corporate tax? (it's the same for why there are large sales tax in developing countries; sure they are less efficient than income taxes, but lots of people evade those)

> Dominion Voting Systems — the voting machine behemoth that President Trump and his allies baselessly attacked after the 2020 election — has been sold to a Missouri-based company run by a former Republican election official, Axios has learned.

> The intrigue: As part of its deal with Dominion, Liberty officials asked Dominion to settle several other defamation lawsuits against Trump allies who attacked the company over the 2020 election results.

Good luck; we probably crossed paths (I was helping a bit on the hiring side at some point). A postdoc is a really great option, I've seen lots of people do that before going for a bschool so it can definitely can work.

BTW one other thing besides what you mentioned is not just the freeze but the firings. FDIC lost 30% of staff, BOG is going to reduce maybe 10%, CFPB is no more, etc. so the market is actually being flooded with senior economists. They won't compete directly with the posts you want but still flood the market.

Not sure prices are changed though. :/

Three reasons IMO.

1. If you take out AI-related companies, SP500 is down YTD

2. Even then, because the exchange rate is down 10-15% against most countries, that mechanically increases the SP500 by 5-7.5% (because half of SP500 firm profits are from abroad)

3. Many ppl expect lots of inflation in the medium term (look at 10yr-3m spreads). Firms are basically Equity+Debt, and because inflation wipes out the value of Debt, then the value of Equity mechanically increases

So, that's my pet story why the market is at an ATH (source: am an economist, can't say where, but I know lots of well-known economists)