benmanns

Thank goodness for globalized economies.

On Polymarket you can be the bookie and put up a yes at $0.72 and no at $0.74 if you’re confident in that 73% estimate.

Ask and ye shall receive: https://manifold.markets/benmanns/aliensgov-prop-bets (play money). I set this up so multiple can resolve and other users can add additional speculations.

I thought Plaid have (had?) a developer account that could connect something like 100 accounts that was free.

Same. The complexity of HA also leaves my family with a bus factor of one re: keeping the lights on.

They were only touched on (and just barely) in my CS education, so don’t feel too left out. Spend an evening or two on the Wiki for Probabilistic data structures[0]. With a CS education you should have the baseline knowledge to find them really fascinating. Enjoy!

Oh, and I don’t find myself actually implementing any of these very often or knowing that they are in use. I occasionally use things like APPROX_COUNT_DISTINCT in Snowflake[1], which is a HyperLogLog (linked in the Wiki).

[0]: https://en.wikipedia.org/wiki/Category:Probabilistic_data_st...

[1]: https://docs.snowflake.com/en/sql-reference/functions/approx...

I've really been enjoying Complex Systems with Patrick McKenzie (patio11) of HN fame. Additionally: Odd Lots, Money Stuff, Chat with Traders (hit or miss, some guests are not great).

Audiobook publishers require/request this when you sell subsidiary rights. We’ve been able to push back citing accessibility concerns. I find it really annoying when not available for my own reading.

For Shopify in particular, they have a headless front-end services called Hydrogen[0] (with optional hosting called Oxygen). It's basically an opinionated wrapper around the storefront[1] and customer account[2] APIs, which allows interacting with the store from the front-end. It allows you to host Shopify on your own domain under your own control and gives a bit more customization than is available via hosted Shopify. It's what I run for the Creature Publishing site[3] and allows some extra customization for wholesale accounts, etc, without the exceptionally expensive enterprise (Shopify Plus) plan. To be completely honest, sometimes I question the decision over a simple hosted shop subdomain. Some light SSR/API calls are necessary for our setup, which is hosted on Cloudflare Pages/Workers.

[0]: https://hydrogen.shopify.dev/

[1]: https://shopify.dev/docs/api/storefront/latest

[2]: https://shopify.dev/docs/api/customer/latest

[3]: https://creaturehorror.com/

It does, however, make providing housing more profitable, which, on the margins, will drive more landlords and home builders into the market, decreasing long term costs (relative to a straight 100% increase relative to the basic income). So you might send everyone $100 per month and costs go up $100 per month, until supply chains shift towards supplying lower income humans with more goods and services than they used to get, at which point costs will decrease (from the $100 increase).

With enough forewarning, suppliers could anticipate the increased demand and prepare for it.

The trouble is, you have to place the outbound call to those contacts to trust them. People could spoof an incoming call from numbers in your contacts and it will look as legitimate to you as a receiver as if the real number was calling you. With voice spoofing, it's now possible to call someone as [grandchild] with [grandchild]'s voice with a pretty horrible story about what's going to happen if some Bitcoin or Google Play gift cards are not purchased and handed over immediately.

IMO the game is to get the highest credit score while spending the least on credit. That way if you e.g. need to acquire housing, a reliable vehicle, business loans, even employment or insurance, you will be able to and can get a good deal. Sure, paying nothing saves you money, but it also costs you the opportunity cost between what you could earn/save by utilizing credit vs waiting months to years to save up and pay cash.

I’m not sure. I’m sophisticated enough to use BNPL and set aside the funds in a HYSA, but I never have as you miss out on: credit card rewards, credit card purchase protection, credit card extended warranties, credit card chargeback infrastructure. Additionally, I’ve seen BNPL offers where the first payment is in 2 weeks compared to 30-50 days for credit cards depending on when your statement closes and how soon you have to pay, so the extra interest is less than you might think. It could make sense for very large purchases, but then, that’s also where credit card features can really come in handy.

Like you, I’m far too risk averse to not fact check everything an LLM outputs, but I’ve also fixed bugs that have been present for 5+ years. Maybe at a certain point you can just wait for the next generation of model to fix the bugs. And wait for the generation after that to fix the newly introduced and/or more subtle bugs.

This is interesting to me—are you an app developer that does this pattern for that reason? I definitely figured it was exclusively because they can request repeatedly for the fake prompt, and only “use” their prompt when the user is inclined to accept. There are times where I’ll accept the first prompt and reject the second prompt when apps are spammy about it.

Another case of Hyrum's Law, where the entire functionality of the S3 SDK and any competing service provider borrowing from it becomes Amazon's problem to fix at their own cost. Maybe it's time for a non-Amazon but S3 API compatible library to emerge among the other cloud storage providers offering S3 compatible APIs. OpenDAL looks interesting. Also another reminder to run thorough integration tests before updating your dependencies.

You probably know more about this than me, but it looks like some lines do: https://tfl.gov.uk/corporate/transparency/freedom-of-informa...

Regenerative braking in cars also keeps a lot of brake dust out of the air. A pair of brake pads lasts about as long as the life of an EV.

Also keep in mind there’s something like 180 days per year of K-12 education in the US each year, so that’s something like 0.62 to 5.4 per school day.

I think so, but not a bad marketing gimmick. It gives a pretty easy way for the general public to interact with ChatGPT on a trial basis without signing up or paying for it using a somewhat hard to acquire identifier (phone number). I'm curious if they're doing anything to avoid abuse from spoofed numbers.

I think we should start doing more taxes combined with subsidies. Give everyone a $1/t carbon tax. Give everyone a ~$1/t farming subsidy based on current carbon production. Nobody loses, but everyone is incentivized to decrease carbon production and the faster ones profit more. Phase out the subsidy over X years if you like.

Otherwise, you’re right. We’re upsetting the balance of a very complex, very important system and causing a regressive tax in the form of price increases.

Email rendering is a great analogy.

Most eBooks are produced by creating a new InDesign ebook document from the existing InDesign print document. Then you fiddle with a bunch of stuff to make it look right (removing the forced line breaks and such you did to get the print document looking how you want it, etc). That's then exported to ePub, which is a zip of some HTML, CSS, images, and fonts. The code outputted is absolutely terrible, as one used to get creating webpages with WYSIWYG Dreamweaver et al. and causes a lot of issues that have to be fiddled with in InDesign, manually corrected in the unzipped export, or frequently just left in the final book.

There really should be a better eBook specific authoring system. Ideally with importing from the print document, as that's the canonical version. It's just a website, after all, and we have great tools for building those.

1.b. When somehow they identify you’ve change IPs the first time, and now that you are home, there’s still some sticky session that doesn’t get cleared without purging all cookies for the domain.

I think there are legitimate cases where a business has a product available at a price that is net positive to both the business (profitable) and customer (consumer surplus) but is having trouble connecting with consumers. What % of advertising this is versus pushing low quality trash, branding, status symbols, and scams, I have no idea.

However, as discovered during a family heart crisis, medical professionals will routinely ignore any kind of heart rate or blood pressure readings that you take at home. In my experience, it's not until they see the same measurement during the ($$$) ambulance ride that they take it seriously.

Am I understanding correctly? You look an English dataset, trained an LLM, machine translated the English dataset to e.g. Spanish, continued training the model, and performance for queries in English improved? That’s really interesting.

On the flip side, I've used plenty of first party SDKs with weird timeout values, poor error handling, and non-idiomatic APIs for the language.

If anyone is interested, there are markets on Manifold trying to predict this: https://manifold.markets/ai

To be honest, even as a participant and enthusiast, I’m not sure how useful or accurate it is. I can push the prediction by 13 years with $1 worth of Manifold currency.

Starting January 1, 2025, existing companies will need to report beneficial owners to FinCEN, which may start the process of cutting down on shell company abuse. https://www.fincen.gov/boi-faqs#B_2

Good news, https://www.litmus.com/ is exactly this, and does work for Gmail.