I'm a big fan of Mochi[1] (also unaffiliated) after getting frustrated with the clunkiness of Anki.
Mochi has great native apps on macOS and iOS (and maybe more?), the cards are formatted in markdown so I can generate them with LLMs with a custom system prompt, and I just found out today they have an API so I might try my hand at getting an LLM to push new cards on its own via. an MCP server.
I solve this problem with a system prompt in my LLMs[1].
ChatGPT is the main spot where I'm going to be trying to understand a new concept, so after groking it I'll ask it to make flashcards which I can then just copy and paste into Mochi.
An improvement would be some sort of MCP integration between the LLM and Mochi so it can just add the card directly. I'm sure we'll get there soon.
I've had the same sort of difficulty with phrases like "most" or "almost all" or "hardly any"—I crave for these to map to unambiguous numbers like the probability yardstick referenced in this article.
I spun up a quick survey[1] that I sent out to friends and family to try to get some numbers on these sorts of phrases. Results so far are inconclusive.
"Almost all" is an interesting one, because it has family of mathematical definitions in addition to any informal definitions. If X is a set, "almost all elements of X" means "all elements of X except those in a negligible subset of X", where "negligible" depends on context but is well-defined.
If there's a finite subset of an infinite set, almost all members of the infinite set are not in the finite set. E.g. Almost all integers are not 5: the set of integers equal to five is finite and the set of integers not equal to five is countably infinite.
Likewise for two infinite sets of different size: Almost all real numbers are not integers.
> "Almost all" in math can mean "except at every integer or fraction" :)
I am a mathematician, but, even so, I think that this is one of those instances where we have to admit that we have mangled everyday terminology when appropriating it, and so non-measure theoretic users should just ignore our definition. (Similarly with "group," where, at the risk of sounding tongue-in-cheek because it's so obvious, if I were trying to analyze how people usually understand its everyday meaning I wouldn't include the requirement that every element have an inverse.)
Sure but that’s because 100% of real numbers, by any standard measure, aren’t integers or fractions. It bothers me if it’s used to mean 95% of something though.
My wife and I have a daughter in the demographic of these shows, though she's a little young for Bluey. There's a YouTube (and now Netflix) show called Ms. Rachel for a younger audience that I'd put in the same positive category as Bluey.
We probably watch one or two hours of Ms. Rachel videos a day with our daughter. We've got several family friends with a household rule of "no screens at all for kids" who would scoff at that but their rule seems both draconian and technophobic to me. Our daughter has picked up many words and concepts from the show and we've learned a lot of the songs as a family and sing them when the context comes up (ex: "baby put your pants on..."). Ms. Rachel has been a hugely positive parenting tool for us.
Every once in a while, though, YouTube will try to autoplay some Cocomelon after a Ms. Rachel video and wow it's just absolute garbage. I think this article captures it well: it feels like slop engineered to keep young eyeballs glued to the screen with no higher purpose than increasing the number of engaged minutes.
Instead of "no screens," the more granular "you can choose from this menu of approved content on your screen for a reasonable amount of time per day" is the better parenting move for our family.
Our son has profound hearing loss and he wears Cochlear implants, and I remember very fondly the time we were hooked on Ms Rachel.
She is great and lots of her videos are a blessing for parents with children with hearing impairment as she uses lots of techniques that our Speech and Language therapist used to teach us.
And signing! I've noticed she signs along with most of what she says. I think it's an inclusivity thing, but it also meant that our daughter could communicate simple ideas long before she could talk.
I don’t have kids yet, but 100% agree with your last paragraph. Controlled, carefully selected content is much much better than no screens approach. When picked carefully, those shows are actually educational, helping with growth.
IMO it's all about privacy. Perhaps also availability if the main LLM providers start pulling shenanigans but it seems like that's not going to be a huge problem with how many big players are in the space.
I think a great use case for this would be in a company that doesn't want all of their employees sending LLM queries about what they're working on outside the company. Buy one or two of these and give everybody a client to connect to it and hey presto you've got a secure private LLM everybody in the company can use while keeping data private.
I’ll add to this that while I couldn’t care less about open AI seeing my general coding questions, I wouldn’t run actual important data through ChatGPT.
With a local model, I could toss anything in there. Database query outputs, private keys, stuff like that. This’ll probably become more relevant as we give LLM’s broader use over certain systems.
Like right now I still mostly just type or paste stuff into ChatGPT. But what about when I have a little database copilot that needs to read query results, and maybe even run its own subset of queries like schema checks? Or some open source computer-use type thingy needs to click around in all sorts of places I don’t want openAI going, like my .env or my bash profile? That’s the kinda thing I’d only use a local model for
I love this! I was similarly inspired by the Wait But Why article[1] and made a chrome plugin several years ago that shows this for my life every time I open a new Chrome window[2]. It's also a handy countdown for big events like vacations or The Singularity (which starts in 7,259 days) that I want some time to prepare for.
I freakin' love Sublime Text but the AI chat features of Cursor make all of the benefits in this post irrelevant.
It's fast? Not as fast as an LLM.
LSP code completion? Not as good as LLM completions aware of your entire codebase at once.
Snippets? These don't matter if an LLM can just make them up on the fly.
We've entered a new paradigm of what it means to be a good code editor. I'd love if Sublime added the LLM chat and code diff from Cursor but I think the new way to edit code is going to look a lot more like having a conversation (text or voice) with an LLM that's making the changes for you.
Deep understanding and being able to reason about problems is more important than speed. Offloading too much to LLMs has a detrimental effect on our brains.
I'm not saying don't use LLMs. Use them in ways that increase your skills, and be wary of using them in ways that atrophy your skills.
AI LSPs are coming. There are about half a dozen of them right now that you can find on GitHub, but I haven't been able to get any of them working really nicely and Sublime. But I have faith that eventually they'll mature up and a good one will surface as the go-to solution.
You can run DDL commands directly on ClickHouse, but after you have a lot of these commands in your codebase, it becomes a pain to manage them. We had these commands defined in a lot of ruby files named <timestamp>_create_table.rb using which we mimicked the way rails and active record manage the app database migrations. Rails and active record also give you a schema.rb which shows the latest state of the app database at all times. After having no schema.rb for a long time, and it becoming a pain to manage the standalone migration files in ruby, we decided to build this to be able to manage migrations for ClickHouse, our analytics database. We love using it and the schema.sql file (with the latest db state) it generates at the end.
Television in particular seems ripe to be reallocated. Didn't we go through a whole analog-to-digital conversion over a decade ago that led to TV going through wires instead of through the air?
Virtual Channel Numbers let a station pretend to be on a particular channel number. The actual RF channel number doesn't need to match. But the channel number you key in using your TV remote does need to match the virtual channel number.
In nearly all populated areas of the US, you can still receive broadcast TV for free over the air with an antenna.
Digital television stations state what "channel" they are in their signal's meta data. That allows them to change frequencies but keep their channel identity. Since TV when digital, many stations have changed frequencies, some several times. You may find the "repacking" of the broadcast TV frequencies an interesting read:
For reasons I don’t entirely understand, it would cost me quite a lot of money to view my local free-to-air TV stations over either cable or the Internet, so antenna it is (for the very few times I need it).
In the USA, over-the-air stations may require cable operators to carry their channel at no charge (to the operator) or they may negotiate a charge to the operator, which the operator may refuse. The major stations have chosen the latter. Part of a cable TV bill pays for this (though the stations would day they’re just getting their fair share of the high cable bill.)
I don't know if it's true but someone told me this is the main reason why the likes of Home Shopping Network and QVC keep their over the air transmitters going in many places even though most of their viewers are on cable. It seems like a waste of spectrum space but it's so that they can force the cable operators to carry them.
That's generally not true. US cable and satellite operators are only required to carry the "primary" video feed [1], which is usually the xx.1 channel. In most markets, home shopping channels typically air on subchannels (xx.2, etc.). The exception, of course, would be if the TV station designates the home shopping channel as their primary channel.
Home shopping is usually used to monetize excess bandwidth.
I've heard about that, but I wonder what the economics here are.
Are enough people really willing to pay for the convenience of, I guess, not having to switch between antenna and cable input, or are living outside of broadcast coverage of the stations they care about?
Weirdly, it's exactly the opposite in Germany: Supposedly the public broadcasters have to pay the cable companies to get them to carry their programs.
The average American barely knows how to turn their TV on and off. Switching inputs is a scary prospect. Having rabbit ears on your tv is also def a social status signaling thing.
> Having rabbit ears on your tv is also def a social status signaling thing.
That's what I've long suspected. No wonder it's a great opportunity to save/waste money :)
Supposedly in some social classes and age groups, broadcast TV is literally unheard of, with Best Buy promoting TV antennas accordingly ("free cable!") and people suspecting it's a scam or illegal.
Mochi has great native apps on macOS and iOS (and maybe more?), the cards are formatted in markdown so I can generate them with LLMs with a custom system prompt, and I just found out today they have an API so I might try my hand at getting an LLM to push new cards on its own via. an MCP server.
1. https://mochi.cards/