> difference between an instruct based re-ranker and laya/jev I just don't see it
Main difference is that laya/jev/et-al give you a zero-shot classifier that requires no training. You can prompt engineer your way to a quick fairly reliable cheap enough decision engine that you can use to iterate quickly (by prompt engineering).
Right now a lot of people are doing this with LLMs and it's too slow and expensive.
Imo the right iterative approach to productionizing these systems is something like:
1. Build it with an LLM. Iterate on the prompt
2. Start building a real-world dataset
3. When the prompt works, turn it into a clear rubric for Jev or similar
4. Keep iterating until desired accuracy achieved
5. Use the real-world evals you've built to train a custom classifier fine-tuned to your needs
You now have a system that has produced useful results in production from the very beginning and by the end it's a reliable super cheap classifier that can make thousands of decisions per second.
I don’t think that’s it. I sincerely doubt most developers are doing side by side comparisons of calibration quality.
OpenAI has a section on their embeddings model api page for zero shot classification. Of course you can choose an open weights embedding too if you’d like.
I think Jev wins on marketing and convenience. Most SWEs don’t want to talk about embeddings, cosine similarity, or precision/recall tradeoffs. They want something which plausibly works and is easy to use.
Yes it turns all that work of building a classifier into an api call. This is hugely valuable for prototyping and while you iterate on what the product should even do.
> IMO high housing prices are destroying civilization.
It's like a game of musical chairs. There aren't enough houses to go around. You go around and the music stops. Everyone scrambles to their nearest chair. Obviously the person on crutches will be the first to fall out of the game. The rest then talk about laziness, drugs, bootstraps, and all the rest. But add 1 more chair and there's no problem, even the dude with the broken leg can sit.
Someone tried to build the extra chair but the endless required studies and legal challenges from existing chair-sitters bankrupted the would-be chair builder.
Also the chair, if actually built, would be so expensive that one legged guy can’t afford it. His best hope is someone with a worse chair takes it, then he tries to take their seat, but there’s already a dozen two-legged guys in front of him.
Everyone wants to blame existing landowners and sure they share part of it but the lion's share of the blame ought to be directed at do-gooders (land owning or otherwise) who are unwilling to be honest (with themselves, or in the internet comments) and admit that every policy that "looks good" based on the one sided statements presented in abstract by its peddlers.
Each one of those policies is part of the system that gets us the results we have and the peddlers all share in the blame.
> Also the chair, if actually built, would be so expensive that one legged guy can’t afford it. His best hope is someone with a worse chair takes it, then he tries to take their seat, but there’s already a dozen two-legged guys in front of him.
Maybe you need 5 chairs? Here in SF supply has fallen decades behind the demand curve to the point it may be impossible to catch up even if all you did was build build build.
decades of reasonable, consistently applied housing policy led to the creation of islands of wealth, surrounded by oceans of poverty. ergo, this is the purpose. this is how the people there want to live. I left many years ago when I realized the futility of it all.
I call SF and Silicon Valley “the world’s only slum for the rich.”
What buys you a trap house or an apartment with hallways that smell like piss in SF or the Valley will buy you a mansion in 90% of the country at least.
> It's like a game of musical chairs. There aren't enough houses to go around.
Actually, there are. In Germany for example, we have millions of units left empty... the catch is, these are in (extremely) rural areas. Areas that used to be populated, but no longer are because the government has let them fall extremely behind in modern amenities. Some even do not have cellphone service.
Normally, you'd expect people and companies to do a pricing arbitrage and move where housing is cheap, but unfortunately, when there is no local employment or the infrastructure does not support remote employment, or other crucial requirements (safe potable water, electrical grid, grocery stores, schools, some sort of entertainment) are lacking, this does not work out any more.
Well, for generations now, exponential growth in all things has been state policy and it was good (tm). Because of population growth, there's always a shortage of housing. Population is projected to peak this century. Whatever will we do when there are already enough houses for everyone to have one?
> Whatever will we do when there are already enough houses for everyone to have one?
In many places in Europe and Asia, the govt will essentially pay you to go live in near-abandoned villages and revitalize them.
In USA it’s just ghost towns. See also Detroit
The problem is better described as "There aren't enough homes where people want to live". We have plenty of houses, even in america. The "where people want to live" part largely comes down to density, zoning laws, and the fact that for most people their primary residence represents their entire net worth and it would be materially impactful if the price went down (by building more/better homes nearby). So everyone resists building.
But it's okay, if you resist building for long enough then your town will eventually collapse and your house will be worth nothing anyway. Many american Towns are essentially insolvent and can't afford realistic upkeep prices for services like sewage, roads, good schools, etc., and are living on borrowed time until infrastructure collapses and nobody wants to buy there. See Strong Towns the book.
> The "where people want to live" part largely comes down to density, zoning laws, and the fact that for most people their primary residence represents their entire net worth and it would be materially impactful if the price went down (by building more/better homes nearby).
Infrastructure is a major deal. You can't really exist as a member of modern society without broadband Internet, unless you're Amish or whatever and even they have begun to shift attitudes. Starlink changed the calculation quite a bit, and it definitely helps, but other things such as cellphone reception and emergency services that arrive in a reasonable time are also important.
Surely at least part of it is the concentration of wealth and economic activity. There's large masses of decaying rural towns where houses are cheap, but at the same time the whole reason they're decaying is because there just isn't enough economic opportunity to afford even these cheap houses.
I have no clue what the overall balance of housing looks like, but surely it doesn't help that big chunks of the housing market are being deprecated at the same time.
See I’ve had this insane wild idea for years that homelessness could be related to inadequate housing supply.
Wild I know. People not having something being related to not enough of that thing.
But we have to stop those developers! If there’s ever a food shortage will people be saying “stop greedy farmers?”
Actually they probably will. Not enough food so obviously farmers are causing it and it’s a conspiracy. I’m sure that’s what the TikTok slop and Xhitter rage bait would say.
For anyone who doesn’t believe this is real, I recommend reading Thinking In Pictures. It’s a great book.
Explains 3 types of thinking that people use in their day-to-day: Linear verbal, schematic conceptual, and photo realistic.
It’s all a spectrum so you likely use all 3 to an extent.
Most people think linearly and sort of describe things in their mind that they’re trying to imagine. It’s a picture but not really.
Engineering types most commonly are strong in conceptual thinking. You can imagine complex systems and how they connect but it’s hard to describe and linearize this into words. You have to translate. Again it’s sort of an imagined picture but not really.
Then there is photo realistic where you see the thing in your mind. Almost as if you were seeing it for realzies with your eyes. Common with visual artists and some types of autism. Linearizing into words is almost impossible.
The book talks about the extreme opposite of aphantasia but it builds a good way of understanding these different qualia imo
According to Ed Catmull surveying Pixar staff, there doesn't seem to be much more than a minor correlation between ability to visualise and artistic ability.
I can personally watch a movie in my head, eyes open and I’m super not artistic but very engineer oriented. To the point that I realized I should be dropping software for mech eng
I'm the same. As a student, I could vividly picture the page of the book containing the details of whatever problem I was working on. I have close to zero artistic and design ability, but am very strong in engineering and spatial reasoning.
I've wondered about this: if you can hold an image clearly in your mind - e.g. the Eifel tower - wouldn't you be able to reproduce it well on paper? I know that's not artistic ability in the creative sense, but still..
People often can't reproduce what is right in front of their open eyes well.
Conversely, I have aphantasia but can draw reasonably well from memory (at one point well above average, but I'm about 30 years out of practice).
There are people who can "trace" what they imagine, but it doesn't appear that being skilled at drawing is that closely linked to ability to visualise.
An interesting thing, though, is that if I draw from memory my lines are more precise, while if I draw from something I can see, I will sketch out and refine. I think the two are quite different skill sets - drawing what you see often requires paying attention to spatial relationships and transferring those to the drawing to get proportions right etc.
> Engineering types most commonly are strong in conceptual thinking. You can imagine complex systems and how they connect but it’s hard to describe and linearize this into words. You have to translate. Again it’s sort of an imagined picture but not really.
I guess I have that ability, in that I can relatively easily figure out spatial relations or can map out code mentally. The photorealism I am a 0 on.
Is linear thinking just a stream of conciousness thing where you basically list attributes? If so then I guess I have that too.
Schematic conceptual, and several other modes not listed. I only do verbal with concentration, and never photo.
NOT a picture, I'm aphantasiac. Map, maybe, sort of works. But mapping (mathematics concept) works in some ways superior to the idea of "2d map". GIS associations for instance are intuitive and obvious to me, but so are dictionaries.
> Were magazines insanely more expensive at the time? What gives?
The standard of living was much lower and people’s expectations of reasonable income were lower too. It was common in those days for blue collar workers to do 12 hours of hard physical labor and live in cramped conditions with multiple people per room. What we would now consider worse than a homeless shelter.
In comparison just working a desk job was already a luxury. Even if you were barely making ends meet
Quick Google search says a newspaper writer would make $400 to $800 per year in Poe’s time. That is about $20k/year in today’s money, which is just above poverty level
I was reading about 19th century Russian writers recently. Very few of them made much money and some like Turgenev and Tolstoy were personally wealthy.
That’s the best part! We’re all NPCs in other people’s stories _and_ the main character in ours.
You can do and be absolutely anything or anyone you want. Just gotta bear the consequences and that’s it.
Wanna live a homely life surrounded by loved ones? Go do it! Nobody is stopping you. Wanna galivant around the world having adventures with no roots? You can!
And when you meet a local npc bartender on your adventures, you’re just another rando tourist npc. It’s great
Commented elsewhere in the thread before reading the comment (a terrible habit of mine) and I fully agree. The way I explained it to my daughter is this:
Every snowflake is unique but snow is still just snow.
Life’s funny that way. You can’t predict what an individual will do and yet predicting what a crowd will do is so easy we have entire engineering and social disciplines based on manipulating crowd behavior.
> The frequent rain and mud makes it to the rim brake shoes and diminish braking power
I'll never forget when I was a kid (13 or so) going for a ride in the city during a downpour, slamming on the brakes because a pedestrian stepped out in front of me (mixed pedestrian/bike zone, no cars, we both assumed no-one else was dumb enough to be out), and ... nothing happened. Damn near shat my pants. Luckily the pedestrian saw me and jumped out of the way.
In retrospect that was too much downpour. Could literally see water spraying out my brakes. The hot cocoa after was fantastic tho
> The most surprising thing my adult ADHD diagnosis revealed was that as I began telling friends that I had been diagnosed positive they told me they had known that since they met me. These were childhood friends.
As someone with peer reviewed ADHD but not official diagnosis, what are friends gonna do about it? They can tell you, you're always late and it's annoying but that's about it. At some point it becomes just who you are as a person, part of your character sheet. If you're not planning to change and it's mostly under control (have job, life, etc), does it even matter to put a label on the thing?
I feel like a diagnosis in my 20's would've been useful. Now it's like I got me my coping strategies and they work pretty well. If I do or don't have ADHD doesn't really change anything.
And for anyone else reading, the leading symptom of adult-diagnosed adhd is saying and thinking things like "I don't get it, ADD can't be real, that's just how everyone is"
> I feel like a diagnosis in my 20's would've been useful. Now it's like I got me my coping strategies and they work pretty well. If I do or don't have ADHD doesn't really change anything.
This is what I told myself before I started taking medication around age 30. But ultimately my coping strategies felt flimsy, and I felt like my life could fall apart at any second.
Getting medicated and getting therapy has improved my life and evened out my emotional state an incredible amount. I still feel like me, just with a bit more stability and consistency.
I don’t think it’s ever too late to start medication. If you think there’s a benefit to it, it’s worth trying out.
I didn't know - but I was adaptively self-medicating.
At one time it was Jolt Cola ("all the sugar and twice the caffeine!"), then energy drinks came along. I can drink 10 a day, because I genuinely function better that way.
But Adderall works really well now, and unlike caffeine, it doesn't set me up to lose sleep.
different people with ADHD have different symptoms or effects. I think yes it absolutely matters because I have always secretly been hard on myself, that I am worse than others, struggled under extreme rejection sensitivity. It matters and helps, if only to be able to be a little kinder to yourself and have the ability to decide to get help in the form of a therapist, or medication, or find strategies that worked for others.
I'm ADHD and really struggle with similar things. I've been leading a really lonely life because I tend to isolate after any friction with people, assuming they don't like me anymore.
I know to people without this problem it probably sounds very childish and silly, but it's a real problem. I lack a lot of the inner social calibration of what is a big problem and what is not so I very often have social issues because I say something hurtful (unintentionally) and don't realize why it's a big problem. Then I say other stuff that other people are actually fine (or mostly fine) with but I stress out about it and agonize and isolate myself.
I also identify with being super hard on myself. For a recent example I replaced a door in my house recently. I had never done this before. I bought a power routing saw and cut the hinge mortises myself. Two of them were basically perfect, one of them was a bit jagged and it was off by 2 millimeters when I went to hang the door. I spent the next four hours pissed off at myself for fucking up the door!
After doing a franky great job for my first ever attempt at hanging a door, and never using a routing saw before. :/
> This is like asking if anyone in the Tour de France has credited training wheels for their success.
And yet every F1 driver credits gokarting for their career. Laying the groundwork in your early years makes a big difference. My career would be very different without Logo and Turbo Pascal
I hate that the overpriced polishing cloth was legit the best screen cleaning thing I’ve ever used. One wipe and my ipad was perfect.
These days I use the cheap microfibers I got for the motorcycle and yeah it works but scrubbing my device like an idiot for several seconds just doesn’t feel quite as magical you know
> For the first part, I suspect that money, and the safety net it provides, is a bigger impediment to entrepreneurship than its collegiate advocacy.
An old HN comment captured this perfectly, it’s my favorite analogy of entrepreneurship [1]
Entrepreneurship is like one of those carnival games where you throw darts or something.
Middle class kids can afford one throw. Most miss. A few hit the target and get a small prize. A very few hit the center bullseye and get a bigger prize. Rags to riches! The American Dream lives on.
Rich kids can afford many throws. If they want to, they can try over and over and over again until they hit something and feel good about themselves. Some keep going until they hit the center bullseye, then they give speeches or write blog posts about "meritocracy" and the salutary effects of hard work.
Poor kids aren't visiting the carnival. They're the ones working it.
Main difference is that laya/jev/et-al give you a zero-shot classifier that requires no training. You can prompt engineer your way to a quick fairly reliable cheap enough decision engine that you can use to iterate quickly (by prompt engineering).
Right now a lot of people are doing this with LLMs and it's too slow and expensive.
Imo the right iterative approach to productionizing these systems is something like:
You now have a system that has produced useful results in production from the very beginning and by the end it's a reliable super cheap classifier that can make thousands of decisions per second.reply