I started off with Terra at first before reading anything basically just picking "the middle one" after a while of use I didn't really notice a difference between Terra-Medium and Luna-High, the benchmarks since have suggested there's no real reason to use terra because Luna is twice the speed and some fraction of a cost while on xhigh reasoning achieving better results than Terra medium
I remember trying Luna on high and finding it spent an enormous amount of time compared to Terra, but after your comment I will try it again and see how it performs. If you're correct, everything will change in my usage.
I was talking about changing banks with my spouse, the following two weeks I get podcast ads and Samsung TV ads about the bank I mentioned.
I use all the usual off the shelf "dont track me" stuff like a pihole and a "resit fingerprinting" Firefox fork and a VPN and everything else.
I'm not amazed it still somehow got me, just sad.
I mean you only need at add the stipulation that there was clear negligence or malice in your instructions to the agent.
like we already do with a bunch of other crimes.
yeah for me it's "I wanna add this new thing to an existing system" and the AI responds "we should just add some arbitrary state here to facilitate this feature".
The real issue is the existing system needs to change entirely to facilitate, I know this, Good developers know this, The AI however knows the shitty solution would solve the immediate problem because it's been trained on shitty solutions.
the problem could simply be the AI doesn't have all nebulous loose context I have about the goals of the project and future plans, but I would have to write a novel to give it that context.
> The real issue is the existing system needs to change entirely to facilitate, I know this, Good developers know this
1. I’ll often include boilerplate in a prompt to tell it to make the broader fix. [1]
2. However, a top HN AGENTS.md post 11 days ago included the standard guidance “As much as possible try to minimize the number of changed lines when implementing a feature.” I.e. some devs want LLMs to avoid broader changes and so some of that likely makes it into the training, even if others like us want the opposite.
[1] As far as whether my boilerplate is effective, I don’t know.
The default case when implementing a change should obviously be to make changes with as minimal a blast radius as is reasonable. This is basic software engineering and current LLMs fail it. LLMs also don't have a good idea of what is "reasonable" and a human judgement call is needed.
The case where a (sub)system needs a complete rewrite to admit a feature without incurring too much technical debt should be the exception. When exactly to make that exception is something that clearly currently requires a human judgement call, as models aren't yet nearly smart enough to make such calls.
It truly has been one of the worst UX experiences and poorly made apps that ever made it to mainstream.
I remember a decade or so ago there being a bunch of buzz about them in software dev because they were a microservice node.js/react stack or something with each UI element or portion of back-end having a separate dev team.
I can vividly remember thinking "oh that's probably why when I change song sometimes the album art doesn't update."
they're living proof your product doesn't even need to be that good you just need to be first to market and not really piss anyone off.
I worked for a startup in 2019 which was really pushing the "Spotify model" of team structure. Apparently it's all about self-organising teams and having no leadership. The fact the technology output of the Spotify model - the Spotify app - seemed to be kind of crap didn't seem to matter. I think the main output of the Spotify team structure was blog posts promoting the Spotify team structure.
> Apparently it's all about self-organising teams and having no leadership.
Gross over-simplification. For anyone curious they have videos out there, though they are 10+ years old now so may no longer be accurate, but in 2019-era this was still a thing (company I worked for did it too).[1]
This video was created in 2014 and was more a description of what they strived for at that point rather than of the full messy reality. Spotify quickly after that grew larger than those ideas would sustain (without changes) and rapidly became more alike any other tech corpo. Certainly around 2019-2020 it was not like in the video by any means. Some structures or names of concepts stayed way longer than the actual practices as described.
(I was active as a coach in Stockholm around this period and know multiple people, both engineers and coaches, who worked at Spotify including Henrik the author of the video.)
The Spotify model is basically just Amazon’s team-owned KPIs and independence, put into a UI so you can really see how much their app is conways law.
This is why all the sections of Amazon.com are barely coherent. AWS does this too but it’s more naturally split and text works better than stateful rich media.
> so you can really see how much their app is conways law
The headache of using the shared playlist "Jam" feature and trying to add a song without playing it directly. There is like a separate queue system for "Jam" and things work differently.
Right now, if you have video clips in Spotify enabled, your lockscreen controls will display 'Artist' - 'Song' - 'Video Available', instead of 'Album Name'. If you disable video all together, it just leaves that area blank instead of displaying the album title.
This has been a thing / issue since 2025. It's pretty clear the video team probably put that in to bump video interactions, then the whole team got put on something else and now this problem is blackholed. The most bizarre to me is: even Spotify's top managers must use their own product when they helicopter to home, right? So they would see this too.
Where is a Jobsian anger for user experience when it is needed. Although in that respect even Apple is in the gutter these days.
Totally agree. They also have some stickiness going for them (at least for me). Now that I've organized my music in Spotify and have my habits in place that deal with the crummy UX, I've gotten numb to the poor UX.
Case in point, I make a daily playlist for running and that UX is frequently changing and often for the worse, but I deal with it because it's a habit that costs me a few seconds maybe a minute per day. The cost of switching to a new thing just feels daunting so I stay.
I also am guilty of enjoying the wrapped summaries at the end of the year :)
I definitely enjoy Wrapped. But also something that keeps me in Spotify is the social aspect. If I want to make a playlist for someone, 95% of people I know use Spotify, so I need to have Spotify.
Yeah this is a good point. They also have a feature that lets you build and mix playlists together with others, not my personal speed but adds to the stickiness.
I also enjoyed wrapped, but at some point they made it only viewable in the app. It was previously available in the browser. That was the last time I thought about it.
Good point. Regrettably, newer phones have eliminated external SD cards. I preferred upgrading my phone storage on demand and having external card for stuff like MP3s and videos and manually doing large data transfers directly on my computer.
I carry my entire music library on my phone in Opus format. Smaller than a similar quality MP3 and good enough for the times I’m likely listening offline (working out, riding transit, on a plane).
Yeah last I knew the Spotify desktop app loads many dependencies in triplicate or more because each pane is its own independent iframe universe, all so the engineering teams don’t need to talk to each other and they can A/B test incessantly. Inefficient madness.
you're still mistaking that Claude "knows" anything, it doesn't know or think, it's a word prediction algorithm and there is nothing stopping a word prediction algorithm from predicting falsehoods.
No, it is really useful to know how a technology works. LLMs work by predicting next tokens.
It is _amazing_ the utility they have given that that is what they are and they are highly useful but suggesting solutions that ignore they are spicy auto-complete is counterproductive on many different levels.
I think it is useful to remember, because enough people think these things have genuine motives desires and treat them in that way because of that misunderstanding.
they think theres a person in there with morals that would or wouldn't lie because of some devious reason and forget simply the context filled up and the truth was "forgotten".
The default framing often over personifies ai, but this framing over alienates the model. It’s good to think with both framings, but both feel like imperfect metaphors.
I'd believe it, my spotify breaks all the time now and I seem to remember them announcing none of their devs code anymore.
not 10 seconds ago I had to force close it to get it to stop playing music on my phone.
I tried "be concise and respond in a blunt tone" and all that did was add a section to the bottom of every response "here's the blunt concise truth <1 million bullet points>"
I do this in Unity because allocations/GC cause hitches.
it's pretty normal thing to do there's even the built in pooling libraries so you can pre-allocate 10,000 gameobjects when the game starts.
I haven't played with ECS/DOTS yet but I assume it does something similar.
it's much much more expensive, people like to repeat the line that solar is the cheapest form of energy while hand-waving away batteries and grid upgrades required.
I did the napkin math a while back but a battery bank for 10's of thousands of homes for 48 hours costed as much (more?) than just building a nuclear power plant.
The Australian scientific body, the CSIRO, did an extensive report on the full life costs of various energy strategies and concluded that nuclear made little to no economic sense in Australia.
Perhaps had they used a napkin they would have agreed with you.
The CSIRO aren't the group who have to cost, pay for, take on the financial risk of or be operationally involved in running a nuclear power plant. Their opinions are welcome but not in any way final. The Chinese might easily be willing to come in and build something as their industry matures, for example, at costs not currently anticipated.
The more pressing issue is the Australian government is a coven of luddites who have taken a bold and consistent stance against any form of advanced industrial progress taking root in Australia whatever and have identified nuclear power as high technology that therefore must be categorically banned. Leaving the whole question an academic one and leaving the CSIRO the only group taking an interest.
The CSIRO renewables plan factors in only 2-4 hours of battery capacity (way less, with some already built) with the rest of the backup coming from existing pumped hydro and an existing Natural Gas plant.
one of the reasons nuclear looked worse in the report is that they would have to actually build a nuclear power plant.
it might not make direct sense to build nuclear given what already physically exists, but apples to apples it's not quite as clear.
I think not everything that makes (purely) financially sense does make economically sense. There's much more to it that just money (e.g. where do you get the uran from for the next 30 years?).
reply