yeah, i kind of stretched the meaning of "real" there, i guess. i just meant software that is designed to accommodate users who are not me, and whom i don't even know.
kokoro is surprisingly great at nuance but it's tough to improve that last ~2% or so. kokoro + rvc is really great too; i use that for ELEMENT47, the LLM-centric comedy podcast i do that i wish more people would listen to. (e47.net , feel free to subscribe!)
yeah, i disagree. i think i am essentially ~non-technical and it is not difficult to build one-user applications at all. it's not even difficult to build "real" software!
ARE we still in the dorks-only phase? i am certainly a gigantic loser, but i am not sure i am the sort of dork being invoked and i have been doing and thinking about the exact same kinda stuff. i think the genie is out of the bottle or whatever
it's true. persuasion is the immoral core of marketing. personally i believe everyone should hate persuasion too but nobody will listen to me when i try to compel them!
fwiw, i built something simple like this into my harness thing (github.com/0gsd/enough). may not be complicated enough to do per application nowadays vs. needing a modularized outside solution, but it is certainly a good idea that seems to work!
I think I probably did 12 whole books before I realized no one else would read them, or that if they did, I would feel guilty. some of them are really good though even knowing they are purely plan-based expression!
i believe you meant something like US$9999 for the 512GB. otherwise, i'm going to feel like QUITE the fool for choosing the 96GB variant at the same price
yeah, on a 96GB Mac Studio and Gemma+Qwen, it's definitely fully doable. fully doable but not really for coding on 16GB. but svelter models and cheaper (eventually) hardware are coming!
it's funny because i made this thing (called enough) that aims to make it easy for non-technical people to get up and running with local models quickly, but it is impossible to figure out how to break through the noise. every thread and comment like this breaks my heart a lil bit
but there is not going to be an inherited and largely non-discardable codebase at every company, right?
and maybe not a team that looks anything like the teams that built and maintain the large codebases that are out there.
the distinction in this article makes all the sense in the world to me, and definitely helps as i try to figure out what term i use to describe my current status as a thing-producer, but part of why i just call myself nothing is it is entirely unclear to me what the new configurations of infra + product vision > actual v0.1.0 launch > new feature development "teamlines" are going to end up looking like.
if i had to guess, one such config might be "the 0.1% of vibe coders who took the HN crickets in response to their projects to figure out how to learn how to do what a product engineering team needs to do end to end to make a self-sustainable product."
(self being that one person, not the product itself)
skills can include python files that do whatever in subfolders. it's actually pretty crazy how much they can do for how blindly a lot of people import them. (i built basically this exact same thing a few months ago)
I just can't believe we dealt with 20+ years of social media and I've never seen something like this.
Sounds like Letterboxd has a Top 4, and MySpace had Top 8 friends, but I feel like an inventory of all-timers like this would express my true self eerily well.
I figure this is a good thing to post on your bearblog or equivalent. Or your social media!
i think i just mean if they are using google models, then i believe every query is going to google no matter what apple claims about "keeping it local." whether google does anything with it, separate question i guess, but i imagine it will ALSO be slow to simulate the protection apple is selling. and sure, it's a catty comment, i'll own that. but that is my read on the announcements and demonstrations.
all the limitations of on-device with none of the benefits, it seems? they gotta get SOMETHING out there and soon but idk, i would probably feel safer running a chinese model through a 3P iOS app shell vs. trusting Geminiri to not snitch if i cared about the sanctity of my personal information.
i definitely considered something like this for the local-first harness i made ... i just don't think most people have the RAM to be able to run two good models yet. maybe i'm wrong though. but i also think a single "agent" can compartmentalize itself into subdivisions better than we imagine (i.e., much much better than any single human can). i ended up creating a broker, though, so at least the tool calls don't eat up as much context. and the auto-reset thing is definitely legit.
i mean, i still read hacker news primarily via RSS in feedly. i kind of never stopped using it, and everybody is much more generous with their feeds nowadays than back in google reader times. bearblog, etc. RSS rules
thanks! it's kind of more like an old radio show. i pre-troll myself in many directions.
and btw, not random pages: each episode is centered on 5 Verge stories, 5 USA Today stories, and 5 heavily edited Wikipedia pages from the past week (or past few days when it was 2/week). so it's kinda like John Oliver, but it has an absurd meta-story about the four writer-performers too.
in episode 5 iirc mark zuckerberg hacks into hacker news
i mean, obviously i don't think it's terrible because i made it and keep producing them. it can be difficult to listen to if you don't like kokoro + rvc voices, but it really is surprising that it isn't terrible. i learn stuff almost every episode.
i produce the scripts with a skill and render the audio with src.xmvp.org which i also made (it's a bunch of frankly confusing CLIs, but you can do a lot of fun stuff with it if you try)
anyway, sorry if you think it's terrible. Firesign Theatre was a big inspiration and that can be a little grating at times too. measured wartime satire isn't always comfortable, or something