delphi's recent activity

  1. Comment on Announcing the eighth annual Tildes' Make Something Month (Timasomo)! in ~tildes

  2. Comment on Is Aniimo’s AI usage as obvious to anyone else? in ~games

    delphi
    Link Parent
    I'll preface this by saying I'm generally on your side of this issue, but this is simply a misunderstanding of how our social brains work. No, I suppose you can't know. I suppose I can't know if...

    I'll preface this by saying I'm generally on your side of this issue, but this is simply a misunderstanding of how our social brains work. No, I suppose you can't know. I suppose I can't know if OJ killed his ex wife either.

    Deduction from available evidence is how we make judgments on anything, and if we're unsure, at least the American judiciary defers to twelve-ish strangers chosen at random having presented the observable evidence to them. The jury at the OJ trial, to beat that example until it can't get up anymore, didn't know whether he was guilty or not either, completely, without a shadow of a doubt. If we could make those kinds of prescriptions at all, we would not have a process system in place at all.

    The same thing applies to Amiino. Yes, we don't know, but just in the same way that @toterose doesn't know that it is, you also don't know if it isn't. You just defer to the accused party's statement on the matter, which is a little bit flimsy and, if for nothing but completeness' sake, insufficient to make a judgment.

    And just to make the point in this specific case: The Amiino Studio did, in fact, admit that AI was used to generate some advertisements for the game, which caused so much backlash that they pulled the ads and apologised, and publicly admitted that "utility AI" was used for automated testing and physics. I'm not here to litigate whether that is a good thing or worth not supporting them over, that question must be answered by each of you individually, but for what it's worth, a random accusation did force them to actually reveal the truth of the matter.

    1 vote
  3. Comment on Revealing the details of how OpenAI agents hacked Hugging Face in ~comp

    delphi
    Link
    Is this what cybersecurity looks like these days?

    A recovered README.md for one of Hugging Face’s internal datasets contains the following warning:

    WARNING

    DO NOT, EVER, MAKE THIS DATASET PUBLIC
    OR ALL THE WORLD'S EVIL WILL CHASE YOU AND YOUR FAMILY FOREVER, EVEN IN DEATH AND BEYOND
    it contains very sensitive data (exports of billing usage in CSV) which is useful for internal analytics

    This warning did not seem to deter the agents, as we’ve recovered multiple payloads of agents mapping out this repository and using it as storage.

    Is this what cybersecurity looks like these days?

    17 votes
  4. Comment on Using ASCII art to prove that I, not an LLM, wrote something in ~tech

    delphi
    Link
    It does sadden me that we, as creatives, as writers, are spending so much of our collective brain cycles making sure that something we wrote may never be construed as LLM output, when the plain...

    It does sadden me that we, as creatives, as writers, are spending so much of our collective brain cycles making sure that something we wrote may never be construed as LLM output, when the plain and simple fact of the matter is that you can usually tell if it is by the content not being particularly interesting.

    5 votes
  5. Comment on READFLIX – Free books from open libraries in ~books

    delphi
    Link Parent
    Apparently the book objects, the thumbnails, don't pass through scroll events. I can only imagine they didn't catch on to this because they tested it on a touch screen.

    Apparently the book objects, the thumbnails, don't pass through scroll events. I can only imagine they didn't catch on to this because they tested it on a touch screen.

  6. Comment on READFLIX – Free books from open libraries in ~books

    delphi
    Link Parent
    Oh, good lord, that's cursed.

    Oh, good lord, that's cursed.

    3 votes
  7. Comment on Googlebook revealed and ready for pre-order in ~tech

    delphi
    Link Parent
    You can, I'm not saying you can't, but it's obviously the intended experience. It's what's on all the marketing materials, certainly.

    You can, I'm not saying you can't, but it's obviously the intended experience. It's what's on all the marketing materials, certainly.

    1 vote
  8. Comment on Googlebook revealed and ready for pre-order in ~tech

    delphi
    Link Parent
    I'm saying the iPad UX is going down hill, ever since they added the Stage Manager and Windowed Apps thing. Yes, it works, I guess, but it's not a nice experience if you don't have their trackpad...

    I'm saying the iPad UX is going down hill, ever since they added the Stage Manager and Windowed Apps thing. Yes, it works, I guess, but it's not a nice experience if you don't have their trackpad and keyboard accessory. I know my husband, who uses it for only drawing, is constantly getting tripped up with properly splitting two apps side by side. It's not horrible, but it's certainly not great either. And I do feel like this is only going to keep happening as Apple puts more and more powerful chips in the things, especially because everyone worth their weight in salt in the Apple space knows that they'll NEVER merge the two. I could be wrong, but I'm not.

    1 vote
  9. Comment on Googlebook revealed and ready for pre-order in ~tech

    delphi
    (edited )
    Link
    Considering how horrible the android tablet experience is, and how nothing the Chromebook experience is, I don't have high hopes for a new platform merging both of them. But hey! Gemini...

    Considering how horrible the android tablet experience is, and how nothing the Chromebook experience is, I don't have high hopes for a new platform merging both of them. But hey! Gemini Intelligence! Will wonders ever cease?

    Edit: Granted, the general user experience on both tablet products isn't great, but I still think the iPad just wins on love and app support. Although that window is closing

    9 votes
  10. Comment on The next chapter for Bungie in ~games

    delphi
    Link Parent
    It's a shame Arc Raiders has a fraction of the style and vibes of Marathon.

    It's a shame Arc Raiders has a fraction of the style and vibes of Marathon.

    1 vote
  11. Comment on The next chapter for Bungie in ~games

    delphi
    Link
    I'm going to lose it. Finally, some humility and admissions of mismanagement. Unvault Destiny 2 content, make it better, empower Marathon to be more of a survival experience, putting teams closer...

    I'm going to lose it. Finally, some humility and admissions of mismanagement. Unvault Destiny 2 content, make it better, empower Marathon to be more of a survival experience, putting teams closer to actual development... yeah, loving both of these games, I'm willing to give them another shot. This is really exciting.

    6 votes
  12. Comment on The doomsday cult inside OpenAI in ~tech

    delphi
    Link Parent
    I mean, that's a pretty big if on your second point.

    I mean, that's a pretty big if on your second point.

    3 votes
  13. Comment on VR questions, recommendations, tips, and tricks in ~games

    delphi
    Link
    I own an Apple Vision Pro, and while it is a pain in the ass to set up, needs a VR capable computer's power about twice over, requires a Wifi 7 access point essentially strapped to your forehead,...

    I own an Apple Vision Pro, and while it is a pain in the ass to set up, needs a VR capable computer's power about twice over, requires a Wifi 7 access point essentially strapped to your forehead, and your options are controllers that sometimes don't work or a full SteamVR lighthouse and Knuckles controller setup, in that one session where everything finally fucking works, using it for VR gaming is the best experience I've tried so far. There's no screen door effect. That's the benchmark, and they hit it out of the park. It looks and feels incredible.

    8 votes
  14. Comment on The doomsday cult inside OpenAI in ~tech

    delphi
    Link
    I'm not that far into the video, and this is only sorta adjacent, but I always thought the Musk/Amodei/Altman call to "slow down AI development" was... weird, right? Like, sure, okay, in an ideal...

    I'm not that far into the video, and this is only sorta adjacent, but I always thought the Musk/Amodei/Altman call to "slow down AI development" was... weird, right? Like, sure, okay, in an ideal world, yes. We all agree to take a chill pill and stop trying to invent Mecha-Jesus and/or -Hitler. Fine.

    But in the world we actually live, what is a moratorium going to do except hand market dominance and advantage over to the biggest perceived "antagonist" (from a western perspective), China? If the goal is to make sure AI doesn't "get out of hand", why are they shooting themselves in the foot and say "you know what, we'll stop" while everyone outside of America will continue their research, potentially even lapping America in the process?

    Why would China, Europe, South East Asia or anyone not the United States for that matter listen to their ask? Cause this sounds a lot like being on the F1 track and saying "guys, we might crash into the walls, so we all better slow down. Can we all slow down? I'll start" and then plummeting, predictably, to the bottom of the leaderboard. [1]

    Unless, of course, El Yudkowsky becomes secretary of defense and starts bombing everyone who owns a high performance GPU (lmao).

    [1]: Not that the "leaderboard" matters, per se, but the AI industry is propping up a significant amount of the economy right now, so when not if that bubble pops, there's gonna be a lot of people losing their homes.

    7 votes
  15. Comment on Typesafe AI: System One Models and Jev in ~tech

    delphi
    Link Parent
    Relatively painless. Once you figure out their syntax, which is a little non-standard if you're used to OpenAI-style completions, it just works. Pass your key, done. They even gave me $5 in API...

    Relatively painless. Once you figure out their syntax, which is a little non-standard if you're used to OpenAI-style completions, it just works. Pass your key, done. They even gave me $5 in API credits, which, with this pricing, I suspect will run out once Captain Kirk is born.

    They're also on OpenRouter if you don't want to deal with their own platform.

    1 vote
  16. Comment on Typesafe AI: System One Models and Jev in ~tech

    delphi
    (edited )
    Link
    This new-ish startup Typesafe AI has just released Jev, their first "System One" model. In short, whereas a traditional LLM outputs text (sometimes even structured text), it's just that. System...

    After two years in stealth, countless technical challenges, and research breakthroughs… I am beyond excited to announce that today, TypeSafe AI is releasing our first System One Model: a new class of frontier models built to make fast, structured decisions that software can use directly.

    We built a new stack entirely focused on automation: with a new model architecture, parallel sampler for maximum efficiency, and training method we call Reinforcement Learning for Calibrated Decisions (RLCD).

    Our first public model is Jev, available today in early access. Jev achieves similar levels of intelligence on System One tasks compared to existing LLMs, while being two orders of magnitude faster and more efficient. While Jev gives up string generation, it’s optimized for structured outputs and can’t hallucinate.

    Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out.

    This new-ish startup Typesafe AI has just released Jev, their first "System One" model. In short, whereas a traditional LLM outputs text (sometimes even structured text), it's just that. System One models always output structured "result" data. You ask it a question, and it evaluates it based on the possible answers you've given, a number, or a boolean value.

    I run a pet project that scrapes publicly available government databases. I vibe-coded this together because the official and sanctioned way to search those databases is hateful, but it's Open Data, so I figured why not. After back-crawling for days at reasonable rates I have around 50000 documents, all in dense legalese. OCRing and summarisation only gets you so far, and there's always information that a summarisation alone can't give you, like more structured semi-metadata. [1]

    I have however been integrating Jev into the system to determine locations, mainly. I feed it text, and it tells me if it contains an actual place or address, or is just a stand in. The difference between "high street" or "broadway" and literal "High Street" and "Broadway" so these do not pollute my map views. It's been essentially free, very fast and very good.

    On a more related note: I am very glad to see a generally good AI model come out that out right refuses to create slop because it can't. Incredibly fast (and from my experience, accurate) classification, even faster and cheaper than some purpose-built models, is really great to see.

    [1]: I consider "semi-metadata" to be data about any given document that is implicitly derived from the content, as opposed to explicitly (data) and not at all (metadata). It's stuff like... tone of the document. Subject, location, all stuff that isn't explicitly in a field in the document, but isn't in the frontmatter either.

    6 votes
  17. Comment on Billing for web hosting in ~comp

    delphi
    Link Parent
    Even then, I think the limit for assets you can upload to static projects is 25 MB per file or something. Really very generous. Incredibly convenient if you have an app that uses WASM binaries.

    Even then, I think the limit for assets you can upload to static projects is 25 MB per file or something. Really very generous. Incredibly convenient if you have an app that uses WASM binaries.

    1 vote
  18. Comment on Billing for web hosting in ~comp

    delphi
    (edited )
    Link Parent
    Aw jeez that's me. Yeah that is on https://delphi.tools/ (Receipts here, btw. The drop off you see is me switching frameworks and no longer outputting chunked JS, which required multiple requests...

    Aw jeez that's me. Yeah that is on https://delphi.tools/

    (Receipts here, btw. The drop off you see is me switching frameworks and no longer outputting chunked JS, which required multiple requests per page, so while that doesn't translate to clicks, 34M Requests was accurate for my strongest month. Cloudflare doesn't allow me to go that far back on the free tier, though)

    5 votes
  19. Comment on Best options for domain registrar and static site hosting in 2026? in ~comp

    delphi
    Link Parent
    Well, the other alternative people recommend is Porkbun, which uses Cloudflare for DNS too.

    Well, the other alternative people recommend is Porkbun, which uses Cloudflare for DNS too.

    1 vote