post_below's recent activity

  1. Things that don't suck

    ~talk Ask
    So much of what the algorithms surface is negative. For all of the reasons that mostly everyone's aware of at this point. It's easy to get the general impression that times are dark without...

    So much of what the algorithms surface is negative. For all of the reasons that mostly everyone's aware of at this point.

    It's easy to get the general impression that times are dark without realizing it. I think sometimes it's good to intentionally offset algorithmic (and general human) negativity bias.

    Lets do a positive news thread, I'll start:

    Hungary votes out Orbán after 16 years

    Perovskite solar cells hit 34.85%

    Portugal hits 80.7% renewable electricity

    Hidden drainage system found in human brain

    First lab-grown oesophagus using hosts own cells (fully incorporated with muscles, nerves, arteries within 6 months)

    And of course Artemis II! Why is space exploration somehow more positive than the sum of its parts?

    Please post anything, it doesn't have to be "news". The full range of the humanities works too

    14 votes

  2. Comment on US begins blockade in Strait of Hormuz in ~society

    post_below
    Link Parent
    Do they? I mean eventually, entropy being what it is, everything collapses. But in actual practice the last time a major world power collapsed it wasn't complete and it was 25 years ago. Before...

    All civilizations collapse

    Do they? I mean eventually, entropy being what it is, everything collapses. But in actual practice the last time a major world power collapsed it wasn't complete and it was 25 years ago. Before that what, the Qing dynasty? Except China didn't really collapse. The Spanish empire? Except really they just contracted back to their original civilization but bigger.

    In actual practice I don't think "all civilizations collapse" is very useful. There's a long way to go before actual collapse is even on the table for the US. Not to say that the current admin won't do more dumb things to get there faster... but in the meantime the world has figure out what to do about the reality we're in now.

    Or to put it another way, there's a whole lot of human life between where we are now and a theoretical "everyone's starving to death" future, with a lot of important decisions to be made along the way, any of which could change the outcome.

    5 votes

  3. Comment on Half-baked idea for metered inline image allowances in ~tildes

    post_below
    Link Parent
    There's no denying it's technically less wrong

    There's no denying it's technically less wrong

    9 votes

  4. Comment on Half-baked idea for metered inline image allowances in ~tildes

    post_below
    Link Parent
    This is not ok. Young Tildans seeing this could get a warped idea about anatomy. The only solution now that you've done it is to allow real images to set the record straight about cat tits.

    This is not ok. Young Tildans seeing this could get a warped idea about anatomy.

    The only solution now that you've done it is to allow real images to set the record straight about cat tits.

    20 votes

  5. Comment on Half-baked idea for metered inline image allowances in ~tildes

    post_below
    Link
    I don't expect images on Tildes to happen anytime soon given that they were intentionally not included at the start. And I don't have a strong opinion one way or the other. I guess I lean...

    I don't expect images on Tildes to happen anytime soon given that they were intentionally not included at the start. And I don't have a strong opinion one way or the other. I guess I lean pro-image.

    But if they were to happen I don't think complicated schemes would be the best way to go. Just quick and easy moderation by superusers and the ability for people to turn them off for their account, maybe with some sort of visual indicator or "show" option where an image would otherwise be.

    I don't think images would hurt the quality of conversation, a little more irreverance would probably be a good thing. It would probably be annoying sometimes but we'd all survive. And also I don't think the lack of images takes away anything important from what Tildes wants to be.

    It would be really interesting, though, to see what Tildanians would do with more expression options. I guess in 2026 we'd see a lot of AI generated images.

    3 votes

  6. Comment on Why Microsoft’s war on Windows’ Control Panel is taking so long in ~tech

    post_below
    Link
    It's a strange choice that MS is making. Rapid enshittification only works if you have a high level of lock in. Social media apps rely on the network effect, Google relies on market dominance....

    It's a strange choice that MS is making. Rapid enshittification only works if you have a high level of lock in. Social media apps rely on the network effect, Google relies on market dominance. Microsoft relies on enterprise inertia and OEM deals.

    But I don't think that's going to be enough in the long term. They'll likely hold onto their enterprise dominance for quite a while longer but if they start to bleed consumers the whole empire could come, slowly, crashing down. They've been losing the tech crowd for a long time already.

    Their strategy seems to be to capture the LLM agent market in both enterprise and development but I don't see any indication this is working. At best they're positioning themselves as middleware, which is a tenuous place to be for a company their size. Even if they're successful they won't have any kind of moat. They don't have a modern equivalent to the Office suite on the horizon and things are changing too fast for anyone to have a safe bet on what that would even look like. It's unlikely to look like Copilot! And very likely to come from the model providers, which MS has so far failed to become.

    They have their cloud and datacenters, where they're second place. If they start to lose their enterprise pipeline into that ecosystem, Google is well prepared to take their spot.

    It's not shocking anymore to see big successful companies make dumb moves, but MS seems particularly out of touch with reality at the moment. It's too bad because Bing is the only realistic competitor to Google search. DuckDuckGo you say? That's essentially Bing in privacy mode.

    5 votes

  7. Comment on Sam Altman may control our future—can he be trusted? in ~tech

    post_below
    Link Parent
    idk... I thought the ending was deft. It moved from catching him clearly contradicting himself, implying that he's after money and power despite his claims to the contrary, and then brings it home...

    idk... I thought the ending was deft. It moved from catching him clearly contradicting himself, implying that he's after money and power despite his claims to the contrary, and then brings it home by using Altman's own words to make a veiled comparison between Altman's 'magic' and LLM sycophancy and hallucination.

    It would be hard to walk away from that read with anything but the impression that Altman is full of shit. Tempered, I guess, by grudging acknowledgement of his Jobsesque salesmanship.

    It reminds me of middle stage Musk with even less attachment to the truth. I wonder if he'll go as completely off the rails?

    3 votes

  8. Comment on Sam Altman may control our future—can he be trusted? in ~tech

    post_below
    Link Parent
    The author didn't call him a liar, journalists shouldn't want us to trust their opinions. Thus the neutral(ish) tone. But they did spend an absurd amount of time getting quotes from people calling...

    The author didn't call him a liar, journalists shouldn't want us to trust their opinions. Thus the neutral(ish) tone. But they did spend an absurd amount of time getting quotes from people calling him a liar. It's essentially the thesis of the whole novella length article. Sociopath was also mentioned more than once.

    Cheers to them for the public service message, even if it was also a guaranteed viral hit.

    4 votes

  9. Comment on AI Coding agents are the opposite of what I want in ~comp

    post_below
    Link Parent
    That's the right direction (meaning automation, not local models, they aren't quite there yet). You can call agents programmatically so if you wanted a collection of agents watching over your...

    That's the right direction (meaning automation, not local models, they aren't quite there yet). You can call agents programmatically so if you wanted a collection of agents watching over your shoulder, ready to head off in various predetermined directions, you could definitely build it.

    But really there are already fairly low friction ways to handle those things without starting from scratch. Custom agents, slash commands, skills, hooks, more recently loops and scheduling. Most of the tools you need to customize the environment any way you want are already built into Claude Code and (always just a little behind) Codex. For anything that's missing, both CLIs have a headless option.

    1 vote

  10. Comment on AI Coding agents are the opposite of what I want in ~comp

    post_below
    (edited )
    Link Parent
    Not really that out of date, November 2025 was the last big leap in capability. Which is odd, if you gave a frontier coding agent clear structure and pseudo code in January 2026, generally it...

    Not really that out of date, November 2025 was the last big leap in capability. Which is odd, if you gave a frontier coding agent clear structure and pseudo code in January 2026, generally it could turn that into working code even if that specific pattern wasn't in the pre-training. The top models are really good at translating from one language to another, even if the target language is code and the source is well structured english. They have a huge corpus of patterns to match against in both languages. The exception maybe being obscure programming languages, I haven't tried to use an agent for a language that wasn't in training but I imagine it would be more trouble than it's worth.

    Context matters a lot in this case, if you were to prompt "write code to solve X problem" and there were no solutions even vaguely similar in the training, the agent might struggle.

    If you said... "Our goal is to solve this [problem]. We want to normalize X data and then do Y with it, based on its relationship with Z. The result needs to conform to [criteria]. Let's talk about how to accomplish that". You'd be on your way to a solution. Or to put it another way, on your way to creating sufficient well structured english that the model could reliably translate it into code. Note that the above is an overly simplified example.

    As for myself, one of the first steps in solving a novel problem is often gathering information. Agents are really good at speeding up that process. By an order of magnitude sometimes.

    2 votes

  11. Comment on AI Coding agents are the opposite of what I want in ~comp

    post_below
    Link Parent
    Indeed. I've said this before, but it's going to bear repeating for while: In practice coding agents aren't simple tools. In theory they're simple, but two different people will get very different...

    They are? Those all exist.

    Indeed. I've said this before, but it's going to bear repeating for while: In practice coding agents aren't simple tools. In theory they're simple, but two different people will get very different results for the same task unless they're using identical scaffolding and similar prompts.

    It's not an all or nothing between pure vibecoding and "just autocomplete". You can decide what the tool is for and when to use and not use it. You can absolutely use agents just to audit for security issues, or just to write tests, or whatever.

    But the key point is that, unless you're doing something very straightforward or someone has dropped you into pre-made scaffolding that provides a specified workflow, you're going to need to play around and iterate to get it working the way you want reliably.

    Possibly part of the problem many have is that a lot of the conversation and marketing around agents is some version of "ask for a thing, get the thing". Which only works sometimes, and pretty rarely for certain kinds of tasks. People read hype pieces and get a distorted idea of how to use the tools.

    One of the things I love about building software is the space where the best answer isn't immediately obvious and there isn't an established canonical way to handle it. That's the most satisfying kind of problem to solve and making agents do what you want is very much that kind of problem. Everyone is still figuring it out in real time, the model providers included. They didn't build the models to deterministically solve particular problems, wasn't even really an option. They just built general models and then paid close attention to how people ended up using them and then worked to support that via fine tuning and harness design.

    It's uncharted territory

    16 votes

  12. Comment on Enjoying reading in the age of LLMs in ~humanities

    post_below
    Link Parent
    Agreed, I should have said qualified it with "completely". No question that it's already replacing a lot of human writing.

    Agreed, I should have said qualified it with "completely". No question that it's already replacing a lot of human writing.

    2 votes

  13. Comment on Scientists uncovered the nutrients bees were missing -- colonies surged fifteen-fold in ~science

    post_below
    Link Parent
    It sounds to me like you're knowledgeable enough. Those are great solutions, and even a somewhat stripped down version of state constrained capitalism works great in a lot of places. Even the more...

    I think putting limits on capitalism and greed would help, encouraging people to think differently about consumption and needs would help, pricing externalities better would help, reconsidering our culture of “I consume therefore I am” would help. But I’m far from smart and knowledgeable enough to tell you which system would embody these things perfectly.

    It sounds to me like you're knowledgeable enough. Those are great solutions, and even a somewhat stripped down version of state constrained capitalism works great in a lot of places. Even the more minimal version of regulated capitalism that is currently being dismantled in the US worked a lot better than the alternative (no regulation) for decades.

    We don't need an alternative to capitalism, we just need a strong enough state to temper the worst of its excesses.

    4 votes

  14. Comment on Enjoying reading in the age of LLMs in ~humanities

    post_below
    Link
    I'm been thinking a lot about writing in the age of LLMs. Well said. I hate the idea that more and more writing will lack those things, and a lot more besides. This is likely true, provided the...

    I'm been thinking a lot about writing in the age of LLMs.

    There seemed to be such a richness in the different ways people would construct arguments, structure those arguments, then deliver those arguments stylistically, not just from the perspective of being persuaded as a reader but also from the perspective of seeing how a given writer thinks, relates to the living tradition of language, and understands the world conceptually.

    Well said. I hate the idea that more and more writing will lack those things, and a lot more besides.

    The reality is, LLMs are capable of writing texts that, if you gave them to a seasoned reader 5 years ago, they'd say it was well written and indicative of a truly thoughtful mind.

    This is likely true, provided the prompt was good, it was a top tier model, and with some iteration. But there's a caveat: There's a big difference between the first impression of AI prose and how it lands once you've seen enough of it to catch the repeating patterns. It's hollow... the signal to noise ratio is usually pretty dismal.

    The problem is that isn't stopping people from using it, and those people are more likely to be from the demographic that has no particular love of writing.

    Personally I don't see AI writing supplanting human writing. At least for now, at the current quality level of AI output. And I'm happy to see that people are still pushing back and rejecting obvious slop. It might be a hopeless fight but it seems totally reasonable to want organic reading material.

    17 votes

  15. Comment on Linux kernel czar says AI bug reports aren't slop anymore in ~comp

    post_below
    Link Parent
    Some extra context... First, distillation doesn't really work, which would be the only important model building technique that would be lost if the big providers folded. The second tier model...

    If OpenAI and Anthropic both implode from the debt, where do these continually improved derivative models for local execution come from?

    Some extra context... First, distillation doesn't really work, which would be the only important model building technique that would be lost if the big providers folded. The second tier model providers are building their models from scratch, they don't need Anthropic, or Open AI or Google to exist.

    Also, a lot of model makers are releasing open weights, which means anyone can use those as a starting point to build their own models. If somehow the frontier AI companies were erased from the timeline, it would slow things down but in no way stop LLM tech from continuing to advance. It would take a complete collapse of civilization to do that.

    5 votes

  16. Comment on Claude Code's source code leaked in ~tech

    post_below
    Link Parent
    Haha, I feel that. Do you mean this one? Definitely not slop.

    Haha, I feel that. Do you mean this one? Definitely not slop.

    2 votes

  17. Comment on Claude Code's source code leaked in ~tech

    post_below
    Link Parent
    Sadly that blog post is AI written too, though with better "don't use these AI tells" prompting and some post inference human editing. Still, it's better slop than the first slop.

    Sadly that blog post is AI written too, though with better "don't use these AI tells" prompting and some post inference human editing.

    Still, it's better slop than the first slop.

    4 votes

  18. Comment on Claude Code's source code leaked in ~tech

    post_below
    Link Parent
    I just want to confirm... the article is LLM slop and the conclusions are mostly just wrong. Everyone already knew how a harness works, and Claude Code had been decompiled before this leak. It's...

    I just want to confirm... the article is LLM slop and the conclusions are mostly just wrong.

    Everyone already knew how a harness works, and Claude Code had been decompiled before this leak. It's still interesting, and there are a few insights a competitor might glean, but the way the article frames the situation is outright misinformation.

    15 votes

  19. Comment on The cognitive dark forest in ~tech

    post_below
    Link Parent
    Like companies trying to drown out the competition could provide long before LLMs. The context of the original article is that implementation is now trivial. For example: Except they can't. Maybe...

    Like companies trying to drown out the competition could provide long before LLMs. The context of the original article is that implementation is now trivial. For example:

    If whole projects can get one-prompted or agent-teamed it becomes just the money game.

    Except they can't. Maybe someday, but until then the whole premise falls apart.

    1 vote

  20. Comment on The cognitive dark forest in ~tech

    post_below
    Link Parent
    It sounds like you're claiming AI coding agents are more capable than they are, and that you're not actually a software engineer. If that's true I imagine nothing I can say will change your vibed...

    It sounds like you're claiming AI coding agents are more capable than they are, and that you're not actually a software engineer. If that's true I imagine nothing I can say will change your vibed thesis.

    3 votes