waterfowl

@ayv4zyan the fact I got locked out in the middle of a coding session when my credit usage was at 34%, because it updated to the new system and immediately shoved me at 100%, is fucking insane. I'm canceling grok and just getting cursor

The problem is not the pool. xAI’s new Grok limit system could have been a great idea. A weekly shared pool across Chat, Build, Imagine, Voice, etc. actually makes sense. A developer may want most of their usage in Build. A creator may want Imagine. A researcher may want Chat/DeepSearch. Flexibility is good. The problem is not the pool. The problem is that the pool seems to have arrived with much worse effective limits, less transparency, and more pressure to top up or upgrade. Before this, SuperGrok felt generous. I almost never hit Grok chat limits. Build felt usable. Imagine felt usable. Each product had its own room to breathe. Now one heavy workflow can eat the same shared allowance and hurt the others. If I spend my week coding, I should not lose normal chat usage. If I spend my week generating images, I should not lose my ability to ask Grok questions. If a generation fails or is blocked, users should not feel like they paid quota for nothing. The right version of this system would be hybrid: Keep guaranteed reserved limits per product. Add a flexible weekly pool on top. Show real numbers, not only a percentage bar. Do not silently reduce what paying users could already do. Make top-ups optional, not the obvious destination of the redesign. For example: give every SuperGrok user a reserved baseline for Chat, Build, and Imagine, then let the weekly pool handle overflow or personal preference. That would make the product better. What happened feels like the opposite: a good product idea used to justify a worse subscription experience. And this matters more because SuperGrok is $30/month. That is more than ChatGPT Plus, Claude Pro, and Google AI Pro. If xAI wants to charge more than the standard $20 AI subscription tier, it should give more value, not more anxiety around usage. I don’t mind a fair limit system. Compute is expensive. Video and coding agents are not free. But paying users should be able to keep the workflows they already had. A new limit system should make usage more flexible, not make the product feel smaller. @xAI @X @XDevelopers please fix the usage limits. We love Grok, but we don’t love feeling robbed by a subscription that suddenly gives us less.

@RockLeeSmile yeah now let's look at the 6% that aren't. Guarantee they're the games people actually care about

@fdksfjdfd I'm getting tired of this. I just refuted this point twice in the last 10 minutes. No, you're wrong. Doesitplay.org found 94% out of 773 PS5 games are fully playable offline directly installed from the disc.

I'm really curious psychologically what is going on with the people spending their time advocating for Sony's side (destruction of physical games) online, making arguments for us all to have less consumer rights. Can anyone explain it?

@mSanterre I hate saying it because it's elitist, but web dev really is just pretend coding. Vibe coding web dev stuff is so freaking easy. I could never imagine vibe coding game code.

Some programmers have never done anything except basic CRUD web apps and it shows.

@FreakZoneGames at least with a PC you can get pirated/cracked versions and put them on a disc. Or buy games from GOG which offer DRM free installers. The real killer is DRM + digital only. That's a recipe for disaster. Steam DRM is trivial to crack, which is good for game preservation

what the fuck

@avidseries let me guess. You would have supported the Confederacy

On the 250th birthday of America, one of the most un-American tweets appearing on X was posted by the owner of the platform.

@effectfully saying a compiler is not deterministic by bringing up a few edge cases is beyond dishonest. LLMs are by definition not deterministic due to the random nature of how tokens are chosen.

"You can't compare compilers to LLMs because compilers are deterministic" -- no. For virtually all intents and purposes compilers aren't deterministic in the same sense that LLMs aren't. Some compilers are literally non-deterministic, for example GHC generates names differently every time (it's a pain in the ass). But even if the compiler is deterministic, tiny changes in the input can result in major differences in compiled code (which is partially what people mean when they say "LLMs are non-deterministic"). But even if you fix the input in place, any compiler involves over time and it is normally not worth ensuring that the new version of the compiler produces the same code as the old one. That's not even mentioning backwards compatibility issues. A compiler, as an ever-evolving tool, rather than a single snapshot taken at an arbitrary time for the sake of a dumb argument, is non-deterministic in practice. Much like an LLM is used non-deterministically in practice, even though there's nothing inherently non-deterministic about a pile of weights. So stop using the determinism argument, it's irrelevant, that's not the defining difference between compilers and LLMs.

@JozsefSzalma @DavidSHolz this is it. When you code, you get to enter a flow state and just bang stuff out. Vibe coding is rapidly testing, solving problems, over and over. It's exhausting

@DavidSHolz I feel the issue is that we are only left with the more cognitively challenging parts of the job, like making architectural decisions, without the zen-like clickety click in between.

my friends are all feeling extremely productive and also extremely drained with the latest coding models. this makes me feel like something is wrong, and also that there might be a big opportunity. does anyone have any strategies they use to make it feel better day-to-day?

@elonmusk one has to wonder how much of that is due to circumstance though. If Germany was next door during WW2, would we have perhaps annexed some of it? It's worth thinking about. We annexed stuff vs Mexico, and vs Spain, in wars with them

Obviously true to any student of history

@notch depends on the project. I'm making a game data tool right now that's basically a JSON editor for bespoke structures and schema, and haven't read a single line of code. The tool outputs diffs + raw output blocks so I can verify it works. I would never do this with actual game code

How much better do the pull request have to get before you stop reading the code?

@ryanbrewer as someone who learned a second language, LLMs are extremely useful for it. But language learning was never a knowledge problem, it's a throughput problem. You literally need to rewire your brain, and no LLM can do that for you.

It’s shocking to me that LLMs didn’t create an educational renaissance. Shouldn’t I be able to learn a language in a month? What did we get wrong?

@DrMacEachern @astropol0 for some reason I don't see this UI when i check usage... it only shows my grok build credits

@astropol0 Customers go elsewhere when things like this happen. I simply stopped using it. I guess I can’t run out of tokens if I don’t use it. 🙄

These weekly limits are actually painful paying $30 for SuperGrok and running out of usage in just two days feels so wrong. xAI is slowly stripping away what made Grok feel open and generous. It’s sad to watch it become just another limited and corporate AI... Grok is losing its soul

@signulll I think in highly coupled systems like game programming, vibe coding simply does not work, and works better as an assistant. There are even more complex things out there than game programming, like avionics programming. Saying SWE is 'ended' is peak retardation and ignorance

fable, i think, likely marks the end of software engineering as we knew it. it’s prolly the penultimate chapter, if not the end of that entire story.

@Darth_Peachez get starlink if you can afford it. Probably cheaper than other shitty rural satellite internet. problem solved.

preservation is dope and all but i havent seen anyone really talk about the fact that a lot of gamers in rural areas wont be able to game anymore because digital only means you need decent net to download these 100’s of GBs in a reasonable amount of time.

@Its_ereko inbred pedophiles

🚨🇸🇦 Saudi delegation in Tehran paying respects at Khamenei's farewell... After their rulers enabled the strikes that killed him. Pure puppet theater. Iran refused to kneel. The thrones switched masks for the cameras. The resistance sees every betrayal. No more bowing.

@ThePrimeagen Been doing this for a very long time, especially for code planning. One correction though: use a pen. The act of striking through something instead of erasing it does something mentally, and let's you see your idea chain in its entirety

unironic productivity hack for 2026 use a pencil and paper to write down your ideas

@quxiaoyin two wrongs don't make a right. All China has ever done is steal, they can't create anything, only copy.

If “distillation” is theft, then almost every model faces the same criticism: they have all learned from vast amounts of human-created content across the internet, news, books, forums, videos, and more. If your own model is built on knowledge taken without permission, can you really claim copyright over it?

@GaryKildal @theo I'm fairly certain he's never done anything besides javascript

@theo And what would be wrong with that Theo? Have you ever compared assembly to C?

At this point I’m genuinely convinced most of you would have kept reading the assembly code after C got popular

@theo it really depends what you're coding though. If you're web dev which hardly counts as coding, yeah sure, vibe coding works really well. If you're doing anything serious, AI is much better as a sort of partner / assistant, and you absolutely need to read the code.