I like that I can hop on at any time of day, make some progress, hop off any time. I don't have to wake with an alarm and stay out gold farming for 8 hours with 2 or 3 short breaks. I don't have to leave the house. I don't have to talk to anyone. I grew up with MMOs like Graal Online, RuneScape, MapleStory, etc. I have been extremely disappointed by the real world's version of jobs and money. Gold in a game often feels more useful as well. I have actual needs for it. Real life by contrast is vaguely like "work forever and keep making money because that's just what you're supposed to do".
On one hand you have slots/roulette/craps gambling where payout or loss is immediate. Next come candy crush style games where time invested per session is minimal but extant, but reward continues to stream with flashy graphics for accomplishing subgoals. Next come longer-term games like 4x or tower defence where there's progression or some sort (often available to skill or how much time you have to spend) where payoff is delayed to the end but there are still progress rewards like new technology or beating a wave of enemies. Finally RPGs and MMOs where there is (often) no final payout but leveling up, completing quests, acquiring special items, etc. happens at a pace that rewards time spent. At no time is the next step mysterious; there's a journal with next steps which will each provide their own rewards.
Contrast that with real life where reward is on a fixed schedule (N paychecks a month) with no short-term feedback on performance. At best a short "good work" reward fairly rarely from management. There are subtler rewards people get like socialization or status, but they aren't on a quest journal and they arrive less frequently for most people.
I think for extroverts in high social contact jobs it may actually be as rewarding as gaming, e.g. direct sales.
Contractors and spot workers (day laborers, etc.) also get a bit more frequent reward feedback, but far fewer rails unless it's a very canned pipeline of clients.
Contrast with the ancestral environment where hunting or gathering had minutes or hours between reward feedback and our natural senses and instincts provided a lot of the rails (hunger, visual attention to animals and edible plants, potential mates).
this sounds really dumb, but it seems that if people are that motivated by progress bars and blinking lights and flashy animations, we should really be developing video game HUDs into AR headsets to get people to grind in real life instead of video games.
Games are designed to incentivize you to stay in them. Real life has you as a captive audience, and to the extent that it's "designed", it is designed by people also playing the game, against you who have no incentive to encourage you to try harder to beat them.
Discord is a mediocre chat medium - I don't mind using it, but it wouldn't be my first choice.
But people use it to store all information about some games - effectively replacing a wiki. And it is AWFUL for that. Everything about Discord makes it bad for that
What do you find so bad about Discord's search functionality?
Because I see this complaint a lot, but I've personally never had any real problems with it, and it seems to cover pretty much all of the bases I'd want in a search for this kind of platform.
Because it's not semantic search, it's text search. You don't search Discord like you search Google, which is a problem when it is used as a wiki replacement.
It's not even text search. I can know the exact text I'm looking for and get messages that use completely different words just because Discord tries to group related words together when indexing. It's like some lobotomized collated word search, since it doesn't support symbols or quotes either.
Discord's text search is a simple intersection-of-all-words search, with some filters slapped on top. And it's not even the word you typed, it's all words that seem maybe semantically equivalent or similar. When I know the exact word I want to search for (e.g. "loves"), it sucks to always get results for every other related word (e.g. "love", "loved", "loving"). Maybe it'd be a useful feature if it had a toggle, but it doesn't. Some other words that have problems with this are time/times, caring/care, wanting/want/wants, and fucking allowance/allow/allows. Seriously?? Allowance is treated the same as allow??
Can't search for or with symbols. Lots of search engines have this issue. Just ignores them. Want to search for instances of ":3"? You're actually searching for instances of "3". Not much else to say on this other than it's annoying.
I assumed some ulterior motive. Did they pick the cheap option?
Why would they have left their userbase in the same lackluster search situation for so many years… maybe genuinely massive backlog and improved search won’t improve their numbers like $otherPriority?
Discord is very hard to understand for me. But the young people love it for some reason. Had to use it for some projects or hackatons, and it's still a mystery why it's so popular. No one found older links that were deeper in the chats, had to constantly tag people to see anything, random chats pushed the important information away. It was overall terrible for everyone yet everyone still loves it.
And weirdly, I then understood that there's no real global platform for a quick chat that everyone knows and CAN actively use. In universities we use Teams, but beyond that people are never active there...
But I also don't understand Twitter or Mastodon. It feels like there's so much better ways to both post stuff and reach people.
Hey, I am that guy that founded that one obscure Linux distro. And yes, it is better than whatever one you use at supply chain security if nothing else. ;)
Really though, I do think people often equate popularity with being a healthy default solution for society.
A lot of people enjoy mcdonalds, and hell even I cave and buy it once in a while, but popular does not mean it is a healthy default for society.
The corpos are incentivized to addict you and trap you. Not to give you the best solution.
If one actually does research beyond "but all my friends use it" they would see Discord is objectively one of the worst things to happen to sovereign and privacy preserving communication on the internet in a long time. Tech communities using it should know better.
Whenever I see an open source project that says "Join my Discord" without any alternatives, I close the tab. If a technical project cannot find an hour to figure out how to setup a chat with their community without mandating use of a surveillance capitalism product, our goals are already too far apart.
> [chat]... without mandating use of a surveillance capitalism product
Sooooooo what is your solution?
You just described every single, non-E2E communication product, because they are all indexed and used by surveillance capitalism, including governments and corporations.
IRC and Matrix are self hostable and end to end encrypted. There is no good reason for tech communities to lock themselves into Discord and exclude anyone that does not agree to their corporate terms of service.
IRC definitely is not E2E in practice. Matrix hasn't reached feature parity with discord at all, unfortunately (and the main server has a big spam problem for any public chat. not to mention a publicly accessible chat and E2E are kind of a contradiction in goals).
I am sure popular public Discord servers get their own share of spam, but with Matrix you can host your own, or join one of a friend you trust. That is what freedom looks like. E2EE is optional.
OTR supports IRC, but in practice either solution beats the privacy of Discord, which requires all traffic to be sent in plain text to their corporate owned databases.
> OTR supports IRC, but in practice either solution beats the privacy of Discord, which requires all traffic to be sent in plain text to their corporate owned databases.
As someone who was an irc operator, I can assure you, everything from IRC was funneled into ever place it was ordered to be...
Nope, but I know people who have well and I know the level of quiet cowardly compliance that industry participates in. I want to enable an independent and private internet and US network providers are not in a position to help with that.
What functionality? Someone technical can do literally anything Discord can do with FOSS tools.
The functionality that should matter to free software projects is... freedom. I can forgive laypeople for not knowing any better, but it is pretty hard to overlook FOSS devs that should know better that -still- endorse Discord for FOSS development.
You _can_, but if you're running a FOSS project you probably already have a significant shortage of spare time. Most FOSS devs are not that idealistic: they want something that works and works reliably and not to be building their own platform just to be able to work on the project they actually want to work on.
What a detailed and thoughtful point. It is "shit" so we should all use corpotools?
You do realize there are like a dozen clients for Matrix right? You can literally have any UX you want or even ask your favorite LLM to make a Discord UX clone that uses matrix protocol under the hood. That is how open networks work.
I'll echo the general impression. Even just using Matrix for direct messaging it's not something I am going to be able to pull any of my friends onto unless they already put the privacy and openness advantages very highly. It's just a pain to use, a pain to debug, and flaky enough you often have to debug it. It certainly seems like there's a lot of effort going into it but it seems to have resulted in a lot of churn but not much polish at any level.
it sounds like you had an awful experience with a specific client and/or server combo (or perhaps multiple ones). can you share which ones? as you say, there's a lot of effort going into it, so it's important to know if impression is based on outdated knowledge or whether you or others have had a bad time with the current generation.
Element (and Element X) and the main server. I still use it daily (to message the one other person I know who uses it) but nothing makes me think it's viable as a discord replacement. (I haven't had too many problems super recently but then I haven't tried to log in on a new device for a while)
Matrix has been my almost exclusive chat solution for the last several years to talk to all friends coworkers and familt and it is a major upgrade to IRC in most ways, while also giving me end to end encryption when I want it, and the ability to self host and federate and bridge to other networks. Self hosted at two companies as well and works fine. Those are the features that matter most to me, and Discord does not do any of them.
Whatever issues you are experiencing will not get better if you are unable to describe them in words more detailed than "sucks."
I'll take a crack at this one. I've been using matrix and element for over a decade as my main messenger, and it's great, but it also does suck. I even use it for my business chat. It's pretty much the most popular federated encrypted client right now, and it seems to have taken over XMPP's lead. So here's a nice list/rant about why matrix/element sucks.
1. I spent 15 minutes last month helping my business partner sign onto ElementX on his second phone. He's signed on his PC and his primary phone, but the second phone was very difficult. Basically type username and password, but now you have to verify with emoji. How do you do that? Well, sometimes it pops up on your other clients that a new client needs verification. Sometimes the popup never shows, and there's no way to find it. But we found it eventually. Verify via emoji. Great! Let's do it. Oops, it timed out before he could finish. Apparently it has a super short timer that even I struggle with it. Eventually we gave up and used the recovery key, which worked, but you know many people are going to lose that key since it's not a password they can just memorize. Most other messengers, it's just username, password, and then click the email confirmation link. Boom, easy.
2. Audio/Video calling/conferencing. There are apparently 3(?) different technologies in Element alone, depending on which version of Element (desktop, web, mobile). The "old" one is jitsi. The new p2p one appears to be only available to clients that aren't Element? Or you can only create conference rooms in those other clients (commet?), but you can join them from any client. But then those new rooms are sometimes not join-able from Element clients. They just don't appear in the room list at all. Even clicking a link to the room says it's the wrong "room version", but it's a freshly-created room. So I delete the room and make another new one. But now it's a Jitsi room and it doesn't work in this other Element client (all first-party actual Element clients, and all broken in slightly different ways!). So... it's working now, but only after lots of trial and error and lots of cussing from my business partner. And I have to somehow justify that this is better than using Slack/Discord/whatever corporate thing.
3. The official client is fragmented. On mobile, there is Element (classic) and ElementX. You're _supposed_ to use ElementX, since it's the new one, but ElementX does not have all the features of Element (classic), so you might need the Element (classic) client for those things. But Element (classic) is not getting updated, so you're left in the cold as far as updates and the new features (meaning you might not even be able to cross-verify your clients, or join video calls, spaces, etc.)
4. The official matrix server (synapse) is pretty slow. I hosted a couple small communities (~100 people or so?), and synapse somehow kept getting overloaded and crashing. Maybe it needed more than 4GiB, but that seems kind of silly. There are decent 3rd-party clients (continuwuity is my current favorite), but there's no migration path from one server to another without wiping your entire database and starting fresh. So.... my main homeserver is kind of stuck on synapse.
5. Bridges. Yes, you can bridge anything to anything, but have you actually tried it? Interface is super clunky. You have to run a separate bridge service, make sure it swaps keys with the matrix server via a yaml configuration dance, then talk to the bridge's bot in a dedicated matrix channel with special commands that may or may not get your bridge working. I've used signal, discord, telegram, and irc bridges, and they're all a mess in one way or another. My most recent attempt was with the signal bridge. I made the mistake of enabling end-to-end encryption. One day, the bridge server restarted and lost all its encryption keys. Cool, now all the bridged rooms are dead. So naturally, I delete the bridge service and re-configure it from scratch. Now it doesn't work because its bridge bot username was already taken by the previous instance with the lost encryption keys. And I can't just delete that bridge bot user because _matrix does not allow deleting users ever!_. So now I have broken signal bridge rooms and broken signal bot users that I can't ever delete, even with the admin panel. So naturally I open up a postgresql client and delete them manually. Oops, now my matrix database is broken in ways I cannot even begin to comprehend. Because deleting users is not allowed. So now I just can't have a signal bridge. _maybe_ there's a way to make it work, but I've been programming and sys-admin'ing for over 20 years, and it's just beyond me. I'm going to have to (again!) wipe out my homeserver and start over, losing all my messages and contacts (again!). XMPP had very nice transport services, and Psi+ had a very nice UI for configuring and organizing them.
6. Multi-identity support is lacking in all the official and most community clients. If I have a work account and a personal account, I need to start two separate Element clients. Each one is a hefty electron app, and they all have the same systray icon, so I need to trial+error every time to open the right one to message someone. So far, there's only one client I tried that had proper multi-account, and I think it was commet. It was ok, but I hated that it was a flutter desktop app where every click felt super unresponsive. So I went back to Element. Most XMPP clients have multi-identity built-in. I'm not going to use my work identity for personal chat, just like I'm not going to use my hackernews account to message co-workers.
7. No message auto-deletion. Many people like to claim that Element is more secure than Signal. It is not. If you live in an oppressive country where things you say can easily result in prison or torture, you need more than E2E encryption. You need messages to be gone. You can't trust that your device or your friend's device won't get confiscated and forced to reveal your password. Even the United States has forced device unlocking at the borders. Signal (and Confide, and even Telegram) have automatically deleting messages, so even with the keys, there's nothing to see. Matrix stores messages forever. Aside from some complicated server-side setting, there's no way to automatically delete them. You have to right-click and manually delete EVERY message if you want future privacy. Nobody's going to do that. And I tried enabling the synapse settings for auto-deleting messages, but they didn't work. It seems that feature isn't really tested, or it's documented poorly.
8. The spam/CSAM problem when running a public community. The admin tools are a bit lacking, so without some considerable expertise, it's often not worth even having a community matrix server. So folks just make a discord instead, which has actual moderation tools builtin. Yes, I know there are external moderation tools you can install, but why should that be required?
9. The protocol is too complex to make full-featured 3rd-party clients. This isn't IRC where you can make a basic client in 3 lines of perl. You'll notice that the 3rd-party clients are all lacking in some features of Element, the official client. Even the official ElementX client lacks features of Element Classic. Most folks are stuck on the official electron-based Element client, which has a fairly clunky UX (try explaining to your grandma how to cross-verify clients).
This list is just off the top of my head. I could probably write a whole book on the problems with the matrix ecosystem. But again, matrix is the best we have. IRCv3 hasn't really caught on, and the need for an IRC bouncer to receive offline messages isn't commonly excepted in 2026. XMPP was/is ok, but it hasn't been commonly used since Google Hangouts pulled the bait+switch on us 10+. I keep thinking I should really just go back to XMPP, or start using delta.chat since "everyone" has an email address.
I often can't even get fairly technical folks (who write code by hand, etc) to use Element, even if I do all the server setup and just hand them the keys. It's too hard for most folk. It's too hard for me, and I've been maintaining my own matrix servers for 7+ years. I maintain 3 matrix servers right now. One synapse, and two continuwuity. NixOS makes it pretty easy to maintain servers, actually, as long as they're not open to the public.
And if anyone asks me "what should we use instead?", I'd probably say XMPP or IRC. And we know neither of those is perfect, either, and nothing open source and federated is as easy to onboard like Discord, Signal, or Slack.
But even though Element/Matrix sucks, it still works well enough for me to use it as my main communication. At least until I finish my own multi-protocol messenger. Which I started about 6 years ago and haven't had time to work on.
Disclosure: I wrote some of the code on Confide's backend back in 2021. I do not currently work at that company and am not promoting or un-promoting the app in any way.
I think the oldest cartridge that would still work in a modern firearm would be .22 Short, which is from 1857.[1] It was developed into .22 Long, which was developed into .22 Long Rifle, which new firearms are still chambered in today. It wouldn't cycle a semi-auto though. Another contender would be .38 Long Colt, which is from 1875 and will work in firearms chambered in .38 Special or .357 Magnum.
The main causes of incompatibility were the transition from ball-and-powder to rimfire cartridges, then the transition from rimfire to centerfire. It's surprising how little firearms have changed in the past century. 9mm Parabellum is 125 years old. The Browning short recoil action (used by most handguns today) is over a century old.
> It's surprising how little firearms have changed in the past century.
Isn't that a function of constraints in the deployment model, which are fundamentally unchanged? A soldier needs <something> with him to hit enemies and <something else> to reload it - and the simpler the better, to facilitate difficult deployments and keep costs down. I mean, the venerable Kalashnikov is still around, despite its well-known shortcomings...
I guess we would see major developments only if we discovered/engineered materials that can perform the function in a significantly better way.
Modern metallurgy does allow for significantly higher chamber pressures, and this has been done in firearms such as the M7 with its 80,000psi .277 Fury.[1] Pistol cartridges can also be made smaller for the same effectiveness (allowing higher capacity in the same form factor), as has been done with .30 Super Carry.[2] But your guess is correct: They don't really catch on because a small improvement in capacity or effectiveness isn't worth changing all the manufacturing and logistics for new cartridges. 5.56mm and 9mm are good enough, and they're cheap due to their large volume of manufacturing. Any replacement would have to be much better to be worth the cost of changing over.
Most improvements have been in things like modularity, sighting systems, and standardization for mounting accessories such as lights and laser aiming modules. Better metallurgy does mean that modern firearms tend to last longer than older ones, but even a firearm from WWI can fire thousands of rounds before the bore erodes enough to hurt accuracy. I doubt we'll see any significant change in small arms unless something revolutionary comes along, such as portable energy weapons or mass drivers.
If you live in a country or region where the gauge didn't change, you can put a 1845 train on a modern track just fine. If the gauge did change, you'll need a train from another country. Standard Gauge was first applied in 1825 after all.
Although on modern railways whilst the locomotive might fit, the train might not be allowed to operate in normal mainline traffic or need modifications to be considered safe. A steam train on the British main lines needs to be retrofitted with an alphabet soup of AWS, TPWS and GSM-R, and the glacial rollout of ERTMS and in cab signalling will mean figuring out how to add a HMI display to a steam engine cab.
And Apple might find software that doesn't get their magical signature to be unsafe and insecure, but the old executables will run if you ignore all those constraints.
> whether you could put a train from 1845 on a modern track
This quite literally happens all the time in the UK. There are heritage railways, and steam rolling stock regularly shows up on various routes as part of heritage rides.
On the other hand, I ride almost daily on a track bed that was laid almost 180 years ago, and through tunnels and on viaducts that are equally as old. The rails, sleepers, signals, and rolling stock have all changed, no doubt, but the base infrastructure is old hat.
The oldest bits of the London Underground are from 1863. The tunnels are exactly the same, the stations and stairways, and even some of the sign boards completely unchanged from then.
Railways are the epitome of forever infrastructure.
Many militaries are still buying rifles that chamber a cartridge introduced in 1891 (7.62x54mmR). I'd argue a cartridge initially developed for a long bolt action rifle (Mosin-Nagant) that can now be fired out of a variant of a AK-patten select-fire bull-pup (Vepr) fits the analogy.
idk why this idea is so prevalent. sure theres weird avant-garde jazz, but the vast majority is usually relatively simple "standard" with improv over it.
Sergey better really go into "founder mode" or else Thomas Kurian (head of GCP) will continue to eat Deepmind's lunch.
Between selling TPUs to external companies and now hosting competitor models, Gemini just doesnt have the full faith of the Alphabet capital allocators, and I can't really blame them. A mature company like google would rather just expand their extremely profitable cloud business rather than dump billions into a black hole on the off chance they achieve RSI or superintelligence or whatever.
> Our cloud business is at any given moment in time forgoing short-term revenues because we are making compute trade-offs to support frontier AI development
> Our frontier AI business is at any given moment in time forgoing short-term development because we are making compute trade-offs to support our cloud business revenues.
Infact all companies, at all times are doing tradeoffs. You and I can be doing something else right now but I'm here on HN replying to you - and that's a tradeoff.
If someone suggests -just do X then you don't have to do a tradeoff, that just sounds wrong to me.
The primary trade off amounts to CEOs enriching themselves by sucking resources out of the company they control. They trade the company's future and their employee's incomes for personal wealth. Investors are the bag holders for CEOs and their insiders.
We’ve been seeing compute allocation failures in GCP us-central1 for about a month now- that can’t be related can it? These aren’t TPUs, this is normal compute for dataflow/etc jobs.
Infra might very well be the right play. Over the past year, it seems like the evidence is on the side that there's no "secret sauce" in AI, and models will all converge towards similar capabilities. Not much margin to be had for having the best model for a month or less until another lab leapfrogs you by 3%.
If there's no point in having the best model, there's also no point in leapfrogging the current best model. Conversely, labs trying to leapfrog each other indicates that they consider the advantages of having the best model (or at least a better one than they had before) worth it, even taking the competition into account.
I mean the labs are in a tricky spot where they have to justify massive valuations. "We're going to stay where we are and focus on inference" isn't the growth story you need when you're trying to justify a 2T valuation on 80B in revenue (unless you dress it up as a safety concern).
Google can afford to take their foot off the gas without killing their business.
What does “capital allocators” mean here? If your implication is that the Gemini org is resource starved and a few extra dollars will put their model in the lead that is 100% incorrect. If anything it’s the bloat and bureaucracy keeping it behind.
Signing partnerships and hosting third party models costs Google nothing, and in fact brings in a ton of revenue that will eventually go back into Gemini.
Google uses Gemini for its AI Overview on Google Search results, though. And it seems to me that this is an area where they really have a big incentive to run their own AI that they completely control. AI Overview sits at the top of the Google Search results first page, so when it comes to marketing it's sitting in the most coveted place on the entire Internet.
It's also marketing that their AI is hilariously terrible because AI Overview is the biggest posterchild for AI slop and often used as an example of how bad Google is at AI. Gemini proper isn't that bad, just the ultra cost cutting version on search is.
Yes, I'm realistic about it too. For simple queries it nails it more often than not. It's only asking it more complex technology related questions that it falls apart for me, and in those cases I already know I should be talking to one of the big boi models for certain lines of questioning.
Google certainly ain't gonna run a several T model for every search that's for sure (well, not yet).
I think it makes sense for them to separate the application/productivity layer from the model layer. And offer models from many model providers, including OpenAI.
To stay competitive with Microsoft and other productivity suites.
AI hosting currently generates revenue, so the choice is trivial. Building an LLM model is company PR, while hosting another’s model provides immediate income.
Blizzard has not been independent since 1994, but people on the internet love picking some arbitrary point in time when "Blizzard died" when in reality the timeline and quality of their releases has almost zero correlation with who they were owned by at the time.
I like Jason Schreier's take that the end of Blizzard began as soon as WoW blew up in popularity.
"Imagine a Blizzard that never made or never found the success they did with WoW and continued on the path of boxed releases every couple years."
I did find myself rather salty back then that it felt like Blizzard abandoned anything outside of WoW until 2010. WoW wasn't something I ever grew an interest in, yet I played an absurd amount of SC and WC3. Then SC2 was a huge miss for me, as well as every game after it.
Yes, of course there's the financial performance of a studio and then the identity of a studio that people get attached to.
I didn't get into Blizzard games because they were a financial powerhouse, I got into them because they were fun as hell and the identity of Blizzard at the time gave me faith that their next release would be just as fun. That faith started deteriorating around 2005/6 for me, and was gone within a month or so of SC2 releasing. The identity of Blizzard of the 90s and early 2000s that I fell in love with was gone by 2010.
I can go into more detail about the shortcomings of all of the games from 2010 on (IMO, obviously), but the overall point is that they are fundamentally different than their older games and might as well have been from a different studio.
Blizzard has not been independent since 1994, but people on the internet love picking some arbitrary point in time when "Blizzard died" when in reality the timeline and quality of their releases has almost zero correlation with who they were owned by at the time.
Yesterday’s frontier is today’s commodity, so I don't think frontier AI is out of reach. With computing and solar energy costs continuing to drop, the current AI supply shortages are just a temporary blip. We’ll probably land right back in standard boom-and-bust cycles.
Either frontier AI will continue to get better steadily, until it's better than people at basically everything, and very weird things start to happen, or, it plateaus and the open weights models will eventually catch up.
There's a third possibility, which to me is also the most plausible. Models will keep getting better and better within certain domains, while never catching up with humans when it comes to other type of tasks. Like that elusive quality-without-a-name some have called "taste". Or knowing what UX is most intuitive to humans. Really, anything where we have to tap into our common humanity and our human experience in order to be of help to other humans and generally to life on earth. Which is an all-important skill, since the flowering of life on the earth is (or should be) the whole point of anything we do, after all.
We've spent a couple of millennia documenting that elusive thing called 'taste' in word, sound, image, lately in moving images. I'm pretty sure that even if you didn't have taste you could make a good simulation based on all that input.
So what you're saying is that since there are so many examples of what humans have preferred in the past for these models to be trained on, surely there will come a time when they will know even better than us what we want and what we like.
Everyone knows the path followed by every single technology mankind has ever developed - including the incredible run that was Moore's law. Sure, some will think "This time is different" - but is it?
It doesn't need to run indefinitely. I don't think anyone sane thinks it's going to run indefinitely. It just needs to run long enough. Moore's law was what, 50 years? If AI keeps improving at the rate it currently is for 20 years, the world will be very different when that curve flattens out, in a way that makes most of our current concerns are entirely irrelevant.
Those cycles have shortened considerably. Just look at the lifespan of media, integrated circuits, software, products. There have been so many big model releases just in the last three months I've lost count. We're now closer to weeks between steps.
Maybe if China stops releasing capable models the pace will slow down a little, but they have lots of incentives not to.
It is, because just like you mentioned, there's no exponential physics to back up growth. LLMs have to stop scaling up compute, because there's only so much resources.
But even if we think back to the previous decade or two (sans AI), most of tech has been a a red queen's race - developing and improving stuff, without any real substantial results.
Microservices and orchestration are not better than old monoliths. Electron apps are not better than native. Most programs people actually use have decades-old codebases.
Video games of today don't look substantially better or play different than ones from 15 years ago.
But it is not a constant point of "this % of people needs the frontier intelligence".
As thje models get better, the % of the pie gets smaller and smaller. So higher and higher cost of getting frontier model out gets divided onto smaller and smaller customer base as other models catch up
Yes, this is the crunchy bit and we're already well underway on that curve. I haven't touched a frontier model in months because I can do with smaller ones without wondering/worrying when I'll get banned next because what I'm doing is dual use. Qwen was the the first time I thought oh, ok, so this is 90% of what I need. Then DS4 made that 95% and with GLM 5.3 it's somewhere between 99% and 100%. Now, I'm not trying to solve any Nobel prize winning stuff here but if that's the kind of customer that will remain I don't think the frontier model providers are going to like the next six months or so.
I am actually doing robotics stuff and yes, GLM 5.3 can do it.
As for the manual feedback: I want that manual feedback, but if I let it rip it will do just fine all by itself, but then I have no clue what is going on under the hood.
At some point you have to let go of trying for full understanding and control, delegate to the AI, and unlock the real productivity boost.
I have no clue about the state of robotics.
For software, I know that Opus 5.5 can handle more than GPT 5.6 Sol. I am willing to pay for the difference rather than use the cheaper DeepSeek Flash.
> At some point you have to let go of trying for full understanding and control, delegate to the AI, and unlock the real productivity boost.
The prerequisite here is that you can trust it not to do something that gives criminals access to your bank account or allows users to consume an unlimited amount of services it costs you money to provide for free or bankrupts you through legal liability etc.
It's basically this:
"A computer can never be held accountable, therefore a computer must never make a management decision."
Notice what this is actually saying. It's not saying that the computer can't generate a given output. It's saying that if the computer makes a mistake and you don't bother to check, you're the one who has to eat it.
I'm getting plenty of mileage without that and anything with a motor in it requires that you do your utmost best to ensure it works within the design envelope.
My argument is that if safety and liability make the use of AI without fully understanding the implementation impossible, the model is arguably not good enough.
For serious engineering work not only physical designs, software also.
When we manually tested nuclear power plant controlling software, each test case was printed out on paper and tested and then signed by the tester. A lot of paper was produced and archived for each software release. (Archived for as long as the software was running in any nuclear power plant).
Everyone understood, that in a case of significant accident at the nuclear power plant, accident analysis would be done and if the root cause was an software error, this could have possible legal consequences for the developers and testers. (Of course nuclear power plants have many independent and redundant hardware controlling mechanisms, but incorrectly working controlling software could damage the power plant in a very costly way).
It was you who shifted the conversation away from software to make an argument that "anything with a motor" requires full understanding of the software implementation, and that for this reason GLM 5.3 is good enough.
I think you're entirely wrong, and we can leave it at that, seeing that you're clearly unwilling to engage constructively.
The practice of deliberately engineering software without full understanding of the implementation, especially when it comes to robotics, should be illegal. Technology serves society alone and it should progress at a pace that is sustainable for adaption. Your excitement for a “singularity” does not outweigh someone else’s health and safety.
And over a increasingly small slice of time, as, due to the commodification of these things, the frontier businesses will see their customer base walk to the next one that does any better on their task.
Every time they throw someone out with a message about how this request is definitely against their safety guidelines they are clinching the deal on someone running something local or switching to a provider that just lets them get on with their day. Their own fearmongering is going to be part of their undoing.
The market for "frontier intelligence" will be too small to serve outside of certain government/military applications that exist in spite of market forces, in the same way that the market for small, portable nuclear reactors is exclusively limited to submarines and aircraft carriers. When I call a plumber, I don't need a frontier plumber, I just need someone who's baseline competent.
> market for small, portable nuclear reactors is exclusively limited to submarines and aircraft carriers
SMRs are being commercialized. Likewise nuclear cargo ships have been seriously studies. Nuclear reactor powered (not just RTG) satellites have been built.
I know this is beside your point, but I think if we figured out how to commercialize this, demand would come.
But why wouldn't you want (and pay) for the very best plumber? A frontier plumber will optimize the time their customers poop, pee, bathe, or do dishes 10x to 200x, and with their freed-up time,they'll outcompete those who don't pay for frontier plumbing!
Sir, I take the impending plumbing singularity very seriously, and wouldn't want anyone to be left behind. When plumbing hits recursive self-improvement, we may see toilets that shit themselves more efficiently than any man can, and finally attain ASI (Auto-Shitting Infrastructure).
I can see you do. I beseech you though to think about the consequences for other industries. For instance, the ratio of toilet paper:organic product will need to be maintained somehow and I suspect that if we follow your pathway to its logical conclusion that there will be a cellulose crisis unlike anything seen in the world since the USSR went bust.
The next move in my investment strategy is obvious though, invest to the hilt in PVC tubing and I thank you for this invaluable insight. Fortunately this forum has no visitors so we can rest secure in the fact that the pigs won't realize it is too late to avoid being slaughtered. Power to the bulls, after all it is their product more than anything that is responsible for the plumbing revolution.
A high margin luxury good with huge r&d costs is a death spiral.
A high per-user r&d means high costs per user. So then users leave to save a buck, qnd the price goes up again for the remaining users.
Maybe it works out OK if you want to signal wealth - the high price is the selling point. But luxury pure utility is a hard sell to consumers, and for luxury B2B it's about high touch sales (e.g. Oracle).
Agreed, but I I wonder what fraction of tasks really require frontier performance, either in terms of sheer difficulty or where speed is a matter of necessity. For many jobs, it's increasingly a question of whether you'd rather have one filet mignon or 20 good cheeseburgers (all from the same cow).
Why is everyone not understanding that good cheap EVs are banned in the US? If people actually got an experience a Chinese EV I dont think it would be nearly this bad.
reply