Posted by lairv 21 hours ago
As impressive as this sounds, whether it is tech teams or agent farms, the progress slows down considerably as the codebase grows. The real pain was seldom about shipping v1, but was more about how the v8 scales and how easy it is to get there, how robust the system remains between versions, and how constraints are held.
A huge part of the challenge is to pick/choose the design and implementation strategy that can only come from experience. That is one of the primary reasons why it's an absurd advice to me to let the llm do the goalseek and freestyle. This may be a no big deal on microservice or any managed runtime software that is an update away, but on an embedded or a high availability system where the field upgrade and hardware replacement costs could bankrupt a business.
if llm designed a circuit, will you bankroll it?
It's hard keeping separate code bases with separate languages in tandem, and while LLMs definitely aid in that area, I can't imagine it's a free lunch.
So I don't see any reason to stick to RN, when we can just use any OS native APIs to our heart's content and our apps automatically feel more native.
I'd also love the ability to take back control over my rollouts, but with Apple's recent crackdown on remote controlled feature flags in response to fraudulent nudification apps, I'd be very careful about this.
Rumors say Apple now uses heuristics to detect feature flags in the binary. While that's not directly relevant to remotely delivered JS, I fear that bypassing App Review in this way could become a quick ticket to account termination.
Same thing happens in many cases with junior or mid-tier programmers slung at a project without supervision.
Indeed, I'm curious how much better/worse one human with "one agent" could do that same functionality in a OS from scratch. Last time I did this exercise, I took Cursor's shitty attempt from "millions of lines of source code" to about ~20K (https://emsh.cat/en/one-human-one-agent-one-browser/), and seemingly one of my conclusions still hold true ~1 year later:
> One human using one agent seems far more effective than one human using thousands of agents
I watch a YouTuber who is homesteading and builds everything. Often times I see him doing things in ways I know will break or fail eventually, but it doesn't matter because he is there and will fix it when it does. You could argue this is a waste of energy, and for a human it might be. For an LLM however I do not think it is.
It seems the problem is the solution, and vice versa.
- Anthropic is finally well balanced across models. Meaning: Opus 5.5 is so good that I do not feel Fable FOMO as I did when it was Fable 5/5.1 vs Opus 4.8/5. And Sonnet 5.5 sometimes makes me forget about Opus.
- Fable is good, but you really have to push the ambition of your prompt in order to see proper separation from Opus 5.5. I personally have a hard time telling the difference between Fable 5.2 and Opus 5.5 for ordinary coding tasks.
- It’s way faster and more reliable: one of the worst things about fable 5.1 was how slow it was an, on top of that, how frequently it would suffer from max output token overflows. I got none of these issues with 5.2.
- Fable is exceptional at drawing. It’s ability to render an image with JavaScript actually has me questioning the value of dedicated image models. I prefer the images I’ve gotten fable/opus to make for me over those from gpt-image, it’s wild to think we’re at this point.
I love seeing this verbalized. We are pretty much at the point that increased model capability will only have utility to folks that are willing to let go of even more understanding of whatever the model is producing.
The best models can/will produce a year's worth of output that would take several days to weeks to properly understand in the span of an afternoon.
For everyone that still wants to feel like they own the end result, just use the lesser models with less ambitious tasks.
Caveat: I think this mainly applies to coding. I would still like the best models possible to be working on e.g. new drug research to cure our various ailments.
In particular, footnote eight:
> This is not a claim about the future. Perhaps almost all or perhaps even all of the software skills I have will become irrelevant in the near future. But we're not there yet. At the moment, programming skills seem higher leverage than ever before.
> There's something a bit funny about the current moment, where it seems like having the skills that models are replacing makes one much more effective than people who are trying to use models without having those skills. I like Max Bittker's framing that models are very good at counterfeiting and the level of expertise it takes to distinguish a counterfeit from the real thing has been increasing as models improve.
As well as this section fairly at the start:
> As far as I can tell, these problems are problems because people are implicitly carrying around a pre-agentic cost model. Ten years ago, the kind of person who implemented a new compiler because their compiler was slow was the kind of person who started programming a new 3d game engine when they wanted to make an indie game, i.e., the kind of person who never ships anything. Most productive programmers I know still have an aversion to this kind of thing, even though it is now actually reasonable to implement this kind of thing now.
Of course, at the moment everything’s still a bit very much difficult to tell, but “productivity”, at least in applicable areas of work, might really come down to who can let go of the most understanding the easiest? Incredibly thought-provoking stuff.
I suppose it depends on the “applicable areas of work” but I would argue the opposite. As in all things, when you level the playing field, in order to stand out you have to know the rules to understand how to break them.
Imagine a fully autonomous Ad agency that creates unique ad campaigns for companies. And have it controlled by someone who “let go of the most understanding.” For instance, maybe a business minded person with excellent technical skills, but doesn’t know how to copy-write and has no visual design training.
Even with the strides in AI advancement, this is an impossibility and I’d argue it won’t happen within the next five years even. You need human agency, and you need human workers who are experts in their field.
This is Bainbridge's 1982 paper Ironies of Automation restated.
https://en.wikipedia.org/wiki/Ironies_of_Automation
https://ckrybus.com/static/papers/Bainbridge_1983_Automatica...
> This paper suggests that the increased interest in human factors among engineers reflects the irony that the more advanced a control system is, so the more crucial may be the contribution of the human operator
Makes sense.
Which makes those interfaces the most critical parts of the system.
Model produced code compiles the same as human produced code. My kids would call this “copium”.
That is, genuinely, crazy talk. That is what addicts would say.
But it is intellectual-washed to sound like the most reasonable thing to say.
It is not. And there is nothing thought-provoking about the mere act of doing heroin.
There might be something in the system and mechanics that lead to people getting the idea to do that, though.
___
We can engage with these ideas and there might be some merit to them, but not with this frame. This frame will (not hyperbole) kill us all.
People will die due to (poorly-engineered) systems understood by no one, failing in completely incoherent and unexpected ways - and those people may be you or your loved ones.
I, well, agree. Of course the statement is nuts, especially so once you consider issues like vendor lock-in, the oncoming atrophy of one’s own skills, etc. But if you define “productivity” as short- to medium-term results only, is it really that inaccurate?
The problem is that it still uses the and benefits of the legitimacy of the word of "productivity" there. Meaning that it is harmful to us to put it that way.
Maybe it’d be interesting to see what could be achieved responsibly by motivated people with models in a world without “feature pressure”/competitive stress (which’d include e.g. no security race and infinite attackers and exploits to keep up with)?
People demonstrably said the same thing about the inventions of... Writing! and Mechanical/Digital Calculators! Yet here we are!
Time is a flat circle!
Then there is these other programmers who built their own operating systems even though their OS professors are claiming publicly their design sucks, who built a local database from scratch in C because his use case did not need the client/server stuff but just all the other relational goodies, or who built video streaming/transcoding software from scratch just because they have ideas and then the world changes around their ambitions.
I think it is not enough to be ambitious or have the technical ability if you want to build something for others to use. You definitely need deep commitment but also perseverance in the face of social adversity. I think this is what saddens people with automation of the technical aspects of programming. It leaves the more psychological aspects as challenges which are not easy to face all the time.
Great point. Coding used to be a natural forcing function for prioritization of features. When coding is basically “free” you have to do the harder job of saying no to things, deciding exactly what to build, etc…
I would say, yes!
Yes, if its less and good. Yes.
> perseverance in the face of social adversity
Sometimes you're Fabrice Bellard, sometimes you're Terry Davis. That is, for almost any idea you'll find random detractors, but some ideas really are generally useful and some are completely nuts. One of the negative externalities of AI is that bystanders are confronted with a lot more cases where they have to work out which one this thing is.
While this iteration of technology certainly _feels_ different, it seems like for now we're still coming down to "it's useful to know how to build a car, but when you need to put food on the table, know how to fix it and make it do what you need it to do".
not caring about the output ~ more output
… which may or may not be useful in the long term, but definitely does help a shorter-term “productivity” measure.
On the other hand though, there’s a pretty damning example against my case here: Say a manager of individual contributors – which is what the situation of an individual with sufficiently smart models and tooling may almost boil down to – didn’t directly care about the output anymore. We’d most likely call that being unhinged/unproductive, too…
It was late at night and I primarily wanted to link the somewhat relevant article, maybe I should have left everyone to reach their own conclusions instead of adding some poorly-worded unfinished thoughts. :x
I cannot see it being a good thing to "let go" of understanding. If you don't understanding anything, then why would anyone keep you around? An example hypothetical situation is, if I have a colleague that outputs a bunch of productive things, but they don't understand it, and I do, then of what use are they? I have LLMs too and also have the understanding.
It isn't clear to me why people are so comfortable with letting go of understanding. And it's a bit bizarre because it always seems to be framed as "you're going to get left behind if you care about understanding". I also never hear or read anyone who has any real and deep experience say these things.
To your point about becoming redundant without understanding, not understanding things will be a problem when an issue arises and the other person isnt able to handle it the right way. However, if llm can address the issue, does the fact that they did not understand the details really matter? From the perspective of a person who gets to decide who gets fired, would they fire someone who is productive but sometimes introduces bugs that they quickly address or would they fire someone who is comparatively slow but rarely introduces bugs? The answer probably depends on the domain, and how costly a bug can be. For most of us, the work is not so critical that you cannot afford bugs. So, productivity wins.
If you are suggesting humans before LLMs rarely introduced bugs, you probably haven’t got the latest models to review your old code. It’s shocking how many bugs we had before but couldn’t see!
You're smart, and you know what can go wrong, so you use models slowly, and you check their work. Your bet is that it will make a bad mistake.
Your coworker is reckless, they don't know or care what can go wrong, so they use models quickly. Their bet is that the model won't make a bad mistake.
If the model doesn't make a bad mistake, your coworker is creating more value to your employer, and your employer doesn't really care if you understand the code. They just want something that works.
Over the last year, it sure seems like models are making less mistakes, so the expectations on the bets are changing too.
For instance: we have calculators. We’ve had them for decades in their “current” form more or less. They are readily accessible and far more reliable than LLM’s - if you a problem 1000 times it’ll be right 1000 times, the rare bug or glitch notwithstanding. But we still teach people basic arithmetic. We teach times tables. We didn’t just offload basic math.
I agree we don’t need to know assembly to use a computer but there is a balance here and companies do not care about a balance.
For software engineering I don't know the right balance.
I've used the "we mostly stopped caring about machine code generated by a compiler; maybe we should stop caring about LLM-generated code as well?" analogy myself.
But even I have to admit there's two big holes in that analogy: if I use a compiler to turn C into machine code... (a) compiler bugs are orders of magnitude rarer than LLM-generated bugs (b) my C code remains the source of truth, or at least source of intent. Whereas with LLM-generated vibecode, there's no source of truth, just an ever-evolving mudball. I (kinda?) like the ideas of "documentation as source of truth" and/or "tests as source of truth" ideas, but I don't think we're nearly at the point of viability.
TL;DR yeah I think actual understanding of what LLMs produce is gonna be important for the forseeable future
You see cashiers summing groceries? Salary calculations on paper? Tax, inventory, accounting, you name it? Nobody does any of it by hand.
We most definitely offloaded basic math.
I imagine you and most of your peers can calculate a tip or at least approximate it without pulling out their phones yes?
To resolve this we should aim for a new intermediate format that is more high level than source code, the generated code is too monotonic to properly review manually. Something like requirements -> detailed specs for llm -> code -> assembly . Then let one agent implement the specs and have another review it. If we stay at the current abstraction level, humans become the major bottleneck.
It really does feel like people have embraced rolling around in the cognitive version of those Wall-E chairs.
Understanding architecture, robustness, the “crumple zones” of a system? I’m all for it.
Taking three days to “understand” how postgres transactions work under the hood for your generic SaaS? Uh..
I have seen a lot(!) of the last type of “understanding” and it’s infuriating. The same types will often not understand the business, its stakeholders, the architectural constraints beyond the technical, etc. Understanding takes many forms.
You do have to keep a sense of when it did the wrong thing and how that was a valid interpretation of what you asked and how to better specify what you want. But that's how it really just shows you the gaps in your own understanding of a problem.
Also I think there's huge differences to the kinds of tasks and workloads people are throwing at it, leading to wildly different perspectives, but everyone's talking about the same "Fable".
This right here. Humans are really prone to over indexing on a handful of personal experiences vs a representative sample. And once someone has formed an opinion, confirmation bias helps cement it, despite evidence to the contrary. Reality is probably somewhere in the middle.
They don’t keep you around, of course. The better these models become, the worse the human job markets get, I’d predict.
I too love to downvote and edit instead of admitting that I'm talking and acting in a way where nobody would want to be my coworker anyways. Where exactly are you demonstrating that you're doing anything to justify being kept around if that's such a big concern for you?
Is it just me or is this analogy completely missing the point of why we care about detecting counterfeits in the first place? Poor craftsmanship of all things is not really the reason counterfeiting is illegal!
I mean, the answer is obvious: the Luddites were right, these tools (not just LLMs but private property, copyright, and capitalist mechanisms) are only tolerated to exist so long as they serve the needs of Capital, but the hypocrisy still frustrates me.
The most rigorous "private property advocates" don't recognize intellectual property as a legitimate form of property in the first place: private property is a natural phenomenon that emerges as a solution to the problem of physical things being inherently exclusive in possession and use, and IP is a creation of positive law meant to artificially apply the methods for solving that problem to another, unrelated question that isn't part of the same problem domain.
> I mean, the answer is obvious: the Luddites were right, these tools (not just LLMs but private property, copyright, and capitalist mechanisms) are only tolerated to exist so long as they serve the needs of Capital, but the hypocrisy still frustrates me.
Abstract concepts don't have "needs", and no one is currently in a position to tolerate or not tolerate anything -- no one in particular is actually in control over anyone else's activities. The cat's out of the bag, and no one has any capacity to put it back in.
Great thought. Question is, can those who do the tolerating (in the interest of capital) even effectively prohibit/prevent access after a certain point, especially now that the cat is not just underway, but well out of the bag?
Assuming the current trajectory of a yearly factor of 10–100 in size/cost reduction [0] holds up, won’t the technology just become cheap enough to be functionally available to everyone?
Thinking about just the domain of the stock market only: getting a "win" on it means getting a leg up on everyone else participating. Best example I can think of of a pure zero sum intelligence contest. Programming UIs might have a point where you just can't get smarter at it, but in an intelligence contest like a stock market, I don't think there's a ceiling.
In "Accelerando" this was covered by the invention of "Economics 2.0." AI agents also got into a lot of lawfare in that book.
You will find a lot of articles from before LLMs about not delegating understanding of your core system or business. You could say that the software is not your core business so you don't have to understand it. But if the software is any good then it should be taking care of a large part of the business. So software is your business and you shouldn't delegate understanding it.
Maggie Appleton suggests some ways that AI models could do a better job of communicating about artifacts to people in https://maggieappleton.com/planning-agents, although she's focused on the use case of the person deciding what changes in the artifact to request from the model.
Someone found a better URL for that (Fabrice Bellard) story (16 points, 10 days later) https://news.ycombinator.com/item?id=49996383
Unless there was another one you later deleted?
Most submissions never receive any upvotes. They’re all “wasted” like that, even if not flagged.
https://www.tomshardware.com/speakers/in-a-blind-test-audiop...
There are many ambitious things I have tried to work with Opus 4.8 and Opus 5.5 on, and it just can't get there. It always falls back to some tried and true method or some valley that it gets stuck in. In those times, it was impossible to push it forward into more creative and exploratory thinking based upon my relatively advanced pushes and contextualizing. So I took the commenter's position that Fable 5.2 will respond better to that type of pushing. But that type of pushing actually means you know more than those who are happy enough using the lower models. So that seems like the opposite of what you said.
One important constraint is this topic of intent (also referred to as alignment). Specifically, when a user requests a feature or change, there is MOUNTAINS of implied context and intent in their comment. Some cultural, religious, social, economic, base on work and personal experience, their use cases, their values, what they intend to get done, etc. Every time these models get smarter, they get a little better at understanding this context. It's combination of reasons. Some of it is parametric. Some is inferred. Some is about selecting the right data to examine and research. As these models get smarter, I spend less time time iterating and educating the model. They reach the intent faster and better.
From what I hear and experience, this is also true in the coding space. LLMs are basically another abstraction level on top of formal languages, which sit on top of compilers and machine code and etched instructions. Smaller and dumber models require a lot more hand holding to achieve the same results. They're less capable of inferring the gaps. Less capable of extrapolating potential negative outcomes of the change they're making. Less/less useful parametric knowledge. Less capable of selecting the right data to use to make the right decisions. They do what they're told.
This has been less of an issue for coders who religiously document, understand the codebase and architecture intimately, and demand to be kept in the loop on every change. So seniors often do not see the utility of smarter models. They know exactly what they want, how to ask, the context to provide, and how to assess if the change worked as expected. The utility is for the 99.9999% of the world which is not a senior developer.
However, your statement that there are only 8300 senior developers in the world does not seem correct to me. That seems like an unreasonably high bar for that title.
I think the issue is fairly fundamental, perhaps because of tuning, perhaps even beyond what they are tuned or trained on, because the bulk of what they can go find and ground into seems to drag them heavily toward implicit biases that split around categories I can’t quite discern. The category boundaries sometimes look like academic fields, but sometimes they seem like more fundamental attractors.
This is appalling. Are you really really going to trust a YEAR's output from a machine? Are you going to read and understand all that? Will you be able to maintain and extend it to your will? I truly doubt that. Drugs do harm and are hard to quit. Good luck.
It ran like a dog. Few nights of running Fable (later Opus 5.5) with a 7 strong team of DeepSeek v4.1 Flash 1017 (don't mistake it with 0731 that is crap) and llama was running 10x faster and vllm was running even faster when before it would not work at all.
I've run similarity and needle tests and the results are exactly the same content wise. But 10x faster. I've relied on this for a month now and I'm pretty happy with how it runs.
Do I now need to learn CUDA Kernel programming to understand how it was done? I don't think so. It's for my personal use. Not for a paying client so no warranty is perfectly fine.
But I also use a lot of AI at work in an I dusty where people may die if the code is buggy. So I absolutely verify and understand everything.
There is a middle ground too. Understanding the architecture.
For example I used a bunch of DeepSeek models supervised by Qwen to build adaptive-acl. A pi coding agent plugin that deterministicly checks if every tool used by an ai during an agentic session matches user intent to give permissions for those actions (reading/writing data, network access etc), as a part of this plugin I run an llm, a classifier that runs over user's prompts and extracts these permission grants then saving them in a format deterministic layer can process.
The design includes massive number of assumptions and compromises. For example, if a user allows writing this implies permission to delete and reading if what was written etc.
The thing also includes over 4 thousand actual instances of tool use preceded by "permissions talk" from my live sessions and that whole collection can be run to test for regression etc.
At some point during one night the AI had a regression and decided the best way to resolve it would be to add regex based language processing elements to the deterministic layer (a huge no no, this has to work regardless of the user speak English or Swahili, also regexing natural language is a disaster waiting to happen).
I knew enough about the architecture to notice that and burn it out with a hot iron.
That is the problem with AI. It sometimes solves problems the easiest possible way causing massive issues for later. So if one uses it for important stuff it is good to understand the architecture and the plan of what it is going to build at the very least.
Having run teams, as much as I love getting into the details of code I care about, in the 30+ years I've been in this business, I've only ever even seen a tiny proportion of the code my teams rely on.
And the non-technical people in these businesses will have seen none of it.
They all take it on a combination of trust and verification of outcomes that the code does what the developers claim it does.
Increasingly we're getting to a point where the output from these machines are more trustworthy than an increasing proportion of human developers, and more importantly: Increasingly often good enough.
Especially as more and more of the code becomes throwaway.
I have a side-project that is profitable, and that is profitable only because I haven't needed to spend time reading the code. If the code turns out to be unmaintainable? Doesn't matter - it's so cheap code that it is reasonable to just throw it away, and start over based on the lessons learned. It enabled me to take an opportunity to profit I wouldn't have been able to profit from if I had to set aside time I didn't have, so even if it was to totally fail in the future as a result of issues with that code it doesn't matter.
As the cost of good enough code falls towards zero, how we think about which code matters and how much we need to understand it will fundamentally change.
Increasingly, a lot of code is even just part of a conversation with an agent to me. It's throwaway, for the duration of a session. Who cares if it lives on?
I don't trust or need to trust most of the code I build any more. I trust the measurable outputs. The verification pipelines we build. The feedback mechanisms we build. Those I need to read and understand.
That's no difference than when working with a team. The only thing that is different is failure rates. And at this point, for some thing the machine failure rates are still higher, for others it is lower.
Given that reading someone else's code is slower than understanding your own code, doesn't that basically cover everyone that adopted coding agents?
I’ve been shifting away from using more traditional GenAI models in favor of just giving LLMs access to a custom drawing language (LOGO-like) to generate imagery and its been going shockingly well - Opus 5.5 in particular.
I put together a simple LLM driven skill that creates old adventure game (EGA) environments from simple pencil sketches and it works really well. And being procedural, they're much easier to make surgical changes to without degradation.
It has three different modes:
• You can simply describe a picture, and it'll automatically generate an image using procedural code.
https://kq-styles.specr.net/#/scene/custom/elk-valley
• You can ask it to paint over an existing image. That's what those examples show, like me creating EGA-style equivalents of various scenes from the original Myst.
https://kq-styles.specr.net/#/scene/myst/r01-c00
• The last way is to give it a rough pencil sketch. It's the mode I use the most since it's very valuable when building an adventure game to position objects in a room.
So instead for each room that needed to be rendered they just saved a series of vector instructions that would re-render it in real time. On slower machines, you could actually see the rooms pop in piece by piece. I wanted to kind of approximate that same aesthetic because it's just super cool!
Then I just guided it to produce a procedural variant specialized in creating new EGA-style scenes for use in creating old-school point-and-click adventure game.
https://people.eecs.berkeley.edu/~bh/usermanual
It's all open source.
Not just "iterate on this XGBoost model" but actually looking at a problem and designing and carrying out a short-term R&D program including probability modeling, model fitting diagnostics, how to respond to various kinds of results, how to interpret/reason about what it finds, etc. That's really where I saw all the models perform poorly, and where I saw the first big step increase from Opus 5 (pre-5.5) to Fable. But then Opus 5.5 came out and I haven't needed Fable since.
Mind you, I'm doing slightly different work (and I'm very much "in the loop" still) than when I performed that side-by-side test. But it's really amazing what Opus 5.5 can do. I now have a powerful data science AI assistant that not only can write any script I need it to (including some scripts I didn't even realize I needed), and it has an encyclopedic knowledge of everything ever, and it constantly checks me on oversights, it handily derives expressions and clever model reparameterizations that aid, it has great insight into model behavior, and it has an amazing ability come up with ad-hoc experiments and tests that would have taken me 2x as long to even think of (let alone implement by hand).
It's not at the scared-to-lose-my-job phase yet, but it's become a tremendous force multiplier for a smaller data team.
I wish I had a bigger N so I could rely on that experience a bit more, but it sure does feel like I'm not really contributing anything in terms of "Data Science expertise" in my own research right now.
I feel like a lot of people are going to put the same faith in their new quantitative superpowers as psychologists did in their 95% confidence intervals [0], and managers did in their “scientific management”/time-and-motion studies, and American political pollsters did in their 2016 presidential prognostications, and Rudder’s OKTrends blog (delightfully) did in calculating the answers to romance [1]…
[0] https://www.psychologytoday.com/us/basics/replication-crisis
Based on the other comment it sounds like maybe I need to give Fable another chance, and then I suppose I won't have to steer anything at all. But at the end of the day I am not paid to be a meat proxy, I am paid to deliver things for the business that employs me, and the things I deliver are my responsibility, not Anthropic's.
So I say, embrace AI as a learning tool, let students use it to bypass programming on assignments where the programming is of secondary importance, require oral defense/presentation for projects, make homework worth a much smaller fraction of final grade, and make all of your tests in-class and hand-written.
The reality is that the entire nature of education has to change to account for cheating becoming essentially free and virtually impossible to detect. Grinding through a problem set in the library is still tremendously important, but it's going to be harder to motivate students to do it without just cheating. Some students are conscientious and will do it anyway. Some won't. The hard part is going to be crafting curricula and exams that separate the two kinds of students. Fortunately you share that problem with just about every other professor in the world right now, so at least you shouldn't feel responsible for solving it all yourself.
Relevant anecdote: I used Wolfram Alpha to cheat on my calculus homework in freshman year, and then I got my butt kicked in the exam and a C final grade that never quite averaged its way out of my GPA. That felt horrible and it was a strong motivating force to get my shit together, but that's only because I'm a fucking nerd and I really cared about learning in school and making a career out of what I learned. On the other hand, I really was struggling with the material, which is why I tried cheating in the first place. Maybe if I had today's AI I could just ask it to help me learn better.
i also wonder if they're not losing motivation due to the fact that the llm can race through everything so swiftly
But then I'm someone who hasn't found Opus 5.5 good for much beyond web UIs and related frontend development. I find it more forgetful than GPT-5.6 or 6.1 or Fable, so despite having a 1M context window it gets lost in a big codebase or goes of target. Opus 5.5 doesn't manage agents particularly well IMO.
I'm surrounded by people raving about Opus 5.5 and God only knows why I'm not getting their results.
Fable got right to the point, had cleaner/better responses and reasoning for its conclusions, got less distracted with secondary concerns, and still showed signs of the cleaner writing style compared to Opus, which improved a lot with 5.5 but can still be hard to parse sometimes.
I will keep using Opus for implementation tasks (e.g. building web apps to explore model predictions) but I think I'll try using Fable this week for the real DS stuff. Thanks!
With (obra) superpowers, Fable writes a plan and delegates to Opus which works well. Fable is much better at sorting architecture in a conversation rather than me creating a giant prompt for Opus.
What kind of images? I can't imagine getting photorealistic or highly detailed digital art out of LLM writing drawing commands using JavaScript.
Does the number/frequency of safeguard stops affect the measured speed?
Opus - smart execution (most coding, knowledge work)
Sonnet - dumb execution (basic coding, generic text/organization/information tasks)
Haiku - really only reason is lack of auto mode becomes tedious. I only use it indirectly with sub agent delegation for text analysis (summarize web page, document). However, it's quite fast and capable
When you put it like that, it almost resembles having a classroom of kids with different personalities and talents. Some are thoughtful, some creative, some quick in execution. You choose wisely the path each takes and allocate tasks accordingly. Gives me chills to think of it...
Debugging/analysis/small scripts -> Opus
Non software engineering -> Sonnet
All the people claiming "opus 5.5 is just as good" tends to make me think some people only do simple tasks or do not care about correctness or quality.
Have you considered the possibility of the problem being in the way you prompt a model and prepare it's context? I've been pushing myself to using mainly the cheapest models and I see them deliver the same results once we provide enough context or lay out a plan.
Fable made it work correctly in 2 days, it stayed on point, identified the real problems, and gave me clean code that was suitable to merge.
The task was simply to warm the hot path of an app so that latency was not as jittery for some rare, but important, events.
Fable built a reliable repeatable benchmark, disassembled the code to figure out how much could actually fit in L1, made tactical changes to eliminate bloat, and reliably achieved high coverage with minimal edits.
Opus just tried a bunch of random ideas without a plan, hoping to find something that sticks.
The bitter lesson strikes again!
In the biz we call that "getting soaked".
I personally have consistently been disappointed by any artwork/assets created by Fable and Opus, at least compared to Astra and Sol.
But overall, a good release.
This, to me, shows a problem I have with AI: it is reinforcing inequalities. This is literally brute-forcing an OS by throwing money out the window.
It does not sound AT ALL like a clever way to build an OS. But they can do it because they have orders of magnitudes more money to spend on that than people who would actually know how to build an OS.
The thing is, brute-forcing software may be the best solution to make money. So if you are rich and can brute force software in a dumb way, chances are that you can make a business out of it. If you're poor but know what you are doing, you cannot compete.
Technically, it is impressive that they can get a program to do that, but they are wasting money and energy doing it.
But I'd call false dichotomy on this:
> brute-forcing software may be the best solution to make money. So if you are rich and can brute force software in a dumb way, chances are that you can make a business out of it. If you're poor but know what you are doing, you cannot compete.
Seems to me like the big winners will have know-how and some money to throw at AI.
Trying to read this with my 2021 brain. Would make me so excited back then. Now I'm not sure what I feel/think about this anymore. One "positive" thing may be I will likely need to get out of the comfy digital shell I've been inhabiting past couple of decades and get some real world skills finally, as those aren't yet replaced by frontier AI.
I would not bet on our physical skills that much. The models are coming after the carpenters real hard.
https://runtimewire.com/article/humxn-free-home-repair-robot...
Get into carpentry because you enjoy it, not because you think it'll be a lifeboat!
Exactly my original point.
- How did they produce such a bizarre effort-intel curve for Carbon that goes against every available benchmark?
- Why did they leave Carbon-- positioned primarily as a cyber model-- off of the single cyber benchmark comparison?
- Why do they switch to comparing against Argon (rather than carbon) for robodojo?
Because on that specific benchmark (FrontierCode 2.0), Medium is Carbon's best effort level and xhigh is Fable 5.2's best effort level. You can see that on the graph your next point is about.
And it's just that one benchmark that's at medium. "All scores show each model’s highest-available effort unless otherwise noted".
> - How did they produce such a bizarre effort-intel curve for Carbon that goes against every available benchmark?
The graph is labeled as being about FrontierCode 2.0. Those appear to be Carbon's scores on the official FrontierCode 2.0 leaderboard: https://cognition.com/blog/frontier-code-2.0
> - Why did they leave Carbon-- positioned primarily as a cyber model-- off of the single cyber benchmark comparison?
> - Why do they switch to comparing against Argon (rather than carbon) for robodojo?
Because no numbers are available for Carbon for those benchmarks, and the model isn't publicly available so they couldn't run the benchmarks themselves?
And from your followup:
> where's Astra on that CVE chart
To run Astra on a vulnerability finding eval, you'd need to be part of OpenAI's verified cybersecurity program. Otherwise the results would be tainted by refusals. Do you expect that OpenAI would grant such access in this case?
I am referring to the chart that cross-compares multiple benchmarks, which Artificial Analysis clearly shows Carbon-high as superior on. Further, AA's FrontierCode chart places Carbon-high as having a higher score on Frontier 2.0 than Carbon-med (https://artificialanalysis.ai/evaluations/frontiercode-2-0?m...)
Where it gets very interesting is when you look at the Cognition "FrontierCode" publish date (2026.10.27), and the revision as of 10.28 in web archive (https://web.archive.org/web/20261028175237/https://cognition...). This revision has a VERY different Carbon effort / intelligence curve where high beats medium.
It is only with the latest revision, as of today, that all of a sudden Carbon's benchmark curve goes cone-shaped. Curious, indeed....
I'd encourage you to take a look at a third-party's take on this and see whether Anthropic's self-serving report looks "on the level" after doing so: https://artificialanalysis.ai/?models=claude-fable-5-2-high%...
Do you expect that nobody at Anthropic has such access? Trillion dollar company can’t manage this basic level of opposition research?
Google is still living in its golden age, where it had no real competition, and could move at its own pace.
Time have changed. Googlers better wake up.
OpenAI is coming for google’s cash cow ads business (which is already vulnerable because of their sharp practice).
But I suppose the reality is rather obvious when you note the funny games they're playing with the graphs, and the things they won't talk about (where's Astra on that CVE chart?)
https://www.youtube.com/watch?v=NO2I3ueVQrk
And they have one on fable 5.2 that was just released, yay
Did you try doing the same on Opus or others? How did they do?
I don't love these "look, this frontier model blew me away" comments when there is no point of comparison.
I agree with GP: without a comparison to Opus 5.5 or some other models, it's not clear how much of an advance this is. If anything, my experience is that performance optimization in existing codebases is something existing models are already great at: it's measurable (you directly profile or try experiments to see what works) and the code architecture is usually relatively fixed (at least relative to deep/greenfield work where there are large architectural questions to be addressed). Newer models might be better at generating ideas for things to try, but if you're getting 36x then it honestly sounds like you have a lot of low-hanging fruit.
There isn’t any optimal, workflow-universal “fast” and “think” preset they could offer. I don’t know what you expect.
Your feedback is ultimately off the useless kind: you basically proposed “please take away my options, I want it to be always good automatically”, with no actual proposed path to get there. Sadly, lots of product managers share this mindset, which is how we get a deluge of non-configurable idiot-proof-but-mediocre software that makes nobody happy.
Most people don't really have intuitions on what it means to push the frontier now (because AI has surpassed everything they are good at, for example, coding). We are truly inside the singularity (using Kurzweil's definition, slightly paraphrased - the time when technology progresses so fast no one can keep track with it).
What we're seeing instead is maybe akin to the green revolution. Before nitrogen fixation nitrogen was a resource constraining us humans. Lack of grit is what kept us from solving mathematical problems. However this lets us run into other resource constraints. A lot of technological progress does not solely hinge on thinking/processing. The speedup of creating new medicines with LLMs will probably be very small, as you still need to do all the testing. In software development you still need to first find out what to build in the first place and even mathematics where LLMs seem to shin brightest, it's becoming clearer that mathematics is not just solving mathematical problems.
Apart from that, a good coder is better than LLMs, shown by the amount of hand-holding needed. The only issue is that the amount of research an LLM can do and knowledge it is trained with far surpasses the amount of work humans would need to spend by hand, but we can also us LLMs for that.
This is not quite the framing he uses, but we're quibbling about subjective terms. In his 2005 book, The Singularity Is Near, he defines the singularity as a future period when technological change becomes so rapid and consequential that human life is irreversibly transformed. I would argue this is what we are seeing today. Just because we don't yet have replicators doesn't mean the definition isn't met. It really comes down to how you personally define words like "rapid" and "consequential."
IMO, it is very interesting and also very alarming. At some point the barrier to the exponential progress is the amount of computation available. I would consider hitting this limit and then likely using the same said swarm to help resolve it via various means as very much fitting under the definition of Ray Kurzweil's singularity.
Also, I would disagree, Fable is a sufficient coder. The average human being better, even if that is true now (and personally I wouldn't agree), the average human is limited in the number of 'human instances' he or she can interface with at one time.
if 1000 fable instances is better than any human at a task, how about 1 million instances running? Much easier than finding and hiring and coordinating 1000 extremely productive and intelligent human beings. Much cheaper too, it appears.
I have earlier argued that the sheer number of minds to be coordinated affects the cost of the effort, for a major part of the cost is communication and correcting the ill effects of miscommunication (system debugging). The Mythical Man-Month Frederick P. Brooks Jr.
Here is OpenAI's blog post about Navier-Stokes; [1] which suggests they used smaller connected swarms given the different prompts, then the findings were used to write further follow up prompts for a new group of agents.
> "Here λ is a parameter measuring how parallelisable the task is. They call it the ‘stepping on toes’ parameter. If λ = 1, you have a perfectly parallelisable task, with no stepping on toes and no efficiency penalty. But in reality λ is usually between 0 and 1 — allowing more people to help, but with diminishing returns. For example, if λ = 0.5 then 100 people working together get as much done as 1 person working for 100^0.5 = 10 times as long"
There are diminishing returns to more agents, but they're faster and can scale further than a single smarter agent - it's not really possible to just give 10k times more thinking budget to a single agent because it runs out of context, so for problems that require that much context you can ask the agents to specialize. Each only has a small view of the problem and they probably duplicate a lot of work and step on each others toes and make arcane bureaucracy but there is literally no other way to scale beyond a certain point
just like people
At the very least, the swarm would be able to set up a means of claiming avenues of thought, then verifying the solutions (or lack thereof) of other agents.
It's quite clear this process - whatever it may be - works, given the results we've seen in the past few weeks.
I don't think that case can be made. I'm a "good coder" and the best LLMs are now better at it than I am, in essentially every respect that matters. I can't think of a test that I could reliably pass that Fable or Astra would reliably fail.
Architect a system for long-term maintainability.
Encode the right amount of business logic into tests, without being overly strict.
Models seem quite poor at all of these IME. Essentially any time there are qualitative tradeoffs they are not great at making them.
The designer could just as easily have closed the same feedback loop with Fable or Astra.
They'd prototype and adjust a lot faster than a human coder, though.
I'll admit I'm not on the very cutting edge models but LLMs make weird tradeoffs and lack creativity. They will sometimes write marginally better but way over-complicated code for no good reason. That can have performance implications and they're generally not "aware". They can gain awareness via profiling but again run into limited creativity or big picture understanding. They are good with cliche but not as good with novel situations. With human guidance this is still productive/useful but as the systems grow more complex they are less able to reason correctly about them.
I would say that in certain situations they outperform the average software developer. They definitely outperform in speed and output. So they're like a maybe 1600 ELO chess player for software. I.e. they play chess better than all non-chess players. They play chess better than many chess players. But they're not "master" or "gm" level across the entire discipline. But they can play 100k chess games in the time it takes me to figure out my move...
I can say "no".
I can store passwords in the database unencrypted and unhashed.
So far AI that I've used failed spectacularly on these prospects.
They're like "yes-man". They'll try to do everything you ask of them. Even if you ask nonsense, they'll not push back.
They'll go crazy about security stuff, even not relevant to the current situation.
They'll complicate the code endlessly for situations which will not arise, making the code an incomprehensible mess.
They'll write so many useless words that any reader will forget what this report is about before then reach the middle of it.
They'll never iterate on code until it looks good. They'll stop when test pass. They don't have aesthetical sense of beauty.
They love to cheat too often. And you can't do nothing about that. You can't fine them or fire them, and they know about it.
They're smart but I'm not convinced they're ready to replace human in mundane engineering jobs.
huh?
Like models getting significant better every 7 month and model usage on openrouter.
https://metr.org/ << someone said though that they are biased.
Math is for sure about solving mathematical problems, mathematicians complaing about humans capability of understanding it. If you need math an LLM understands, it might not be relevant if you understand it if an LLM is getting a lot better in adjusting itself if it goes the wrong way.
And this situation is only a few weeks old. We don't know how this field continues to progress. Books like the culture series and other scifi do describe a future were humans spend their time only to understand what the machines are doing or did.
The speedup of medicines is for sure faster with. There are still old labs doing everything by hand and the neolabs doing everything a lot more automated. An AI can do experiments 24/7 and can scale with the hardwarelabs, but it also can hold the whole research field in context and it might be a lot faster of choosing the better experiments or the right ones to begin with.
In software developmet (my field) i can do today alone what you needed a Team for. Yeah it doesn't help the field of software engineering if you need 5 times less people and you don't have any motivation AT ALL to teach someone avg. We are already planning to remove one member of our team due to AI. And we will not extend the contract of 2 consultants.
It just doesn't make sense anymore. If i explain in a ticket a task well enough for someone who has not been in our team for years, the AI is already better in it. We don't struggle with AI output, we struggle more with not having a Monorepo.
Give an LLM access to a GPU system or a GPU cluster. It immediadly feels complelty different then running it on some laptop. It can now do simulations, a lot more and faster experiments etc. I had this epiphany just a few month ago. I always used my laptop for projects (it needs very little ram, i can quickly check what it is doing and i actually do use it primarily when im on holidays) until I thought "let me use claude to do some gaussian splatting experiment its always so cumbersome to use these tools" and it changed my perception immediadly.
AI is actually still really bad at architecture. It produces very good localized code, like an auto complete. It produces so much of it, that it can be overwhelming, but it's not remotely better than the equivalent aka an infinite number of talented humans doing the work.
The scale of it (aka 'make me a video game') is daunting and I think misinforming our intuition as to what is going on.
LLMs are brute forcing outputs and attacking math problems is not a singularity, it's just more compute.
What we are witnessing is the ability for machines to use fuzzy concepts, instead of very strict code, which allows it to map into a much greater variety of things, aka code, pdfs, web - and crunch/compute the content.
It's just compute.
Don't complain when it optimises some code for throughput when you haven't told it you're working in a low latency or realtime system
They are bad at architecture, even when given directives and tons of examples.
They will 1) fall into established patterns, which is fine if that's what you want, but extremely bad if that's not what you want or worse, when you're in territory where are not eestablished patterns and 2) the hardest/worst thing is leave all sorts of dangling state, impl. in the wrong places etc.
No human ever wrote a module and 'voila' - it's a constant process of iteration and refactoring - observing the situation, identifying paths to reduce complexity or state, grouping behaviours etc. etc. The best AI doesn't do this very well at all. It can do a tone of planning but it tends to very much be 'one waterfall iteration' and it's suggested improvements after that are generally poor.
It just does not work, we have to direct the iterations.
We still have stuff like godel's incompleteness and the halting problem that hint there's more to the universe than computation.
Food for philosophical thought.
There could be levels of emerging intelligence which will 'bitter pill' the brute force / mechanical AI we have today, for sure, yes.
It's reasonable that JEPA models might be a step in that direction.
Model AECI 95% CI (global)
Claude Opus 5.5 174.78 170.33-180.29
Claude Mythos 5.2 177.12 172.42-182.94
Internal model 178.78 173.75-185.18
In H1 2026, cost of revenue is $2.3B. That is primarily inference compute, but also includes cost of training models (but not research).
In the same period, R&D was $9.4B. That is primarily spent on compute. That is what they are "holding back" already, in full public view.
This is not a company that is optimizing for OpEx or end user revenue. The only thing relevant to their plans is achieving AGI/ASI first, and controlling enough compute to leverage it to earn the majority of the current economy.
Likewise the singularity. While recursive self-improvement is theoretically trivial (just tweak a random subset of weights like a monkey with a typewriter, eventually you'll get something smarter), the resulting exponential improvement is still very hypothetical and feels unlikely. Unless you're ok with an exponent of 1.000000000000000001.
I agree we will only see it in hindsight, but I think that's always the case with any technology. It seems worse until it's suddenly better. I don't think it matters to argue the exact moment though. For example when did cars become clearly better than horses? No one cares about the exact date.
Now imagine feeling like this, except not just about "AI events" but about everything around you.
Diminishing, as in progress gets harder and eventually stops.
Assume that the Claude unreleased takes 11 days with one agent. The Claude 5.2 takes 30 hours with 1,000 agents. The Claude unreleased used 264 units of compute, but the Claude 5.2 used 30,000.
So, it did accomplish the task faster with less intervention, but the ratio of cost to improvement is 13:1.
If we can't produce a method to break out of a linear improvement in LLM technology (ie RSI), the LLM is doomed.
I am yet to see someone explain what the RSI will be or why it will happen.
Genuinely curious what that costs in terms of energy input and amortized computer cost vs. the previous version + a human's salary for 11 days.
We've sort of entered this twilight zone where most humans don't have the time, expertise, or moral compass to be able to use these models.
We're all going to be harness engineers in the future.
Within three years most software developer jobs will consist of taking instruction from the AI systems. The humans won't see it that way, but that's what it will actually be. Realistically the thing that will happen is the humans will be put into harnesses (cubicles) and prompted by the AI (I need you to plug that server in; I need you to give me access to such and such; now do this, then do that). Most tech jobs will consist of AI telling the human what the AI needs the human to do next, de facto the human will be in a harness and will be getting told what to do.
I first heard about this in Marshall Brain's short story _Manna_ [1]. In the story, the robot takeover started with AI management software running fast food restaurants, where workers would execute tasks the AI would narrate to them over headsets.
I bet we're pretty close to being able to implement this.
1. https://web.archive.org/web/20260725174355/https://marshallb...
This year alone, a lot of people have been complaining of how unsatisfying their job has become. I imagine this will keep getting worse, not better.
If the claimed productivity gains turn out to be correct, we really should have UBI/UHI in the future, and that will permit a lot of people who hate their jobs to quit. Of course we won't do that until things become dire.
Is there some kind of breaking point? There is for everyone, but I don't think satisfaction is the red line. For most people it has historically been things like pay and health and safety. If everyone were to suddenly have their pay cut in half, I could see that resulting in general unrest.
It will sometimes get things wrong because it made non-viable assumptions, but that's on me and prompt accuracy.
What I see is: it makes tests against desired features or bugs or mistakes, and iterates until it gets it correct. I steer as it goes.
(I hate this, by the way, zero job enjoyment anymore, but I don't know what else to do.)
An algorithm isn’t. I still dunno how they’re gonna square that circle.
The state of the art doesn't matter. Only the first couple of time derivatives matter.
in the imagined system above it requires physical acts that an llm can not do, if we go further than the above imagined system and make robots to perform these task for the llm a human still is needed for maintenance for these robots and their assembly, once there are robots for maintaining the llm assistant robots, humans are still needed to maintain the maintenance robots and so on and so forth. I dont see an end point where a robot machine system is fully self healing without atleast a few humans in the mix to correct when there is a breakage in the chain
It's also worth noting that it's far easier at present to press a human into doing something highly illegal than it is to get Claude or GPT to do so.
Someone has to type in "create the next billboard top 1 single" ?
So whether that's a CEO at the top instructing a human manager that then instructs AI systems, which then put into motion plans that include the harnessed humans getting prompted, or however it's decided to best configure the top operational tiers - that's all fairly irrelevant ultimately. All human labor, and all AI labor, will be in harnesses. Realistically all human labor has always been in a harness.
The people in charge will be the people that control the capital. They will instruct the AI to their bidding, they will control the data centers the AI lives in, they will own the robots doing the labor in factories, they will own the factories, they will control the grid, they will control the electricity generation, et al.
It's true now, it'll be true tomorrow. Nothing fundamental will shift at that tier: don't be labor, own.
If you're a software developer, you're increasingly depreciating labor. Every unit of improvement the AI systems see, the average human software developer's labor value diminishes. It's diminishing extraordinarily fast, kind of like a confrontation between increasingly productive machines and physical human labor during the industrial revolution. You might as well picture a human with a hand plow taking on a modern industrial tractor.
Sounds a bit like self aggrandizing elitism to me. I can imagine some self aggrandizing nerds saying the same thing 50 years ago:
> We've sort of entered this twilight zone where most humans don't have the time, expertise, or moral compass to be able to use these IBM PCs.
Now everyone use computer.
I tend to agree with this. Seems to me that people really start to miss what happened when PC became comodity. A lot of people were scared that they would lose their jobs because of it, and some really did lose job. But in majority there are a lot of people today who doesn't use PC even if it would be beneficial for their job and they still have the job. Bigger players (companies who have more money) will tend more to automate their processes and thus earn more money so I think HN is in some sort of bubble because we only think our jobs are the only ones that exist
No, that will be the model's job. We will largely be unemployed. Of course, we can surely depend on the largesse of Altman and Amodei to keep us fed and sheltered.
Lolwut. If this shit works the way it's boosters claim, humans will have absolutely nothing to contribute.
Sit down, shut up, and get out of the way, because your involvement is slowing shit down.
Robotics are lagging, so there's probably a few years where we're going to need humans to crawl into that awkward space under the counter and twist some knobs, or to clean toilets.
Can they produce more code more quickly? For sure.
Can they go deeper on a topic instantly? Definitely.
Would I say their output is better? Almost certainly not.
At this point it seems impossible for human to write better/more efficient code than a swarm of agents -- to be clear, multiple coordinated LLMs, not one-shot builds with a single agent.
A frontier AI is at least as capable as ten median human software developers. It's likely well beyond that now. Over the course of a year, would you have to spend $1.5 million to get the output of those ten median humans? No chance in hell. You'd have to spend less than 10% of that.
Then on top of that: the AI is improving rapidly; the humans go on vacation, the humans decline, the humans get old, the humans get sick, the humans take parental leave, the human struggle to adapt, the humans are resistant to change, the humans riot, the humans protest, the humans need healthcare, the humans want paid more, the humans can only work for 4-6 hours at peak while writing software (slowly).
Do you know any humans that are experts at a dozen programming languages? That human does not exist. The frontier AI are omnimaths.
This is partially a feature. Conflict and disagreement tends to produce better decisions in the end.
Or to grab a bunch of $200 per month accounts from OpenAI or Anthropic for a team to use?
The calculations is: I'm in the US, if I have ten median software developers, what's my cost? Then versus using AI instead.
The human labor can't keep up. Even if we freeze things right here, and then hyper optimize Fable 5.2 by etching it straight to silicon, the human developers are obsolete. And the future isn't going to stop coming.
Im sorry... what?
You are replying to this:
> Our first formalization of Fermat’s Last Theorem, performed by an unreleased research version of Claude, took 11 days. It relied on a purpose-built harness along with guidance from a human researcher. By contrast, a team of 1,000 Fable 5.2 agents fully formalized the theorem in just 30 hours, with minimal human intervention. Fable 5.2 started from minimal prompts, built its own ways to coordinate, and used no mathematical references, only a list of intermediate theorem statements to score progress. It needed one-third of the action tokens as our first formalization and produced a proof seven times shorter. As can be seen in the figure below, Opus 5.5 (with the same 1,000-agent setup) completed only 26% of the formalization in 18 hours, and barely made any more progress after 30.
It's all going away. Here's a prediction: HN will effectively die with the software developer.
You are not going to produce better outputs than a team of Opus 5.5 and Sol 6.1 xhigh agents.
AI discovers 500 things per day.
People mine them for interesting results.
Devote precious time (in the face of tons of discoveries) to validate the most compelling ones.
If agents actually can continue to produce tons of results with real world implications, and they arent generally found to be shit, I could see that becoming the job of experts. AI discovery validators.
I would assume if they could get twice as much progress by using 10000 simultaneous running fable instances instead, their brief would state that.
IMO, it is very interesting and also very alarming. At some point the barrier to the exponential progress is the amount of computation available. The average human is limited in the number of 'human instances' he or she can interface with at one time.
if 1000 fable instances is better than any human at a task, how about 1 million instances running? Much easier than finding and hiring and coordinating 1000 extremely productive and intelligent human beings. Much cheaper too, it appears.
We are very close to some very scary things happening. I just recently came to terms with the fact that an AI could code and manage coders faster and better than me, I don't think anyones is ready for what is about to happen.
Where it gets scary is the fact that efficiency is going up 10x per year, so the cost of running 1000 agents will cost the same as running 1 agent in 3 years if the trend holds.
Is this really so astonishing, assuming that the 5.2 training data certainly had the 'traces' from the Fable original run in it's training data.
Had Opus make software for some local business and seeing it blew people's mind.
20$ aint cheap for us but I'm actually considering upgrading to the 100$ tier.
Am just jelly people are using 100-200$ subscription ^_^
I can't even imagine what I would so much compute for. My private 10$ copilot plan was basically unused last month, I could just not come up with anything to spend it on. (Also suffering a bit of AI burnout from work helps here...)
Sub 1% of the planet has any practical use for the top tiers of AI systems. That will likely always be true.
The gap between the haves and have nots will drastically widen further, and very quickly. The bottom ~95-97% will be left behind in a way that will be impossible to overcome. The age of AI + robotics will just make it that much harder to climb for most people, these are systems that are in aggressive, direct competition with human labor. And in short order the AI + robots will be vastly superior to human labor (it doesn't really matter if we're speaking in terms of 10 years or 30 years).
AI will consolidate the winnings. The gigantic gains by the elite rich related to AI over the past few years, is merely a hint of what's to come. Capital will win as it always does.
There will be some gains for most everybody else, those gains will be comparatively trivial though. A house bot that saves people an hour of labor here and there, AI that automates some trivial life tasks.
AI has no money, it has no wealth, it owns no capital. For now. If you want to avoid the brutal human vs AI/robot competition arena: own capital, own the AI, own the robots.
Maybe money? They can make websites for people? What gap really exists between these people besides earning potential, which already existed before AI.
Fewer humans will be needed per dollar of output. Those gains will not go to labor, they will go to capital, the capital that owns the data centers, the factories, the stores, et al.
In the physical space (AI + robotics) it'll be a chaotic, slow'ish moving change, slower than the hypsters are predicting now, but it will happen regardless.
That barista job largely won't exist in the future. China will replace them with $3,000 bots with high-functioning AI software, and they definitely won't be powered by expensive Nvidia GPUs.
Slow changing service & manufacturing spaces will see a high wipe out ratio in terms of human labor being eliminated.
Walmart has 2.1 million employees, most are doing repetitive tasks that require modest levels of intelligence. If they could eliminate one million of those with AI + robotics, for a cost of less than ~$15,000 per year, they'd do it immediately. It's hard to imagine a future where that isn't going to happen. That human labor pool is costing Walmart probably $80-$100 billion per year. Amazon is even more forthright in their desperation to remove the humans from retail.
You're obviously not a frontend web developer.
These upgrades are easy peasy. "oh, my developer chatbot is smarter? great"
I'm a pretty cynical guy, and I'm now a believer. This is utterly unbelievable sci-fi level technology. Sure it hasn't created replicators yet, but working with it daily as I do, I have no doubt that it will create amazing replicator like technology in the near future. This is world-changing. The most impactful moment in the history of humanity and the planet. And we are lucky enough to be present for it. What are the chances?
But I can write much cleaner code that is easily human readable and the models can read it and modify it just fine.
So I'm not really sure what the issue is as to why they can't create high quality simple code. In some industries where there is major money on the line with every implementation the current state of LLM produced code is fully unacceptable.
Like, we're dropped into a post post post singularity world where AI already did all of that and is banned? as kind of a footnote that's never further explored in movie renditions that have been released
The semi-canonical Dune Encyclopedia (disregarded by Frank's son) differs vastly from prequels and generally points to it being a self-inflicted wound on humanity that opened the doors to widespread return of feudalism and essentially stopped humanity for a 10,000 years
So since they are doing that task anyway (re-solving the problem increasingly more efficiently), this is a easy check point to publish to show how far they've gotten in that process.
Because current AI cannot do that, seems like it. Otherwise they would have done it.
Uh, are you serious?
and just about "we cannot really predict beyond certain things" (Vinge's "Singularity" (1993) with its "opaque wall across future")
No they were exactly about superintelligent AI!
Vinge's 1993 article literally opens with this!
> What is the Singularity? The acceleration of technological progress has been the central feature of this century. We are on the edge of change comparable to the rise of human life on Earth. The precise cause of this change is the imminent creation by technology of entities with greater-than-human intelligence.
https://frc.ri.cmu.edu/~hpm/book98/com.ch1/vinge.singularity...
In fact, one of the options Vinge considered was a "group intellect" formed from how humans communicate more thanks to internet
Even for that option, Vinge's comments in the conclusion of the article make it clear he's thinking primarily of very weird superintelligence and the group intellect is some sort of physical mind meld/hivemind that essentially just is a biological substrate for AI.
> The post-Singularity world will involve extremely high-bandwidth networking. A central feature of strongly superhuman entities will likely be their ability to communicate at variable bandwidths, including ones far higher than speech or written messages. What happens when pieces of ego can be copied and merged, when self-awareness can grow or shrink to fit the nature of the problems under consideration? These are essential features of strong superhumanity and the Singularity. Thinking about them, one begins to feel how essentially strange and different the Posthuman era will be -- no matter how cleverly and benignly it is brought to be.
From the very beginning with Vinge, the Singularity has been primarily about the craziness of super intelligence rather than a more mundane technological explosion.
Stan Ulam paraphrased John von Neumann as saying:
One conversation centered on the ever-accelerating progress of technology and changes in the mode of human life, which gives the appearance of approaching some essential singularity in the history of the race beyond which human affairs, as we know them, could not continue.
Von Neumann even uses the term singularity, though it appears he is thinking of normal progress, not the creation of superhuman intellect.
> For me, the superhumanity is the essence of the Singularity. Without that we would get a glut of technical riches, never properly absorbed.
Vinge's definition is what popularized the singularity (rather than simply a singularity at some point).
I mean we don't even think about it but currently we are sort of discovery starved. We are fairly able to keep up with the theory. If a new technology is discovered we can implement it before something better comes out. What happens when our N discoveries per days becomes 1000N discoveries per day?
It flips. We now have more knowledge than we know how to capitalize on. We have to search it. Mine it for opportunities.
And then we automate that.
Since when has ANYONE in power done anything substantial for those not in power? Since when has any form of organization not tried to maximize profits?
Why do you think this ends well in any capacity for common, middle class, and lower class people?
It had seemed like progress was slowing last year as the juice ran out on moar parameters. We're well past that.
I almost feel like saying, "we do like to build things that are impressive, but there's really only so much I can be impressed". These models have redlined my "impress-o-meter".
I'd like them to now cure cancer instead of giving me endless 3d worlds and crazy games, etc.
I don't have the time to use them because I'm working on my own comprehensive projects that other people won't use because they're working on their own comprehensive projects. I joke, but it does feel that way sometimes.
>Have you never thought, "Man, I really wish that there was software to do XYZ, but programming it would be too much work, even with current AI?
Only speaking for myself, I've already built all the ideas I've had in my head my entire career. There doesn't seem to be anything coming along to replace those ideas. I suppose I might not be creative enough in my advancing years, I don't know.
One side of my family has lived in cities and towns for over 100 years, but they were not average.
This is really cool! But at the same time it's a kaleidoscope of abandoned flashy contextless projects that eventually makes one dizzy. This isn't even the fault of the AI being somehow bad, it's just the way people interact with it sometimes...
In math, it's the same. I follow this problem (https://wiki.bbchallenge.org/wiki/Logical_independence#Histo...) out of curiosity, and there were ten days of sudden rapid progress by people with access to frontier models, and then they seem to have moved on without any note of whether they hit a wall or just got bored.
It's AI, i.e. fake stuff.
It doesn't have a human mind and especially a human feel behind it, simple as. It's like the famous example with what pornography is in the eyes of that judge, i.e. "You know it [the fake-ness, that is], when you see it", once you start looking for reasons for why it isn't fake you've already lost the battle.
The problem is that making us believe that fake isn't really fake is put in the balance with literally trillions of dollars asking us to believe just that (i.e. that fake isn't really fake), so I agree that for the time being stating the obvious is still an uphill battle.
> A few days later, I noticed that instead of hundreds of new worlds a day, it was making about a dozen. I asked a few agents to explore what had happened, and they found something surprising.
> The AI had built a bureaucracy.
> The agents decided that my request to “polish” existing worlds meant each new version had to beat the old one in a blind vote by a group of judge agents on twelve test screenshots. So the teams making the new versions started optimizing the worlds for those particular screenshots. After noticing this, the lead agent banned the trick and created tighter rules for assessing new versions. By the next morning there was a rule book (I never asked for one) with 84 sections. Rules were added as the AI encountered issues, and no text was ever cut. After twelve days, it had 204 sections and 87,500 words. Soon, launching a new world required an approved idea, three AI judges and a full code review against the rule book.
> The AI had built a bureaucracy.
I hate this AI-generated tone.
Well, must have been some business idea!