We can use a non-deterministic system because we have test suites that act as validation gates.
Other industries don't, so they have no way of knowing if the AI is bullshitting them outside of manual validation, which makes it more expensive and slower than doing it yourself.
It's a good point but in most of the places i've worked an average person could leverage any 2026 llm to improve things 50% easily. And I'm more anti-llm than not somehow.
- fixing regular computers issues (confusing browser or os behavior)
- finding more possibilities with your current programs (around 2020 someone once asked me if I knew a way to capitalize a word on a computer, when she realized there was a dedicated shortcut she became ecstatic)
- having a rubberduck to discuss your daily work dead-ends
- thus decoupling from team friction as you need to ask less questions to people that might hurt you.
In more technical places, having a llm write small automation would also add up.
As a supporting example for the first point: I recently had a dozen ore more colleagues need to be hand-held through getting basic pieces of their local Python environment working, when the multiple agent tools they have access to could have easily solved it for them if they had just asked - but instead they came to me. Also, these were not swe's but still technical people, so I would have assumed they'd be the group who didn't need that help.
I strongly agree that search engine is the best application of LLM so far. 90% of time I don't look at Google search results, I look at AI Mode output only. And you don't need agents, harness, guardrails, etc for it — basic small clean LLM does the job well without overdoing it.
There are many tasks humans perform which have no or slow feedback loop. Even if they can audit/validate AI output, they don't know if it will be good when it reaches other humans.
Your information seems to be outdated, unless every single street I’ve been on for the past year is “vetted”. All the other (consumer) car brand’s self-driving systems do have that limitation though, maybe that's where the confusion came from. Tesla is self-driving 98% of my mileage, with the 2% being when I take over to choose a specific parking spot or I’m towing/carrying something big.
Don’t know much about how it does in the rain since moving to SoCal but in clear weather it hasn’t come close to hitting anyone yet. Even proactively warned me a few times while I was driving. Maybe I should let it take over in parking lots too :)
It'll be hard, when I'm never causing any accidents or speeding or running reds/stop signs. Maybe they could get me for inattentive driving but that's what the tinted windows are for.
In most cases it’s only the object recognition systems that use machine learning. I know that Tesla was talking about using non-deterministic AI for the decision making as well, but it’s a worry if they’re already doing it.
Earlier Tesla self-driving software used search-based planning and conventional optimization alongside neural networks. Tesla has publicly demonstrated A* in its path planning work.
For those who won't act ethically or at least reasonably, regulation is the answer. We're not too far away from flipping the congress into something far more willing to do just that, especially with anything connected to Elon Musk.
IRS has been starved of resources following the "starve the beast" playbook.
Also the only reason we have to pay for Turbotax (etc) is due to lobbying. We had IRS Direct File for two years and it was great before being gutted by the current administration.
Games have already begun to remove Linux support, most notably Apex Legends and Wardogs. Embark (The Finals, Arc Raiders) recently gave a talk that userspace anti-cheat is dead and they plan to enforce similar requirements.
And it's all a shame, in my opinion. They're too lazy to develop effective behavior-based anticheat (should be easier than ever with better NN architectures and more compute), and so they resolve to screwing over consumers or spying on them. Did not know about Embark's stance, that's too bad.
How is Valve’s VACNET doing? I think you should take a look at what’s happening in Counter-Strike 2 these days to see what the true frontier of ML based anticheat looks like.
Frankly at this point I’m pretty sure John McDonald plays HvH too
I play CS2 plenty. The cheating problem is overstated as all hell. It seems to be concentrated around accounts that are being sold for lots of cash, so Valve is apparently taking that route for banning accounts more often lately.
Now, if we want an interesting case study in intrusive anticheat not necessarily being more effective than behavior-based, we should compare FACEIT to Valve servers. FACEIT is riddled with cheaters, and yet requires ring-0 anticheat on Windows. So fascinating!
What ranks are you playing on? While a player of any rank can buy Neverlose and win and if they wanna just pwn cleans they can buy Midnight (speaking of buying P2Cs: if Faceit is apparently riddled with cheaters, where can I buy a P2C that I can pwn on Faceit with for an affordable price?), said cyber athletes will naturally gravitate towards higher ranks.
> FACEIT is riddled with cheaters
According to sources of… what? In any case it’s sure as shit less riddled with cheaters than Premier LMAO
Yes. Trump is not a rational actor, it is textbook narcissism, not as a critique of his personality but I think he meets the criteria for clinical diagnosis.
If you don't listen to him then it results in narcissistic injury and he will retaliate.
I do think a lot of startups, especially on HN, are in for a reckoning that your new vibecoded product/service or one with an AI feature or two added will survive. Everything you do is going to feed the model, that's the training data now. That's why Claude is so cheap for what it does.
You cannot prompt Claude better than your customer can. There is no realistic moat there. You will not survive.
> You cannot prompt Claude better than your customer can.
I actually think you still can. We're still in the "centaur chess" phase where a knowledgeable human sometimes needs to know the right things to ask for.
Will you be able to in 6 months? 2 years? I don't know, but it's not looking great.
> Will you be able to in 6 months? 2 years? I don't know, but it's not looking great.
I don't think the person you are replying to meant everyone won't survive as of literally today.
We're still in a somewhat calm period where something fundamental changed in the world but it changed so fast most people have barely reacted to it as of yet so things feel relatively normal, but with a growing sense that they aren't.
In the beginning of 2025 if anyone suggested "You cannot prompt Claude better than your customer can. There is no realistic moat there. You will not survive" most people on HN would have laughed at it, even the people who were considered AI-pilled for the time.
Today that sort of claim may spark some debate and some people are clinging to denial, but nobody who actually uses this stuff is laughing it off anymore. Even the staunchest deniers recognize there is enough to be concerned about there that it isn't a joke anymore.
Tbh I'm not very concerned. Unlike chess, which has very finite state, the "build something appealing" kind of game has a very vast amount of state. And the current architecture of LLMs doesn't make them very "creative" (whatever you define it as). If given a scoped task, they'll relentlessly try to complete it but the design, as we speak, is still your job. You can get AI to clone something already existing, which does decrease moats, yes. Making it do something new is way harder.
All those one-shot games and product demos are clones of existing things, and while they're very impressive, they aren't exactly new ideas, you know?
Of course all this might change and the labs might figure out the secret to creativity, but as things stand today, humans can keep "moving up" the value stack to just make more complex things.
Right, but iteration has become cheaper. If one gives Claude a shitty prompt and it does a shitty job, one will know soon. Then they can refine the prompt and get to where they want to be.
When iteration had a real costs, you couldn't afford too many mistakes.
I'm scared to find out what systems built using that methodology will be like.
Product design is mostly saying no to ideas. And occasionally yes, but only to a very select cohesive set.
Most people are terrible at that - though I think it's learnable to some extend. And at least current-day agents will generally just do whatever dumb shit the human says it wants.
I expect you’ll find the outputs of those processes to be entirely unconscionable and, baffling, the users will not see a problem with it. Eventually you will conclude that it’s not a lack of information or a resignation to their fate; it is much like the pigs, who just like playing in mud. (This is how I’ve felt twice now, first about AI writing, and then about AI code.)
> You cannot prompt Claude better than your customer can
That is simply and obviously not true. Technical expertise is still highly relevant when prompting, reviewing, understanding consequences, and approving.
When it comes to trades it's also a bit of "most people do not want to do this job because it's unglamorous"
Many people would choose not to be a plumber for various reasons. A shocking number of people seem to deem manual work as beneath them. Many others just prefer to do other work.
I have no problem doing plumbing or building. I've installed floors, ceiling fixtures, drywall, whatever. But I am best at software and would prefer to do that for my career.
I've been thinking a lot about this. Many technical barriers that could justify businesses are much lower now, so we are left with other kinds of barriers: red tape, regulatory, legal, etc. If you have access to go past those barriers, you are worth something. If all you do is provide a technical solution, you are probably going to stop making sense.
You think all companies are going to vibe code and then vibe maintain for years and years all the software they require to do a business that has nothing to do with software? And you think the LLM is going to do it itself?
I fully acknowledge this is a mess right now, but vibe-fixing or vibe-rebuiliding-from-scratch is only going to get better.
More realistically, I think other companies will vibe code and sell these solutions, but the cheapness of building will drive down costs so much that it'll be effectively the same result.
The main reason Claude is "cheap" because it has to compete with OpenAI, Deepseek, GLM, Qwen, Muse (who is also massively subsidised their model to get training data), etc. If they increased prices 10x, people would instantly jump ship to the next competitor. There's no moat for Anthropic either, short of constantly training new models in a red queen's race.
> You cannot prompt Claude better than your customer can.
You may currently have enough technical knowledge about computers and services to collaborate with a customer better than either of you working alone.
Thinking back to a previous discussion, a veterinary clinic likely does know better than you what they need in their CRM, but they may wildly overestimate the hardware required to do that and just accept the first thing Claude comes up with that's scaled for a city hospital with 4000 beds occupied at any given time, when they're actually just one person in a village with 100 pets.
Or the other way around, which is even worse.
Right now, experienced engineers can help with this kind of problem. I don't know how long this will remain true, but right now it seems to be. I think.
For most yes, but the art of picking the right features that interact in the right way is still a delicate craft. AI is like a sledgehammer. It can demolish technical challenges and it can follow patterns but is still completely blind to the "why".
But yeah the majority of startups don't have well thought out reasoning either. In the end, there's no substitute for hard work. AI can reduce the iteration time maybe.
> the art of picking the right features that interact in the right way is still a delicate craft.
It is.
But after you expend the human-in-the-loop effort to build the right software with the right features using your impeccable judgement, what is going to stop your would-be customer from telling an LLM to analyze your solution and reproduce it feature-for-feature for a few dollars in tokens instead of paying you?
In a way its worse- nothing is stopping claude from decompiling all the things- and then adding the source code it decompiled to the training data, as if it was just another github to crawl. So if one of your installers is out there- you have provided the perfect example of something working.
That's true, but the installers for most of the really valuable proprietary native software are still locked behind paywalls. (Some have been leaked.)
I suspect we'll increasingly see vibe coding targeting assembly language. If humans aren't going to read or write the code then using higher level languages becomes kind of pointless, except for being able to use the toolchains.
Right now, I can definitely prompt Claude better than my customer. Claude's abilities, when steered, are definitely impressive, but I have never seen it correctly architecture around a problem unless explicitly critiqued.
i've been experiencing this as well. It does however make me think that maybe the code architecture doesnt matter if nobody is going to read the code. Like maybe the focus should be on better tests and metric tools to determine quality of code (i.e. performance) along with other tests for security, and then let the code look like whatever claude uses to match those targets.
But I can probably prompt Claude better than your customers too, so can at least 50 of the people I know (advantage of a tech heavy social circle). Good luck turning your prompt into a business when me and 50 of my friends can replicate your business model in about 10 second. Don't believe? Go ask Jev devs how they feel about Clef.
My business is going just fine. Don't underestimate the business value of handling schlep. As I said, "a prompt" is not sufficient to run my business, even on a technical level.
Yeah and anthropic can't train Claude better than anyone else either so no moat there either. Knowledge is not scarce it just takes work to uncover once uncovered it is not useful if not published. If published it can't be owned.
How does that make sense? The expertise is what allows me to tell the LLM what to do in the first place, inspect its outputs, and iterate, which is still faster than coding by hand.
Godot is the best open source engine because it is dead simple to use, comes bundled with its own IDE and documentation, the editor runs on any platform that Godot runs (including the web) and the performance benefits for something like Bevy only apply to a handful of games that really need it.
Godot gives GDScript simplicity for the vast majority of use cases which is that iteration boon, provides C# for games that need the additional guardrails. For performance critical game scripts you can just jump to C++, Rust, Swift, etc.
But most game scripts are not performance critical. Clair Obscur: Expedition 33 swept the game awards last year as an Unreal Engine game made using almost entirely visual scripting (blueprints).
It still bothers me that game engines come with their own bespoke languages like GDScript. Godot is a bit better than others for having C# support and forks in other languages but really no other language will ever be as integrated with the framework as GDScript, And GDScript is never going to be a general purpose programming language, and that necessarily creates lock-in.
I'm currently messing with some old projects written in Game Maker 5 and let me tell you it would be so much easier if I didn't have to do internet archaeology just to figure out how GML worked 20 years ago.
Because it is much easier to have a custom language, than try to subsetting an existing mainstream language, while having to deal with all the mess of people trying to use random stuff on the engine, that naturally isn't supported.
Example, see all the bad rep .NET and C# get from devs that only know them from Unity, full of legacy stuff and restrictions, instead of using the real product.
That seems like more of a feature than a bug to me. Using C# in Godot means you have all of the Nuget libraries at your disposal, whereas no one is writing libraries in GDScript, because no one is using it outside of Godot.
Meanwhile to extend Godot usefully you have to use C++ and a plugin API that's arguably no less complicated than adapting an existing language as the scripting language would have been.
But that is the thing, you don't have access to all the NuGet libraries, or even CLR/MSIL abilities, it is a hit and miss depending on what platforms are being targeted.
Sure if you only want to target desktop/laptop computers, you're safe.
I searched for something, it told me that according to a YouTube video, the entire point of my search was wrong. I asked it for the source, I watched the video, it never made the claims Gemini hallucinated. I asked again and it claimed it scrubbed the video and found the point it made multiple times. I said those timecodes were wrong and it admitted it couldn't actually parse videos and just guessed.
I don't understand how this isn't considered as active malice. Like purposefully outputting random stuff. Any other sort of computer system would get lot more flak than these are getting.
Throughout my career, I've almost always been close enough to the user that I hear about it quickly when something is wrong. It's a tough problem that so many Google engineers are typically so far removed from end users. Or maybe it's just that a small part of the company has long subsidized the rest of the employees to the point that it doesn't matter how good their work is because they'll get paid anyway.
The engineers are pressured to significantly reduce "dependencies" for projects. Anything that could become risk or create friction is dramatically less appetizing.
Simply because of how many people that *must* agree with your proposal. Getting all the relevant tech leads, some you have never heard of or ever spoken with, to agree on a proposal for your team's project is a nightmare.
So you keep it as simple and agreeable as possible. Given the circumstances, it makes sense as one of the engineers. It's fairly fine advice in general wherever you work, but it just haunts all the work you do at Google in particular. Nothing gets done otherwise.
If you have a dependency that can be dropped from an engineering perspective, that's the route the 9 leads reviewing your design doc will take:
"Let's iterate and start with just the basics (no user testing)", "let's get this working and user test in a later phase", "I think this problem is obvious enough we don't need to consult with users about it."
I worked in Ads Integrity at the time, and for one of my projects I was concerned how it would impact the manual reviewers. Then I learned I couldn't talk with them, only by proxy through another person if we really had to. And that proxy takes time, so...
Very insightful. I've worked in only one company of Google's size, but it was a vastly different industry. I personally never got to speak with a user on a massive internal application my team worked on for years. :)
Right. In a global financial services organization, a strategic analysis app used by thousands of internal users was pretty high touch. SME business analysts would sit with the users to make sure they liked it and had no trouble. Every user generates significant revenue. Worldwide B2C is a different ball game, and Google barely makes money off of each individual user.
> The engineers are pressured to significantly reduce "dependencies" for projects.
Is a major factor of this Google's tendency to kill entire products, so you obviously don't want to make your hot new thing dependent on some other internal thing that might be killed off?
Interestingly there is less opportunity for that to happen.
Most of the internal infrastructure is quite available and stable, and there are clear choices for almost all of your application-level and backend-level needs.
Guys, your argument reads so weird. It outputs PLAUSIBLE text. It doesn't actually have a connection to causally correct information, although plausiblness often coincidents with causally correctness.
When it says it can read videos you don’t trust it.
When it says it can’t read videos you think that’s an accurate introspection on its abilities?
(Rather than a statistically likely continuation of a conversation where one side seems to be reading videos and the other side says the read is inaccurate)
> The scary (or funny) part is people use that for serious questions.
More scary than funny, IMO.
The US military almost took an action that could very plausibly have escalated into a hot war with China because people are already relying too heavily on these systems.
Despite the reporting, nobody in power seems sufficiently freaked out about this.
Gemini is the worst gaslighter. It hallucinated a feature in an open source product. I said no that does not exist. It invented very realistic git commit messages. I said commit does not exist, so it said I was out of sync. I linked the repo and it said it might be a fork. Then it invented a mailing list thread where they supposedly removed the feature, that’s why I couldn’t find it.
Same experience with the interaction I described in my comment... after I finally got the correct answer, I asked where it got the faulty information from. It said it had just simply fabricated it. That's actively worse than just saying "I don't know", for something that's sold to us as an easy way to look up information.
Maybe it's because they are trained on Internet comments, and the most rare thing to find on the Internet is someone admitting they don't know something.
But if they had been trained on comments saying "I don't know", they'd probably act the same as they do now but they'd treat "I don't know" as the answer.
This is such a 2024 take. Model will push back and tell you the knowledge gap they have, but people are just used to ask gooogle leading questions, which port badly to llm as it makes them defend the position instead of research data
The saddest part is when people take their experience with Google's idiotic AI implementation and assume that's how all LLMs work. Frontier-class models will, in fact, generally admit when they don't know something. That includes the one I run at home on my own graphics cards, but it seems that Google just doesn't GAF.
Your point about "predicting the next word" mostly means that your post was very easy to predict.
I still believe this AI push out of nowhere is due to the current Government in power state side. Making everyone question themselves and each other and being uncertain about facts while being inundated with techbro fake news called hallucinations is a recipe for disaster for older populations that don't 'trust but verify' like most technology inclined people. This is all by design and we'll falling for it.
The more simple explanation is that the current government in power is over exposed in their AI investment. The normal conservative mainstream media was already doing a great job of propagandizing older populations.
Yup. I'm most familiar with Jill Lepore and Quinn Slobodian (and others in their respective orbits). They're both historians who've written extensively about Musk and Muskism (et al). Wild stuff.
Apparently the plan is to use AI slop, mediated thru social medias, to defeat the woke mind virus, perpetuated by the Anti-Christ, in order to safe guard humanity's future.
If you don't think the world is rife with criminal conspiracies you weren't paying attention in history class.
There's an enormous difference between outlandish claims about extraterrestrials or turning frog gays and believing there's a category of politicians trying to get wealthy off of their position. It's somewhat reasonable to hypothesize about criminal conspiracies in that 2nd scenario.
It takes either being scared to death to admit what's going on in the world, or an attention span of a guppy, to keep calling people conspiracy theorists these days.
Other industries don't, so they have no way of knowing if the AI is bullshitting them outside of manual validation, which makes it more expensive and slower than doing it yourself.
reply