This seems to be the case with most industries that LLMs are impacting, coding, data input, legal etc.
Whilst it's great to get the time saving from not delegating to a junior (LLMs are much faster at the basic level 1 work) there's going to be this big knowledge gap that you really only gain from actually doing the mundane boring shit.
A lot of this _should_ be automated away, but we really need to think about how the next group of people are going to learn the skills and nuance to applying them. I don't really have an answer to that one other than still giving that work to junior staff, but they will then just utilise an LLM to help complete the job, and rightfully so, I'd have done exactly the same thing.
I think OpenAI tried doing this with some early chatgpt settings with "learning mode" or whatever that is. But an LLM just telling you what to do will never teach someone, there was something about having to go out and hunt for the answer through the sea of crap responses or wrong SO questions that really taught you something. How do you replicate that learning with an LLM and how do you apply that to different industries like Law?
For example, redlining an OEM agreement with my vendor. They pretty much all use the same template, and the objections are always the same. I tried using Claude first, and even my lawyer was impressed that it found things they missed. Then I used Claude to redline that type of contract and the other side said "your lawyer was surprisingly thorough". I had my lawyer do the final check, but saved a bunch of money doing the first three rounds with Claude instead. And my lawyer basically just said "looks good to me" by the time they reviewed it.
Or even writing a simple will or estate document.
It seems like LLMs are great for the kind of generic work you'd usually pawn off to a new associate or paralegal, but not for the important things that partners usually do.
- Quoted phrase in the opening sentence
- Opening quote: “learned”
I literally just ask a frontier model : "Pull the latest log and root cause the issue.". That's it, 10' later it comes back with a correct answer. Sometimes I might have to prompt a second time to steer.
No skills, agent interconnections, graphs, vector search or whatever else.
The progress is very scary.
The "lazy answer box" is a tool, learn to use the tool, understand its limitations, as act accordingly. If you prefer not to use it because you like coming up with answers yourself, because it is against your principles, or because you can do better without it, that's fine. Hand tools still have a place in the workshop after all, and craftsmen know to use them. But if AI is helpful to you and lets you spare some cognitive load for other tasks, by all means use it, it would be stupid not to.
The meta at this point seems to be mostly around "knowing what to ask for". Eg: I suck at prompting for specific design tweaks, because I don't have the design vocabulary to be specific and articulate what I want. It's a lot of trial and error. I hear people who are developing full scale applications without a tech background run into the issue writ large. For now, I have an advantage over these people when building applications. But perhaps that's just one round of RL away from being obsolete too. Supposedly the model will eventually have such good taste in design/architecture/whatever there will be little point in asking for tweaks.
The only thing that seems relatively future-proof is just being able to communicate clearly what you're trying to do. Not how to do it, or even what success looks like. Perhaps an element of it is even as basic as being honest with yourself.
It seems that some of the newer lab machines require internet access because you can't run lab tests unless you have a paid subscription to use the machine. No credits on the subscription, no tests.
I mention this b/c a lot of commenters here are saying "but why would you ever hook up a freezer to the internet??"
I find that type of comment fascinating as many of the people on HN worked for firms that built the technology required to use internet based subscription models.
- Quoted phrase in the opening sentence
- Opening quote: “Danfoss CVE”
Why don't they just--- no margin. But there were CVEs-- no margin.
The laptop class doesn't always grok that the shift-work labor class has nothing in terms of luxuries like security reviews and gear replaced on cycles and other basics taken for granted.
But I'd setup Tailscale and-- would you, the grocery store manager running on 3% margin, really? What if you're part of a chain and it's centrally controlled elsewhere, like all the other decade-old tech you use but have to call and wait for a dispatched tech to handle?
Secure-by-default laws will eventually help, but there's always going to be groaning infrastructure particularly in commodity industries, and we still haven't onboarded most of the disaffected men to the Internet that will commit innumerable more hacks and disruptions.
I wonder if the incremental efficiency gains from this outweigh the damages. That doesn't even sound like a bad idea, unless of course nobody stops to think "but what if someone hacks this".
Banking apps closed 99% of high street banks (here in the UK at least).
Amazon is basically destroying the shopping high streets.
Technology has a way of disrupting what we've been doing for generations. I think the whole AI thing isn't made up - we're going to see rapid change in a very short time frame. Personally I'm trying my hardest to adjust, as I don't want AI to "happen" to me, by learning the skills that I can only predict will be important.
* Some figures may have been pulled out of the air :)
It's not only ahistorical to fail to consider preferences and politics, but it's also an indictment of the author's ability to judge the quality of a machine output versus that of a competent person. No doubt that like any of the examples above, some displacement of resources and effort will happen. But that doesn't indicate that something as central as "thinking" (lol) will fade into the past.
- Quoted phrase in the opening sentence
- Opening quote: “Four things had to change, and they stack:”
When you write your blog post with claude, I start hitting these phrases and then start only skimming and looking for them, and stop reading what you have to write.
Since a large fraction of MMPs sold were genre fiction and a large fraction of that has shifted to ebooks (romance readers were the biggest early adopters of the kindle), along with the decline of book sales outside of bookstore channels, the loss of MMPs isn’t the cultural cataclysm the article claims. Meanwhile, it’s worth noting that the trade paperback space has expanded. It’s not uncommon for a new writer’s first book to be released as a trade paperback rather than hardcover (Emily St John Mandel is one notable example of a writer who debuted in TP but had later books released in HC).
⸻
1. It was typically pure wood-pulp paper with high acid content, leading these books to be described as “pulp novels” and since a large fraction of these were genre novels, genre fiction would be referred to as pulp fiction and yes, that’s where the movie gets its name from.
Everything the post mentions is 100% true and well-reasoned, especially with the inclusion of pledge/chroot/etc. for dropping privileges.
There's two diametrically opposed deployment mechanisms:
1) stack towers of glop on top of each other until you're running a simple `echo "hello world"` on top of a docker container inside a vm in a proxmox server as part of a k8s cluster under a hypervisor running on remote compute...
2) just run `bash -c 'echo "hello world"'` and front it with CGI (stateless, single binary, yada yada yada...)
Take a peek at https://github.com/Domain/bash.cgi ... not that i'm suggesting that 'bash over busybox using cgi-bin' is the "proper architecture", it's more trying to illustrate that the amount of code that holds you back from being nearly truly "frameworkless" is tiny and easy to reason about.
For some of the internal tooling I'm doing for work, I've been tending towards:
some-cmd [--json|--markdown] ...yet-more-args...
- Quoted phrase in the opening sentence
- Opening quote: “hallucinates”
- Quoted phrase in the opening sentence
- Opening quote: “Show HN”
> Why people pick Doory
For a project with 6 commits over 3 days? Who are these people?
Anyone reading this, take it as a lesson to always read your LLM produced copy. They will make up $hit and it may cost you your reputation.
Many reasons:
1. The people on HN tend to think they're very clever, but they're often not as clever as they think they are. That has a lot of wide-ranging implications, from classic "engineer's disease" thinking to stuff a lot dumber.
2. I wouldn't be surprised at some unexpectedly large amount of AstroTurfing, especially considering the sheer amount money is flowing into the AI space. My understanding is HN has some influential people reading it, and is influential on a certain fraction of the wider group of software developers.
3. I think there's a certain amount of defensiveness from AI enthusiasts. They've enthusiastically adopted AI to a certain degree, so critique/skepticism of their choice is felt as a personal criticism.
4. HN is a pedantic place (see 1), so "everybody all the time in favour of AI" will be interpreted literally as false, when it really should be interpreted as "why is there so much pro-AI activity."
Most often we’re satisfied with signals and indicators that it works and we build theories and mental models that use to try to avoid the effort of proving anything.
Worse is what happens when you start using LLM models to generate the code for you. I find you start losing the mental model short-cuts; you start to trick yourself into understanding the code. You understand little.
And what few studies there are of code review suggest that we have very little impact on code quality after reading more than a couple hundred lines of code over an hour.
So the current mental leap is, “the AI is a better programmer than I am.”
I think it’s still worth becoming a better programmer regardless. And there’s no royal road. You just have to write code and struggle with writing bad code.
- Quoted phrase in the opening sentence
- Opening quote: “feel”
> I think that is fine-tuning. Many, many, many hours and days spent actually sitting down and playing the game.
Turns out quality takes time. Who would have thought. Very few games reach this level of fine-tuning, because it takes dedication (and budget).
Funnily, it's almost conflicting with this answer:
> Was much changed during development?
> Apart from the number of polygons use, the characters were first-off – once they were designed, they went straight into the game. There wasn’t really a lot of changing, artists just kept supplying stuff that just got put in the game, conveyor-belt style. That was more or less because of a very, very tight deadline – it had to be out right when the N64 came out.
So it's more nuanced. You fine-tune the fundamentals, but not necessarily everything.
The internal models are more eager, and there is the scale difference of being able to spin up 10.000 vs 10, but to get something valuable out requires skill.
Agents that produce good results 90% of the time vs 93% of the time are not enough of a difference. Multiplied too many times, that error rate is too high for scale to become "magic".
Finding proofs is one of the exceptions.
Just look at the codex/claude harness. They regularly lag behind 'external' users.
I feel my main purpose isn't that I can technically do something my clients can't, but that I know how to interpret their needs and find the solutions that fit for them.
Now they just query some AI which spits out >some< solution and they won't know that they are wrong until it can't even be tracked back to their AI.
This will result in "but the computer said" excuses all over the place.
Alternative interpretation of the numbers; there's only a subset of productive work that LLM's are 'capable' of doing at this point. That is, work that's verifiable by a machine or doesn't require verification. Thus, the cleavage points are along that axis and-- naturally-- most human beings don't have that kind of work.
Your car is still the same. Your dishwasher is still the same. The train you take to work is still the same and never comes on time. Fuel is more expensive. The roads are still congested with traffic. Your kids are (probably) doing worse at school. Food costs more. Houses cost more. Rent is higher. Buying a computer or a PS5 is more expensive. Politics is still full of mentally unstable people. The environment is getting worse. You will still die of heart disease or cancer. Food quality is worse. Your kids can't find work. Wealth inequality is accelerating.
But hey...on the flip side...a lot of people are rapidly building software (that nobody is using), and AI is solving mathematics problems (that nobody - including mathematicians - wanted it to solve)
Imagine what a "country of geniuses in a datacenter" will be able to do? Apparently...nothing.
That's convenient.
Also, I received a similar response when inquiring about a financial account that had been overlooked (there are reasons, I'm not getting into it) for a couple of decades. "Can't help you. Too old for us to have records." Well, my records are printed documents. Why can't you tell me anything? Reminds me of when the USA government stole my money from a bank transfer claiming suspicion of whatever (my landlord had a Middle Eastern sounding name and it was the 2000s) and then said they had no record of it when I provided all the required documentation to recover it.
In both cases, unfortunately, I concluded that the odds are so stacked against me that I gave up. Convenient system if you're fabulously wealthy or a massive org (such as corporations and government).
In Japan, this is already the reality. The FSA required brokerages to implement passkey authentication, and the effect is that it's no longer possible to have a brokerage account without a passkey. (Some brokerages, like Nomura Securities, will offer two separate client portals. The portal with money and trading-related functionality requires a passkey).
The thing that sucks a lot is that brokerages seem to also restrict the allowed software/hardware. Don't know the details, but none of my FIDO2 hardware keys were able to be used. I consistently use these hardware keys through WebAuthn, so I was surprised when my brokerage explicitly told me my hardware key is unsupported and I must use either Windows Hello or the iOS Passwords app.
The FSA's official explanation for mandating passkey implementation is that people kept falling for phishing attacks, and the attacks were getting too sophisticated thanks to "AI" or something.
- Quoted phrase in the opening sentence
- Opening quote: “good guys”
It immediately creates a huge lock-in in a hard-to-notice way. Luckily, I did noticed early on and manually removed the 20 or so passkeys I had created.
Earlier this year, I lost my iPhone and restored from backup. I was shocked to see that I could log into my 1Password account using only the master-key, and that the secret-key was not required. 1Password clearly stated that the secret-key was required to setup a new device, but it turns out that they also (perhaps recently) configure it to upload to all devices in your iCloud account[1]. Ever since then, I've lost faith in 1Password's ability to communicate their security model to me and started looking for an off-ramp.
1. You can't use Firefox with our website (when I got this popup, I thought I had been transported back in time 20 years, but with Chrome substituted for IE).
2. You can't sign up for accounts with an e-mail from a "custom domain"
2a. I ran some tests; it's the domain itself, not the host, so e.g. foo@gmail.com is allowed but not a foo@somewhereelse.com, even if hosted by google).
2b. After a bit of back-and-forth with support: "No, we won't tell you the list of allowed domains, but gmail and hotmail both will work."
But there’s definitely a kernel of truth here. 1Password recently began to enforce userVerification: required[1][2], and suddenly on the desktop without TouchID I have to enter my login password to use the passkey on a good number of websites. I didn’t sign up for 1Password 15 years ago so that I have to enter both my very complex master password and very complex login password to sign into websites, the “one password” contract is suddenly broken after 15 years. This experience independently alarmed me about the additional constraints potential of the passkey standard and power wielded by the FIDO Alliance and partnering organizations addressed in TFA, which are, to be clear, useful for security in an organization. (We use userVerification: preferred now btw.)
Therefore, I believe the passkey technology is great for enforcing a standard of security in an organization and stand by my choice for my org. But as a general purpose B2C auth tech, if the additional constraints become widely adopted, user freedom will be seriously harmed.
Edit: I asked Opus 5.5 to vibe code a passkey auth solution in Go a few times without saying anything specific about userVerification. It chose required each time. So expect to see more of these. Oh and vibe coding auth without understanding it entirely is bad, please don’t do that.
[1] https://web.dev/articles/webauthn-user-verification
[2] https://www.1password.community/1password-at-home-31/user-ve...
Therefore, the author's scenario that sites will wait until most of their users are using attested passkeys and then "flip the switch and start rejecting non-attested passkeys" isn't very plausible.
Attestation is certainly a stain on the WebAuthn spec, and it would have been better if Chrome and Firefox didn't support attestation (unless perhaps the browser is enrolled in enterprise management), but this blog post is presenting a hypothetical and unlikely scenario. That's not a good enough reason to recommend TOTP instead, given that TOTP is vulnerable to phishing and passkeys aren't.
If consumer-facing sites start requesting attestation when registering passkeys, they deserve to be vigorously called out, and maybe then the recommendation to use passkeys should be revisited. But until that happens, don't tell people not to use passkeys.
- Ordinary term used more than once in the same comment
- “hologram” appears 3 times
"Hologram" is one of those words that has a wide variety of overlapping meanings. You can be a strict prescriptionist about it and insist that a real hologram involves laser light, interference patterns, 3D scene replication, phase preservation etc, but there will always be something that doesn't quite meet your definition that nevertheless is called a "hologram" by large numbers of people, even expert optical scientists:
A digital holographic stereogram prints an array of hogels (holographic pixels); the hogels themselves are holograms, but are mutually incoherent across the plate. When implemented to the highest level, the result is perceptually and behaviorally indistinguishable from an analog hologram.
A rainbow hologram (such as are found on credit cards) has single parallax only - it certainly does not capture all the properties of any particular electromagnetic field. No purity here, it's just a nifty tech for making things look 3D that happens to exploit wave interference. And everyone calls those holograms.
A scratch hologram is, to a first approximation, an extremely coarse version of a rainbow hologram, made by hand by engraving curved grooves in a reflective material. Now we're away from lasers and wave interference, and for this we sacrifice color - but geometrically we are still dealing with directional microreflections on a surface. Hologram? Depends who you ask.
Lenticular holograms use refraction rather than diffraction to achieve essentially the same imaging model as a rainbow hologram, but in a decidedly less finicky way, at the expense of some resolution. The same imaging data could be used to make a lenticular or a rainbow hologram. No lasers, no diffraction. Hologram? Depends who you ask...
My advice, as someone who has worked with holograms for a decade, is to give up. Saying "AKTUALLY Tupac/Abba isn't a hologram" gets old after the 1000th time. In the real world, "hologram" means "intangible optical simulacra", regardless of how it was achieved. If you want more precision than that, you have to specify the actual technology - Denisyuk, rainbow, lenticular, etc...
When someone asks "can you make a hologram of this", you don't lecture them on optical physics. You say "sure, you want the expensive kind or the plastic kind?"
3blue1brown has a great explainer video on what holograms are: https://youtu.be/EmKQsSDlaa4
- Quoted phrase in the opening sentence
- Opening quote: “Pace the Frontier!!1”
For the IPO, I assume they're going to try to get everyone to focus on inference - which is profitable in isolation. But I don't know how they'll distract the market from the other half of the equation.
- Quoted phrase in the opening sentence
- Opening quote: “leave the multibillion dollar company alone”
- Quoted phrase in the opening sentence
- Opening quote: “Vibe coders are paedophiles”
Felt like that undermined some of the more cogent points the article was making.
I've actually been using HarmonyOS for half a year now, and it feels pretty good — the notifications and system smoothness are on par with iOS. The reason I don't use an iPhone is that I need NFC access control, which iOS doesn't support.
Huawei has invested a large amount of manpower to station people in major companies to help develop HarmonyOS. I think they have already broken the vicious cycle of "no users -> no apps -> even fewer users."
Moreover, in the current AI era, it has become easier for Chinese developers to port software to HarmonyOS. For example, Moonlight, the streaming software I often use, has several natively customized and optimized versions available in the HarmonyOS app store.
It is extremely far from a working solution for someone who uses these apps for just casual non-professional use. For professional use, I borrow Wolfgang Pauli's "not even wrong" to say it is not even something to think of. People who think this is a possible replacement just don't have a clue about how big a software like Photoshop and MS Word are, and how much they actually provide their users with.
I know that "Hey I open-source'd Adobe and Microsoft" can get people some hype on TikTok and Twitter, and maybe have some media reporting about them for fun, but I am really surprised that people here are passing on these like they are really what they claim to be. Genuinely, if anything can be used as a counterargument for the promised LLMs revolution, it would be software like this.
If you go out and illegally cut down a tree, that's not "only a procedural issue". They illegally logged 815 acres of forest.
But it wasn't their own high standards, it was a legal obligation that they neglected?
I hope in vain that the powers that be are able to punish them in a way they will actually feel...
snopes.com (1994)
gamefaqs.com (1995)
archive.org (1996)
wikipedia.org (2001)
tvtropes.org (2004)
openstreetmap.org (2004)
github.com (2008)
I really think they encapsulate the best of what I like about the internet. I've been formulating this list for quite some time. Limiting to just 7 keeps things interesting! For example, it's a pity steampowered.com had to be left out but I see it less as a "website" proper. I wonder (ha) how my list differs from everyone else's.
I expect more businesses will transition to this strategy to counter NYC's new rule. They won't necessarily lower the price, but you will have to sign up for a one-year term, then they will hit you with a massive cancellation fee if you "click to cancel".
Source: I'm an EU digital policy analyst in Brussels.
- Quoted phrase in the opening sentence
- Opening quote: “annoying”
But don't feel bad if you have not been able to do any of them, e.g. Most People with kids can often do very little than job and take care of kids and that's it. What is more important is to not beat yourself up about it, give yourself grace and do one or two things that make you and your near and dear people/pets happy.
- Quoted phrase in the opening sentence
- Opening quote: “as soon as we started thinking for you it really became our civilization”
remember this timeless Agent Smith quote
- Quoted phrase in the opening sentence
- Opening quote: “reverse engineered”
I ask nearly every doctor/researched in the medical field: if you had a AI-created drug that tremendously improved cancer treatment outcomes for your patient, would you hesitate to prescribe it because nobody understood how it worked? I have yet to hear "yes, I would hesitate", most people say "it would be cruel to deny a person a treatment that worked".
I think Tao is focusing too much on second order effects of AI on math and other fields; humans are terrible at reasoning about second order effects, especially ones that are happening dynamically, in real time, using the most advanced mathematical models the world has yet created.
One idea Terrance Tao conjectures, which is highly doubtful, is that spamming the AI button will solve open problems without producing insightful new methods. But the OpenAI drop would seem to disprove this. The sub O(nlogn) proof for DFT for example violated very old human assumptions. Decades of work in the field was incremental progress on sub optimal method that nobody questioned hard enough. More generally, we should always be able to go back to a super-human AI and say, "Attack this problem, but don't use a method tried before."
In all seriousness, anyone remember when every library, business, government agency, and institution, etc was buying $20,000 cisco routers to "get on the internet" during the dot com boom/bust in the late 90's ?
Only to see them all end up on the surplus market, unused, for a couple hundred ?
We're in the AI boom/bust cycle, and its my understanding DRAM fab is specialized enough, that the supply is getting squeezed.
As for "doing it on purpose" ? It's being milked for all its worth. That's what companies do. They milk the market for what it will pay.
Also, for everyone who is trying to do local AI, on Xeon or EPYC platforms, memory bandwidth scales with module count, making them that much more in demand in terms of processing capacity / wafer area.
I live in the US, so I'd probably rather have the CCP snoop through my data than have the US government do it. If I lived in China, it would be the opposite.