> But I think we're in world where very soon, companies will be demanding their engineering departments converge to the lived experience of the people who are seeing something like the author.
I think this part is very real.
If you’re in this thread saying “I don’t get it” you are in danger much faster than your coworker who is using it every day and succeeding at getting around AI’s quirks to be productive.
> The real danger is if management sees this as acceptable. If so best of luck to everyone.
Already happening. It's just an extension of the "move fast and break stuff" mantra, only faster. I think the jury is still out on if more or less things will break, but it's starting to look like not enough to pump the brakes.
> Be careful here. I have more coworkers contributing slop and causing production issues than 10x’ing themselves.
Sure, many such cases. We'll all have work for a while, if only so that management has someone to yell at when things break in prod. And break they will -- the technology is not perfected and many are now moving faster than they can actually vet the results. There is obvious risk here.
But the curve we're on is also obvious now. I'm seeing massive improvements in reliability with every model drop. And the model drops are happening faster now. There is less of an excuse than ever for not using the tools to improve your productivity.
I think the near future is going to be something like a high-speed drag race. Going slow isn't an option. Everyone will have to go fast. Many will crash. Some won't and they will win.
> I think the near future is going to be something like a high-speed drag race. Going slow isn't an option. Everyone will have to go fast. Many will crash. Some won't and they will win.
I think this is right. This is what we as engineers have to wrap our minds around. This is the game we're in now, like it or not.
> Many will crash.
Aside from alignment, and some of these bigger picture concerns, prompt injection looms large. It's an astoundingly large, possibly unsolvable vector for all sorts of mayhem. But many people are making the judgment that there's too much to be gained before the shocks hit them. So far, they're right.
If a company lets faulty code get to production, that's an issue no matter how it is produced. Agentic coding can produce code at much higher volumes, but I think we're still in the early days of figuring out how to scale quality and the other nonfunctional requirements. (I do believe that we're literally talking about days, though, when it comes to some facets of some of these problems.)
But there's nothing inherent about agentic coding to lead to slop outcomes. If you're steering it as a human, you can tweak the output, by hand or agentically, until it matches your expectations. It's not currently a silver bullet.
That said, my experience is that the compressing of the research, initial draft process, and revision--all which used to be the bulk of my job--is radical.
This is often a commonly blamed reason, but I think the data at this point pretty strongly suggests that the more affluent a country is the less kids they have.
You look at some of the most third world places in the world without strong economic security, yet somehow they manage to have babies at a higher rate than Western countries do.
Seems like when you give women the choice, many elect to have fewer kids than replacement level.
Hell, in many countries in Europe, they basically throw money at anyone having kids and their birthrate has plummeted which would indicate that economics is not only reason.
I don't think there's a country in Europe that funds childcare remotely to the level of cost. The most generous I'm aware of is certain states / cities in Germany that provide free 'Kita', essentially Kindergarten. In addition to maternity leave, national insurance etc. But this certainly doesn't cover the numerous costs (including time off work etc) associated with having kids.
Would be an interesting experiment to actually pay people to have kids - i.e.: financially reward them in accordance with the costs involved. I suspect, as with an actual liveable UBI, the results would differ radically.
We do pay people to have kids in the USA - once you're on welfare. Your WIC and EBT allowances go up per kid.
And even if you're not that poor, you get subsidized kids through things like the earned income tax credit and the child tax credit. It's annoying that while some of those support 3+ kids, many "top out" at three and stop increasing.
I've often thought of searching for "sponsorships" for additional children (though we'd probably have them anyway) - not sure I want my son to be named Facebook X AI though ;)
You've missed my point... Those allowances and subsidies don't remotely cover the cost of having children. Especially in the US with the wild costs of hospital childbirth itself.
There needs to be a societal change where motherhood is not only respected but celebrated - why we are now in a society where it's looked down upon (not verbally but by actions) could be pondered.
Yes, the "cost of having kids" argument is 100% bunk. Africans in abject poverty are having 6-7 kids, while individuals living in the richest countries are having 1 or none even though they clearly can afford many more.
Even within Western countries income is negatively correlated with fertility - those most able to afford kids are having the least number of kids.
Don’t get me wrong Gemini 3 is very impressive! It just seems to always need to give you an answer, even if it has to make it up.
This was also largely how ChatGPT behaved before 5, but OpenAI has gotten much much better at having the model admit it doesn’t know or tell you that the thing you’re looking for doesn’t exist instead of hallucinating something plausible sounding.
Recent example, I was trying to fetch some specific data using an API, and after reading the API docs, I couldn’t figure out how to get it. I asked Gemini 3 since my company pays for that. Gemini gave me a plausible sounding API call to make… which did not work and was completely made up.
Okay, I haven't really tested hallucinations like this, that may well be true. There is another weakness of GPT-5 (including 5.1 and 5.2) I discovered: I have a neat philosophical paradox about information value. This is not in the pre-training data, because I came up with the paradox myself, and I haven't posted it online. So asking a model to solve the paradox is a nice little intelligence test about informal/philosophical reasoning ability.
If I ask ChatGPT to solve it, the non-thinking GPT-5 model usually starts out confidently with a completely wrong answer and then smoothly transitions into the correct answer. Though without flagging that half the answer was wrong. Overall not too bad.
But if I choose the reasoning GPT-5 model, it thinks hardly at all (6 seconds when I just tried) and then gives a completely wrong answer, e.g. about why a premiss technically doesn't hold under contrived conditions, ignoring the fact that the paradox persists even with those circumstances excluded. Basically, it both over- and underthinks the problem. When you tell it that it can ignore those edge cases because they don't affect the paradox, it overthinks things even more and comes up with other wrong solutions that get increasingly technical and confused.
So in this case the GPT-5 reasoning model is actually worse than the version without reasoning. Which is kind of impressive. Gemini 3 Pro generally just gives the correct answer here (it always uses reasoning).
Though I admit this is just a single example and hardly significant. I guess it reveals that the reasoning training is trained hard on more verifiable things like math and coding but very brittle at philosophical thinking that isn't just repeating knowledge it gained during pre-training.
Maybe another interesting data point: If you ask either of ChatGPT/Gemini why there are so many dark mode websites (black background with white text) but basically no dark mode books, both models come up with contrived explanations involving printing costs. Which would be highly irrelevant for modern printers. There is a far better explanation than that, but both LLMs a) can't think of it (which isn't too bad, the explanation isn't trivial) and b) are unable to say "Sorry, I don't really know", which is much worse.
Basically, if you ask either LLM for an explanation for something, they seem to always try to answer (with complete confidence) with some explanation, even if it is a terrible explanation. That seems related to the hallucination you mentioned, because in both cases the model can't express its uncertainty.
Anyone who’s been a developer for more than 10 minutes knows that best practices are hard to always follow through on when there’s pressure to ship.
But there’s more time to do some of these other things if the actual coding time is trending toward zero.
And the importance of it can go up with AI systems because they do actually use the documentation you write as part of their context! Direct visible value can lead people to finally take more seriously things that previously felt like luxuries they didn’t have time for.
Again if you’ve been a developer for more than 10 minutes, you’ve had the discouraging experience of pain-stakingly writing very good documentation only for it to be ignored by the next guy. This isn’t how LLMs work. They read your docs.
> Anyone who’s been a developer for more than 10 minutes knows that best practices are hard to always follow through on when there’s pressure to ship.
>
But there’s more time to do some of these other things if the actual coding time is trending toward zero.
I think you'll find even less time - as "AI" drives the target time to ship toward zero.
I agree that this will be the end result over time, maybe even faster than we expect. And as those speed pressures increase, AI will take over more and more of the development process.
Where is all the amazing software and/or improvements in software quality that is supposed to be coming from this revolution?
So far the only output is the "How I use AI blogs", AI marketing blogs, more CVEs, more outages, degraded software quality, and not much of shipping anything.
Is there any examples of real products and not just anecdotes of "I'm 10x more productive!"?
I was in the same mindset until I actually took the Claude code course they offer. I was doing so much wrong.
The two main takeaways. Create a CLAUDE.md file that defines everything about the project. Have Claude feed back into the file when it makes mistakes and how to fix them.
Now it creates well structured code and production level applications. I still double check everything of course, but the level of errors is much lower.
An example application it created from a CLAUDE.md I wrote. The application reads multiple PDF's, finds the key stakeholders and related data, then generates a network graph across those files and renders it in an explorable graph in Godot.
That took 3 hours to make, test. It also supports OpenAI (lmstudio), Claude and Ollama for its LLM callouts.
What issue I can see happening is the duplication of assets in work. Instead of finding an asset someone built, people have been creating their own.
Do you find reading hard? I'm asking for examples. Why isn't anyone showing this off in blog posts. Or a youtube video or something. It's always this vague, it's faster, just trust me bro bullshit and I'm sick of it. Show me or don't reply.
So you want a video of me coding at work using AI? There are entire YouTube channels dedicated to this already. There are already copious blogs about people's AI workflows -- this very post you're commenting in is one (do you find reading hard?)
Clarify the actual thing you need to believe the technology is real or don't reply.
Its only revolutionary if you think engineers were slow before or software was not being delivered fast enough. Its revolutionary for some people sure, but everyone is in a different situation, so one man's trash can be other man's treasure. Most people are treading both paths as automation threatens their livelihood and work they loved, also still not able to understand why would people pay to companies that are actively trying to convince your employer that your job is worthless.
Even If I like this tech, I still dont want to support the companies who make it. Yet to pay a cent to these companies, still using the credits given to me by my employer.
I think there are four fundamental issues here for us...
1. There are actually less software jobs out there, with huge layoffs still going on, so software engineering as a profession doesn't seem to profit from AI.
2. The remaining engineers are expected by their employers to ship more. Even if they can manage that using AI, there will be higher pressure and higher stress on them, which makes their work less fulfilling, more prone to burnout etc.
3. Tied to the previous - this increases workism, measuring people, engineers by some output benchmark alone, treating them more like factory workers instead of expert, free-thinking individuals (often with higher education degrees). Which again degrades this profession as a whole.
3. Measuring developer productivity hasn't really been cracked before either, and still after AI, there is not a lot of real data proving that these tools actually make us more productive, whatever that may be. There is only anecdotal evidence: I did this in X time, when it would have taken me otherwise Y time - but at the same time it's well known that estimating software delivery timelines is next to impossible, meaning, the estimation of "Y" is probably flawed.
So a lot of things going on apart from "the world will surely need more software".
Yes, but who did we automate out of a job by building crappy software? Accountants are more threatened by AI than any of the software we created before, same with Lawyers, teachers. We didnt automate any physical labourers out of a job too.
It's insane! We are so far beyond gpt-3.5 and gpt-4. If you're not approaching Claude Code and other agentic coding agents with an open mind with the goal of deriving as much value from them as possible, you are missing out on super powers.
On the flip side, anyone who believes you can create quality products with these tools without actually working hard is also deluded. My productivity is insane, what I can create in a long coding session is incredible, but I am working hard the whole time, reviewing outputs, devising GOOD integration/e2e tests to actually test the system, manually testing the whole time, keeping my eyes open for stereotypically bad model behaviors like creating fallbacks, deleting code to fulfill some objective.
It's actually downright a pain in the ass and a very unpleasant experience working in this way. I remember the sheer flow state I used to get into when doing deep programming where you are so immersed in managing the states and modeling the system. The current way of programming for me doesn't seem to provide that with the models. So there are aspects of how I have programmed my whole life that I dearly miss. Hours used to fly past me without me being the wiser due to flow. Now that's no longer the case most of the times.
The real value that AI provides is the speed at which it works, and its almost human-like ability to “get it” and reasonably handle ambiguity. Almost like tasking a fellow engineer. That’s the value.
By the time you do everything outlined here you’ve basically recreated waterfall and lost all speed advantage. Might as well write the code yourself and just use AI as first-pass peer review on the code you’ve written.
A lot of the things the writer points out also feel like safeguards against the pitfalls of older models.
I do agree with their 12th point. The smaller your task the easier to verify that the model hasn’t lost the plot. It’s better to go fast with smaller updates that can be validated, and the combination of those small updates gives you your final result. That is still agile without going full “specifications document” waterfall.
It’s a solid post overall and even for people with a lot of experience there’s some good ideas in here. “Identify and mark functions that have a high security risk, such as authentication, authorization” is one such good idea - I take more time when the code is in these areas but an explicit marking system is a great suggestion. In addition to immediate review benefits, it means that future updates will have that context.
“Break things down” is something most of us do instinctively now but it’s something I see less experienced people fail at all the time.
Brain surgery is probably a bad example... or maybe a good one, but for different reasons?
Brain surgery is highly technical AND highly vibe based.
You need both in extremely high quantities. Every brain is different, so the super detailed technical anatomies that we have is never enough, and the surgeon needs constant feedback (and insanely long/deep focus).
This is the way. People are unfortunately starting to divide themselves into camps on this — it’s human nature we’re tribal - but we should try to avoid turning this into a Yankees Redsox.
Both companies are producing incredible models and I’m glad they have strengths because if you use them both where appropriate it means you have more coverage for important work.
I get so annoyed by this Socratic line of questioning because it’s extremely obvious.
Terrorist has plans and contacts on laptop/phone. Society has a very reasonable interest in that information.
But of course there is the rational counter argument of “the government designates who is a terrorist”, and the Trump admin has gleefully flouted norms around that designation endangering rule of law.
So all of us are adults here and we understand this is complicated. People have a vested interest in privacy protections. Society and government often have reasonable interest in going after bad guys.
Mediating this clear tension is what makes this so hard and silly lines of questioning like this try to pretend it’s simple.
The better rational counter argument is that "privacy is a human right enshrined in international law". Society has zero business knowing anyone's private communications, whether or not that person is a terrorist. There is nothing natural about being unable to talk to people privately without your speech being recorded for millions of people to view forever. Moreover, giving society absolute access to private communications is a short road to absolute dystopia as government uses it to completely wipe out all dissent, execute all the Jews or whatever arbitrary enemy of the state they decide on, etc.
You do not get to dispense with human rights because terrorists use them too. Terrorists use knives, cars, computers, phones, clothes... where will we be if we take away everything because we have a vested interested in denying anything a terrorist might take advantage of?
Who decided absolute privacy in all circumstances is a fundamental human right? I don’t think any government endorses that position. I don’t know what international law you speak of. You’re basing your argument on an axiom that I don’t think everyone would agree with.
This sounds like a Tim Cook aphorism (right before he hands the iCloud keys to the CCP) — not anything with any real legal basis.
Article 12 of the United Nation's Declaration of Human Rights:
> No one shall be subjected to arbitrary interference with his privacy [...]
which has later been affirmed to include digital privacy.
> I don’t think any government endorses that position.
Many governments are in flagrant violation of even their own privacy laws, but that does not make those laws any less real.
The UN's notion of human rights were an "axiom" founded from learned experience and the horrors that were committed in the years preceding their formation. Discarding them is to discard the wisdom we gained from the loss of tens of millions of people. And while you claim that society has a vested interest in violating a terrorist's privacy, you can only come to that conclusion if you engage in short-term thinking that terminates at exactly the step you violate the terrorist's rights and do not consider the consequences of anything beyond that; if you do consider the consequences it becomes clear that society collectively has a bigger vested interest in protecting the existence of human rights.
Admittedly "arbitrary" is something of a legal weasel word that leaves a lot of room for interpretation. I lean towards a strong interpretation for two reasons: the first is because it is logically obvious why you must give it a strong interpretation; if the people responsible for enforcing human rights can arbitrarily decide you don't have them, you don't have human rights. The second is because we have seen this play out in the real world and it is abundantly clear that the damage to society is greater than any potential benefits. The US in particular has made an adventure out of arbitrarily suspending human rights, giving us wonderful treats like Guantanamo Bay and the black sites across the Middle East. I don't know what part of that experiment looked remotely convincing to you, but to me they only reinforced how clearly necessary inviolable human rights are for the greater good of society.
>if the people responsible for enforcing human rights can arbitrarily decide you don't have them, you don't have human rights
But the "arbitrary" there is too account for the situation where the democratic application of the law wants to inspect the communications of suspected terrorists, and where a judge agrees there is sufficient evidence to grant a warrant.
Unfortunately, that law does nothing against situations like the USA/Russia regime where a ruler dispenses with the rule of law (and democratic legal processes too).
You can't practically have that sort of liberalism, where society just shrugs and chooses not to read terrorists communications, those who wish to use violence make it unworkable.
But if you want to make it possible for the Feds to break into a terrorist's secure phone, you have to make it impossible for anyone to have a secure phone.
That is arbitrary interference with all our privacy.
Usually such "international laws" are only advisory and not binding on member nations. After decades of member nations flouting UN "laws" I can't see them as reliable or effective support in most arguments. I support the policy behind the privacy "laws" of the UN, but enforcing them seems to fall short.
Enforcement mechanisms are weak, but they still exist to set a cultural norm and an ideal to strive towards. Regardless, I have also laid out an argument at length as to why society would logically want to have this be a human right for its own good, regardless of any appeal to existing authority.
There are just things some people want and the reasons they want them.
So the question that you are so annoyed by remains unanswered (by you anyway), and so, valid, to all of us adults.
@hypfer gives a valid concern, but it's based on a different facet of lockdown. The concern is not that the rest of us should be able to break into your phone for our safety, it's the opposite, that you are not the final authority of your own property, and must simply trust Apple and the entire rest of society via our ability to compel Apple, not to break into your phone or it's backup.
At the risk of being kind of ass, which I've been trying to be better about lately, I'm going to offer some advice. If you can't even respond to a question about secure computing without bringing American presidential politics into things, perhaps you need to take a break from the news for a few weeks.
The reason I asked that question is because I don't think it's complicated. I should be able to lock down my device such that no other human being on the planet can see or access anything on it. It's mine. I own it. I can do with it whatever I please, and any government that says otherwise is diametrically opposed to my rights as a human being.
You are more likely to be struck by lightning while holding two winning lottery tickets from different lotteries than you are to be killed by an act of terrorism today. This is pearl-clutching, authoritarian nonsense. To echo the sibling comment, society does not get to destroy my civil rights because some inbred religious fanatics in a cave somewhere want to blow up a train.
Edit: And asking for someone to says "there are concerns!" to proffer even a single one is not a Socratic line of questioning, it's basic inquiry.
The line of reasoning is more like this: if you make and sell safe-cracking tools then it would not be unreasonable for the government to regulate it so only registered locksmiths could buy it. You don't want people profiting from the support of criminal acts.
The government could similarly argue that if a company provides communication as a service, they should be able to provide access to the government given they have a warrant.
If you explicitly create a service to circumvent this then you're trying to profit from and aid those with criminal intent. Silkroad/drug sales and child sexual content are more common, but terrorism would also be on the list.
I disagree with this logic, but those are the well-known, often cited concerns.
There is a trade-off in personal privacy versus police ability to investigate and enforce laws.
> ...the Trump admin has gleefully flouted norms around that designation...
One would have to hold a fairly uninformed view of history to think the norms around that designation are anything but invasive. The list since FDR is utterly extensive.
I didn’t say he was the first to abuse powers. Indeed it’s kind of silly to even have to clarify “but other administrations…” because that’s fairly obvious to anyone old enough to have seen more than one president.
But the article is literally referencing the Trump administration seizing a reporter’s phone so the current administration’s overreach seems relevant here.
My point was that your stated assumption of what the norms are is inaccurate. If nearly every modern administration does it, that is literally the norm. The present administration, like many before it, is following the norm. The norm is the broader issue.
Which makes the rest of it (and your followup) come across as needlessly tribal, as both major parties are consistently guilty of tending to object to something only when the other side does it.
Frankly I really don’t care about both sides-ism anymore. I can agree with you that a lot of administrations have been irresponsible on this point while also believing that the current administration is particularly dangerous in this area.
If I lose you here because of “needless tribalism” oh well.
I think this part is very real.
If you’re in this thread saying “I don’t get it” you are in danger much faster than your coworker who is using it every day and succeeding at getting around AI’s quirks to be productive.
reply