Not sure why this is being downvoted. It’s spot on imo. Engineers who don’t want to understand the domain and the customers won’t be as effective in an engineering organization as those who do.
It always baffles me when someone wants to only think about the code as if it exists in a vacuum. (Although for junior engineers it’s a bit more acceptable than for senior engineers).
We're assuming we all somehow have perfect customers with technical knowledge who know exactly what they want and can express it as such, while gracefully accepting pushback over constraints brought up.
Anyone who's worked in a "bikeshed sensitive" stack of programming knows how quickly things railroad off when such customers get direct access to an engineer. Think being a fullstack dev but you constantly get requests over button colors while you're trying to get the database setup.
Okay. I'm glad you're privileged enough to where you can choose your customers. Customers that aren't abusive or otherwise out of their league thinking they know everything just because they have money.
Calling me "privileged" or "lucky" feels like a cheap attack on my competence.
I am certain that I went through the same problems you did in the past, maybe I just have a different way of dealing with them, or maybe I had even worse problems than you did but I have a different frame of comparison. We never stopped to compared notes.
All I'm saying is: for me dealing with business owners, end-users, CEOs and CTOs was always way easier than dealing with proxies. That's all.
>I am certain that I went through the same problems you did in the past,
And I'm certain you haven't if you really, never wanted a layer of separation between certain clients over behavioral issues that got in the way of the actual work. And I'm still male, so I'm sure I still have it better than certain other experiences I only heard third hand in my industry.
I don't see it as a cheap attack. Any teacher would love to be in a classroom exclusively made up of motivated honors students so they can focus on teaching and nurturing. Instead, most teachers tend to become parental proxies without the authority to actually discipline children. So they see a chair fly and at best they need to hope a principal handles it. But sometimes the kid is back in class the next day.
>you are making assumptions about my experience. Can you please stop?
You're 2 days into responding to a comment that amounted to "X depends on your exoerience". Is there something else you wish to get out of this thread?
Your complaint is an opinion. I disagree with that opinion. Unless you wish to ask about my experiences or go into yours, what's there to discuss here? Without that, I feel I said all I could on the topic.
>please stop dismissing my experience as lucky or privileged
I'm not gonna harp on it. I'll go to bed and wake up completely forgetting about this thread unless I get another notification.
But you're basically telling me to shut off my feelings. Hard to do. I don't know your experiences, so my feelings can be wrong.
I'm unsure why you are putting so much stock into an uninformed feeling on the internet. It doesn't seem like we want to expand on our stories so there's not much more to go on. And that's fine.
I don't try to assert everything about you, but I'm just explaining the vibes I got. But that's all my words are: vibes.
>Just that there might be worse problems elsewhere
I love my industry and in my personal experience I can count on one hand how many truly problematic coilkeages I've worked with or under. I am lucky in that regard for my industry.
Meanwhile, clients and consumers constantly make me question if I want to continue this career long term. My plan was always to focus more on a B2B angle to insulate from that, but the current winds blowing suggest that angle might not even exist in a decade. So I want to at least have a side hustle ready.
And despite those notions, I'm still on the lucky end in terms of what third and even secondhand accounts I've heard of. Diving more into that pool is unsettling for me, but it might still be more stable than what's going on right now.
+1, customers want their problem solved but at times they struggle to articulate that.
When a customer starts saying “we need to build X”, first ask what the actual problem is etc. It takes actual effort, and you need to speak their language (understand the domain).
But if you have a PM in the middle, now you just start playing telephone and I don’t believe that’s great for anyone involved.
Exactly. The game of telephone is prone to misinterpretation and, when this happens too much, it often answers with rigidity and lack of flexibility, out of fear.
Isn't it a bit of both? When it comes to noticing whether or not code will be a security nightmare, a performance nightmare, an architectural nightmare, etc, haven't experienced developers already learned to watch out for these issues?
Too right. Drilling into the domain from first principles and with critical faculties enabled unlocks so much more value, because the engineer can then see much better ways to solve problems.
Customer interaction has imo always been one of the most important parts in good engineering organizations. Delegating that to Product Managers adds unnecessary friction.
Having spent more hours than I care to count struggling to control my facial expressions in client-facing meetings your assertion that that friction is unnecessary is highly questionable. Having a "face man" who's sufficiently tech literate to ask decent questions manage the soft side of client relations frees up a ton of engineering resources that would otherwise be squandered replying to routine emails.
No that’s not what I meant at all. Amazon Professional Services are made up of full time “blue badge” employees who get the same type of base + bonus + RSUs that all other blue badge employees get.
I've never wanted to adapt a house that significantly. But yeah, I much prefer the cinderblock homes and miss them. Something about the wood and drywall houses just feels incredibly cheap, and I don't like the aesthetic (de gustibus et coloribus..)
Houses change over time. A house could have been build in 1920 without a toilet or central heating. Then over time it got a fireplace on the second floor, an indoor toilet, indoor bathroom, then central heating with gas, extra insulation, a couple decades later double paned windows, hybrid heating with a heat pump, then full electric heating, underfloor heating, solar panels, home battery.
Houses change a lot over time, it is nice to be adaptable and not need to carve out stone and concrete every time you add a feature to a home.
The most beautiful homes I have been inside in Europe were wooden cabins in Sweden. The exposed wood ceiling beams, the unpainted wooden panels everywhere, the little details. I never had that with stone or brick buildings. Mainly because they got plastered and painted over, you almost never see the raw materials on the inside.
It's not the unavailability of trees. European countries have wisely decides that cities built of wooden houses are prone to massive fires. USians haven't learned that lesson and the Los Angeles fire isn't going to be the last one.
A yes, the wise Europeans like the Dutch who have homes in Amsterdam that are sinking into the ground due to rotting wooden beams sinking in swamp ground and homes in Groningen with cracks all over due to the earthquakes that came with pumping gas out of the ground.
Or the dozens of structures in Italy that came crashing down, like the various bridges over the past twenty years (250 bridge collapse events in Italy between January 2000 and July 2025).
Yes us Europeans are indeed superior and we never pick the wrong building material ever.
> To each their own I guess. I’ll happily move walls, add or remove a bathroom, add windows, etc.
A sign of the restlesness. Once you find a house to settle in, why would you need to change it ? European houses are typically versatile, US houses aren't due to having closets (which make a room's layout very inflexible) as well as electrical outlets being mandated exactly in the middle of the wall precisely where one would like to place furniture. US building codes are beyond stupid.
> Terrible carbon footprint for concrete too.
Carbon footprint is not that important. I want comfort. More specifically: if you are somewhat wealthy (in the top 10% of incomes, like most of the people here), in the continental Europe you can nowadays easily buy an apartment in a Passivhaus (or almost if renovated) building, with underfloor heating throughout the place, supplied by a geothermal heat pump, with triple-glazed windows and external covers that give you the utmost quietness even when there's traffic just outside. You can't get that in the US because even if you were willing to pay, there exist only a handful of construction companies that know how to build that, and they're all booked for years.
> I know modern structures are better but I also don’t entirely trust block in an earthquake. Obviously less of a concern in most (not all!) of Europe.
You can take a look at Japan. Modern buildings can withstand earthquakes. The issue in the US is that developers are allowed to just build without a civil engineer or architect designing the building. I wouldn't trust that either.
They also claimed ChatGPT solved novel erdös problems when that wasn’t the case. Will take with a grain of salt until more external validation happened. But very cool if true!
How was that not the case? As far as I understand it ChatGPT was instrumental to solving a problem. Even if it did not entirely solve it by itself, the combination with other tools such as Lean is still very impressive, no?
My understanding is there's been around 10 erdos problems solved by GPT by now. Most of them have been found to be either in literature or a very similar problem was solved in literature. But one or two solutions are quite novel.
I am not aware of any unsolved Erdos problem that was solved via an LLM. I am aware of LLMs contributing to variations on known proofs of previously solved Erdos problems. But the issue with having an LLM combine existing solutions or modify existing published solutions is that the previous solutions are in the training data of the LLM, and in general there are many options to make variations on known proofs. Most proofs go through many iterations and simplifications over time, most of which are not sufficiently novel to even warrant publication. The proof you read in a textbook is likely a highly revised and simplified proof of what was first published.
If I'm wrong, please let me know which previously unsolved problem was solved, I would be genuinely curious to see an example of that.
"We tentatively believe Aletheia’s solution to Erdős-1051 represents an early example of an AI system autonomously resolving a slightly non-trivial open Erdős problem of somewhat broader (mild) mathematical interest, for which there exists past literature on closely-related problems [KN16], but none fully resolves Erdős-1051. Moreover, it does not appear to us that Aletheia’s solution is directly inspired by any previous human argument (unlike in
many previously discussed cases), but it does appear to involve a classical idea of moving to the series tail and applying Mahler’s criterion. The solution to Erdős-1051 was generalized further, in a collaborative effort by Aletheia together with human mathematicians and Gemini Deep Think, to produce the research paper [BKK+26]."
"The erdosproblems website shows 851 was proved in 1934." I disagree with this characterization of the Erdos problem. The statement proven in 1934 was weaker. As evidence for this, you can see that Erdos posed this problem after 1934.
You recommended I look at the erdosproblems website.
But evidence that it was posed after 1934 is not really evidence it was not solved, because one of the things we learned from LLMs was that many of these problems were already solved in the literature, or are relatively straightforward applications of known, yet obscure, results. Particularly in the world of Erdos problems, the majority of which can be described as "off the beaten path" and are basically musings in papers that Erdos was asking -- many of these are in fact solved in more obscure articles and no one made the connection until LLMs allowed us to do systematic literature searches. This was the primary source of "solutions" of these problems by LLMs in the cited paper.
The Erdos Problem site also does not say it was solved in 1934. If you read the full sentence there, it refers to a different statement proven which is related.
Yeah that was also my take-away when I was following the developments on it. But then again I don't follow it very closely so _maybe_ some novel solutions are discovered. But given how LLMs work, I'm skeptical about that.
I honestly don't see the point of the red data points. By now all the erdos problems have been attempted by AIs--so every unsolved one can be a red data point.
what? this article is making a different point if you read past the title.
> Conventional leadership advice suggests looking at decisions as reversible or non-reversible. Many important, non-reversible, decisions are recoverable, though.
I don’t think it’s different. Recoverable == Reversible to an extend. Unless you take reversible in the strictest sense of “undo” it’s different. But you can’t “undo” a leadership decision, all you can do is later correct it and recover.
So imo it’s splitting hairs over the same outcome.
An example - say you introduce 5 day return to office. Half you staff leaves and you now go back to a flexible work from home model. You don’t “undo” the damage done, but you can recover. It was a costly 2-way door.
A decision can be irreversible, yet it's negative impact recoverable.
Those words mean entirely different things.
One is about reversing a decision. One is about reversing an outcome.
Irreversible decisions mean decisions without backsies. Recoverable decisions mean decisions that leave the potential for survival and a comeback.
--
You cannot (typically) reverse a poorly-reasoned midlife-crisis divorce. But you can recover by getting one's life in order.
You cannot (typically) reverse a disastrous corporate merger. But you can recover, by identifying a better path forward.
You cannot (typically) reverse a money transfer after buying into a scam. But you can recover, by continuing to earn and save.
You can change your streaming service name from "HBO" to "Kansas" to "Pumpernickle" to "Cheese Please", and then revert to "HBO". And leverage the PR waves from a mea culpa. And, snowball the PR into a reputation as a lovable rake, by doing the same thing every year around April 1st.
But, you cannot reverse adding caffeine to 7-up after adding caffeine to 7-up, and resume your famous marketing campaign of "No caffeine. Never had it. Never will."
It took Twitter 10 years before it was profitable [1]. I'd guess that Anthropic will be one of the companies left standing when it's all said and done, assuming nothing catastrophic happens.
Also the switching cost. If its negligible theres no reason for Anthropic to be considered a going-concern in the long term. So its valuation makes no sense from a DCF basis unless you are expecting a liquidation in future. But even then, the liquidation value still doesn't justify its valuation today.
It always baffles me when someone wants to only think about the code as if it exists in a vacuum. (Although for junior engineers it’s a bit more acceptable than for senior engineers).
reply