angarg12

One the main theses is

> the models generalize well only on tasks within a small neighborhood of the specific tasks they've been trained on

Unless frontier labs have surprisingly trained their models in the exact tasks my team works on, this is patently false. We are getting very good results on automation and I'm bullish we will be able to mostly remove humans in the loop for most of our infra tasks by the end of the year.

I have no opinion on the other theses, but given that OP doesn't back up these claims in any way, I have my doubts about the conclusions of this article.

Funnily enough this very morning I asked an LLM to implement an algorithm for an NP-hard problem (a variation of the knapsack problem). I gave it 2 directives:

* Do not implement an np solution trying to get the perfect score. Implement a fast solution that gets within x% of optimal

* If a solution seems impossible or it takes too long, return the closes solution you can find, and a warning about the solution being suboptimal

I got the code in a few minutes. On a sample of random inputs, the algorithm produces a solution within 1% of optimal in ~99.9% of the cases. p95 execution time is well below 2ms in my laptop.

That's it, that's everything you need for a production system. "close enough" very fast is sufficient, and the impossible cases very rarely happen. Even when they do, you can simply work around them.

Mostly yes. But there were also some take home assignments, some culture fit loops, and many ML rounds, since I was applying for an ML role.

My last job search 2 years ago took me 9 months of grinding while working full time at Big Tech and +15 years of experience. After loops with ~30 companies I ended up with about 6 offers.

The offers were all over the shop. Some places downleveled me to mid level while others upleveled me. The comp range was about 3x between the lowest and the highest.

Also I couldn't find much of a pattern on the processes and outcomes. No name companies rejected my application without even a phone screen. Lower tier companies offered me lower levels while famously hard to crack ones offered me higher levels. I also didn't see much correlation on how hard interviews were.

I'm totally convinced the job market is a numbers game and there is a high degree of randomness and luck involved. Discipline and hard work help, but outcomes are far from guaranteed. Last time I almost didn't leave my job since most offers were worse than my then current one, until I got 2 high quality offers at the end.

I have only skimmed through the article but I'm impressed OP got a job as a Research Engineer with no formal background, given that I got doors slammed even after PhD + years of relevant experience. I don't want to diminish the accomplishment, but I would be very wary of anecdata and formulas for success. OP quit his job and focused 100% on the search, and it paid off. I know many more similar stories that didn't pan out, and people had to rush to get lower quality jobs or, in the worst cases, got into long term unemployment.

Just my 2 cents.

I got a Bachelor, Master, and PhD in Computer Science, with a total of 11 years of education. It's the biggest waste of time of my entire life.

As I progress in my professional career I'm more convinced that pretty much everything in tech is on-the-job learning, and universities are little more than a social club. Nowadays you can learn everything you do at university and far more online and for free.

Universities (elite ones particularly) still give you credentials that have some value getting a job. However I wonder for how long that will still be true. Learning by doing and building a portfolio sounds like a better way of getting in the industry today than getting a multi-year degree with nothing or little to show for it.

Nowadays I wouldn't recommend anyone to get a tech degree in a university unless it's a world class one. And even then, I would focus on networking and finding like-minded people rather than necessarily getting good grades.

> This exact thing is what software developers have been begging for since the beginning of the profession: Receiving a detailed outline of the problem and what the end result should look like.

> This is often the part that slows down software development. Trying to figure out what a vague, title only, feature request actually means.

But that is exactly what Software Engineering is!. It's 2026 and the notion that you can get detailed enough requirements and specifications that you can one-shot a perfect solution needs to die.

In my experience AI has made us able to iterate on features or ideas much faster. Now most of the friction comes from alignment and coordination with other teams. My take is that to accelerate processes we should reduce coordination overhead and empower individuals and teams to make decisions and execute on them.

This sounds like surface level wisdom if you are in your earlier career. I see several problems with this advice, but here is the most obvious one: this only works if you want to become a manager.

It used to be the case that the only way for engineers to advance their career. But we've long moved since and now you can have a long career and get very high in a company without management responsibilities. The examples given in the blog post are exactly what I would expect ICs to do, not managers.

Do you want advice to get promoted? if your company has a formal career ladder, look into the process and optimize for it. Despite people grievances, this is still the fastest and easiest method to get a promotion (shocker!).

As a european living in the US, the idea there is a market at all is laughable. I tried to get price quotes for treatments several time just to get a "well, it's hard to say" or "it's very complicated".

I'm hispanic and my two last names are Garcia Garcia. That is two last names that just happen to be the same.

When I moved to the US I could have dropped one or hyphenate them. I decided to keep it as-is, and use "Garcia Garcia" as my last name (space and all).

Besides confusing amongs americans and people always confusing one Garcia for middle name and one for last name, I had almost no problems. One time an airline messed up my plane ticket (again by dropping one of the Garcias) but that's it.

I appreciate other people have different experiences, I definitely met folks who have changed their names to conform to american customs and make things easier.

I left my home country over 10 years ago, and ever since I've travelled back once every 1 or 2 years.

Since 4-5 years ago I started to notice these betting houses cropping up where my family and friends live. They are impossible to miss, with big pictures of different sports and no windows.

The most important thing to notice is where these place are and are not. They proliferate in working class and less well off neighborhoods, while they tend to be absent from more affluent ones.

These places get a lot of foot traffic, all the locals barely making ends meet, blowing a few tens of euros here and there, with the eventual payoff. It's not difficult to hear stories of people getting into the deep end and developing a real addiction with devastating consequences.

And it's not only the business itself, but what they attract. All sort of sketchy characters frequent these places, and tend to attract drugs, violence...

Legal or not these places make the communities they inhabit worse, not better. I personally would be very happy if family didn't have to live exposed to them.

The argument that the article uses to critique AI coding is that coding is not the bottleneck, instead the testing and validation of ideas is.

This is a first-degree smart argument. It presents a seemingly non-obvious idea that makes sense in retrospect.

However I happen to work at the experimentation team of a hyperscaler so I have a different perspective.

First, we aren't always saturating all of the potential experiments we could be running. The reasons are different, but essentially it takes time and effort to build those experimental features. If that cost trended to 0, we could make sure to have a queue of experiments deep enough.

Also in our side we need to do development work to support new features and products. We have a backlog long enough to keep us perpetually busy. If dev cost trended to 0, we could always be ready to provide our customers what they need.

Speaking of new products, each one our company comes up with comes with extra effort to support in our side, and yet more effort to produce dozens of AB test to validate new functionality.

This is not talking about ongoing maintenance effort. Bugfixes, upgrades etc. take a non-trivial amount of effort to keep up.

And this is only inside our little experimentation team. What about security, reliability, scalability, efficiency... it makes me wonder if OP has experience running products at scale.

Instead I'd like to think that dropping the cost of development by orders of magnitude changes the equation of how we create products.

We drove from Seattle to SF and back on a Model 3, and I would say there were only a couple of situations where I wish we could keep driving instead of recharging. Otherwise almost all breaks for charging aligned with breaks to go to the toilet, eat, walk the dog, or simply have a rest.

Definitely a very minor inconvenience. And, compared with frequent trips to the gas station vs charging at home overnight, a total net positive in time saved (if you want to measure it that way).

I also thought this was a pretty silly way of bringing up range anxiety. It's like a dealership selling you a gas car with an empty tank. Obviously the very first thing you would do is to refill it before driving home.

This isn't discussed enough. One argument I've heard is that "this only applies to people who break the law".

One thing to consider is how easy is to make minor mistakes that technically count as an infraction. When acting in good faith, the administration can acknowledge this and promptly fix it, as it happened to me during my immigration process.

Then there are random mistakes out of your control. For example, when I first moved to the US and tried to get insurance for my car, I received extremely high quotes from the insurer. When I inquired why, they replied that my file showed several traffic infractions years ago in a different state. Simply clarifying that they'd mistaken me for another person was enough to fix it. Imagine if instead they deported me to a prison in El Salvador without a chance to defend myself.

And this is not talking shadier practices, such as changing the rules so that certain things suddenly become offenses, or simply fabricating evidence against someone.