cadamsdotcom

Full disclaimer I am a huge fan of Linear's philosophy and products.

What I suspect Linear doesn't notice; the thing that makes their lives easier; the David Foster Wallace water they swim in everyday - is they're their own customer.

Most software makers have to translate requirements between domains where terminology is incompatible, dogfood to find the micro pain that helps a product become great, etc. but Linear can just... track the build of Linear in Linear.

When the someone behind the software is its customer they'll get loads of judgment calls right first time. Great!

If you aren't your own customer but plan to leverage intuition, plan on being wrong lots- and keep your company and architecture and products compatible with frequent wrongness so once discovered you can pivot to the right thing. No ego.

I agree the people needed goes down - but will add a nuance.

Roles go away and fewer people are needed to do a thing, but I think it's good! For example the number of people on farms decreased orders of magnitude over the past say, century - yet we still feed everyone. Farming got mostly automated and got insanely more productive per human.

What happened was those people who would've been needed on farms went and did other stuff and the economy got way more complex but also that new complexity let a lot of more awesome things got provided for a lot more people by a lot more companies, a trend that continues today.

If you want to be a farmer good luck; it's harder to get started than it was say, 50 years ago - so why bother when there is no net new demand for farmers and there are also many better ways to make a living.

That's what'll happen for cognitively drudgerous work vs. creative new work.

AI needs to be instructed to do a thing. To use the "steam engine designing another steam engine" analogy, why do the big labs give the highest pay of all to their prized researchers? Why do they take steps to prevent poaching? Surely a non-expert could prompt a model "find a superparameter combo that makes an amazing model" and the next gen model would emerge! The AI should be able to design a better AI on its own!

Why do the labs have thousands of employees? Lets think about this...

Who's designing the evals needed to validate the model quality, who's deciding what goes in the system prompt? Who's negotiating to buy training data from hundreds or even thousands of data vendors? Who's responding when the Internet goes apeshit over some prompt tweak that makes models over-verbose? Who's out there listening to users and feeding what they say back into the product pipeline?

Computers????

However automated things get, that quickly becomes the new baseline. Customer expectations are always going up.

And as soon as humans figure out how to automate any of the things listed above - it becomes part of the new baseline too. It's always advancing. To beat the baseline and be competitive, you need humans. No matter what. No matter how automatic the automated things get. Always and forever. Because you need someone to drive, to find the next automatable thing and automate it - and to prioritize where to search for automatable things.

No matter whether AI caps out or when, humans will be driving.

That said, other humans can and will put AI to use in ways that don't benefit humanity. It's actually already happening - newsfeeds waste hours of people's lives and attention; give teens low self esteem etc etc. and guess what. Society is adjusting - governments via legislation, device makers via screen time limits, communities by spreading the word and raising awareness

Smart AI wielded by people without your best interests in mind may be tough to counter; but you don't have to do it alone: humanity is getting wise to the threat. We don't have great tools but it'd be against the grain of history to assume we won't come up with them in time. Can't give specifics - as they say, predictions are hard, especially about the future..

We've made it through bigger more violent upheavals and we'll make it through this.

A couple of thoughts in agreement..

First, "cognitive labor" is a form of labor - it's different vs. creative work, even though both use your mind. If your math theorem can be written up in partnership with an LLM, wow! You can be done with that, and go do other great work. Cognitive labor is a form of labor. Let's talk about physical labor and leverage to understand how removing labor translates to time. On your farm you buy a tractor to replace your backhoe. If it is a hobby farm you get in the tractor and you're done super fast. Or, you can take that tractor, buy a bigger farm, and spend the same amount of time you'd have spent backhoeing, but in a far more leveraged way. You can trade removed labor for time in a way that suits you.

Second; the joyful moments we get from finding generalizable abstractions and nice solutions are an adaptation. Humans can climb infinitely high conceptual hierarchies - we'll never stop climbing, using the previous ones as tools to find the next. As we do, we gain joy with each discovery because. We get so much done with brains that run on 20 watts - are we tuned to receive a neurochemical reward for finding cognitive efficiency? If AI means more concepts are discovered the tower gets taller - and we find those joyful moments more often. If there are infinite moments to be found, why not go as fast as possible!

Third; everyone else is doing it, it's a baseline skill. Knowing how to work without AI is like knowing how to ride a horse. Useful and fun! Bedsides - only 150ish years ago horses were the only viable transport. Cars were new; toys for the rich. But cars got so good so fast that we eventually reconfigured whole cities. Your homework is to consider why and how horses got disrupted.

New divisions are emerging in the lump of roles humans had at work. AI takes away roles from your job and lets us do more with less. That's identical to every mechanization before it. It's killing some roles that make up your work, even as it creates new ones.

There's so much to be excited about. Fear is a rational response to change. But so is hope :)

Sounds like there's a gap in the market for an honest player to fill.

You should go learn about test driven development and test first development in general.

It absolutely can be industrialized: the agent writes tests that conform to your requirements; the harness makes sure those tests pass before work can stop.

Have a look at https://codeleash.dev/docs/tdd-guard for my (dated but still interesting) take on this. Claude Code hooks are really useful to do the enforcement.

It's open source so no need to read my crappy generated docs - you can point your agent at the repo and ask it questions.

Numbers quoted without context.

What if it now competes among a greater number of countries, or the low is because other countries got better, not because the US got worse?

But hey, it drives the clicks!

The repetitive piece - the piece that needs that process-control - is moving from shipping code to its creation.

That's new. All analogies to 2022 software engineering need to be thrown out.

Software construction is now more like an industrial process, with the artisanal work shifting towards both

1. aligning what's built with what's needed (talking to customers!)

as well as

2. implementing and tuning your software construction process (ie. hi - we're Anthropic. We're burying ourselves in a mountain of code but it's okay because the fix is a second mountain of code, try some tokens the first hit's free!)

Back to my original point...

Writing code is now on the production line; making quality output demands industrial grade/scale verification (some involving humans of course).. which is a fascinating engineering challenge, and it's brand new!

Ha! Turns out AI-washing isn't just for layoffs!

I'm sure it works great for its creator - but, without benchmarks of quality vs. an agent that can grep, it's hard to know if this project adds value.

I disagree pretty strongly that all those lessons are trained into an LLM.

Not only does it not know everything you know, you don't know what it does and doesn't know, and can't find out thoroughly enough to know that you know what it does and doesn't know. So you can't cover all the bases you can cover by just using some code.

You also won't reliably get the lessons it does know out of it every time, because it's a non-deterministic system.

Worse, LLMs are being replaced all the time. The next generation might be trained forget stuff because memory efficiency and also, hey - it can just go look up online what it doesn't know about! Yep - that's definitely gonna be true until the end of time!

You would be blown away by this wild thing called determinism!

Let's talk about a lawnmower. Bought one the other day. Effective tool?

Nope! Now it's "automation". Why ?

Using a lawnmower on your lawn doesn't give you control over the grass it cuts! It just cuts willy nilly wherever you point it, you can't get into the pebbles of your garden bed and forget that curvy edge work you had put in last month. And that's not even the worst. A lawnmower denies you the chance to learn how different breeds of grass make up your lawn or how each breed responds to a scythe. You never get to spend achingly long stretches admiring how grass glints differently in the sun at different times of the day, nor how it smells when cut - you'll mostly smell two stroke. A lawnmower uses insane amounts of oil and gas compared to the simple silent pleasure of walking around swinging a scythe. Being done with your lawn after a short while denies you a backbreaking day in the sun and you might even avoid a sunburn. Besides when you're done mowing your lawn what are you gonna do with all the time you got back! Read a book!?

Lawnmowers are a scam!

Frameworks, linters, compilers.. they contain codified lessons you'd be silly to not leverage in your work - agentic or otherwise. Using what others have built before you helps you get there faster.

Making your own framework is great if you're in no hurry at all - maybe you want to learn, or maybe that's part of what you're selling. For everyone else, use the approach that gives you the most leverage.

You can lift stuff with a crane, or go to the gym and lift stuff for fun.

Both are useful! They're not mutually exclusive!

But thanks to technology you no longer need to lift heavy things to survive or live a good life.

So shall it become for the mind. All the banal, unpleasant "cognitive labor" people perform now - copywriting, TPS reports, documentation.. that's all gonna get automated.

You used to need to be strong to wield a sword and shield while wearing armor and go into battle and try to not die defending your city or country. Now thanks to a great heaving pile of innovation and quality of life improvements we've won over centuries, we can decide one day to compete in bodybuilding. Same muscle gains, less danger, way more fun. All the country defending is happening elsewhere. Maybe the same will happen for our minds.

Imagine a future of brain leisure - one where we can go compete in bodybuilding competitions for the mind! Perhaps, a live stream where people do cool brain things. Who would line up to compete to compose the most engaging poem - on the spot, no assistance - in some obscure language from a continent they weren't born on. How cool would that be to watch!

Uh oh.

Oh dear. Oops. Sorry team. I forgot we were doing the doom and gloom thing, only looking 1cm from our noses and seeing disaster.

Sorry we made you grind LeetCode.

Here's a slop bot to talk to instead.

Sounds like a massive evolutionary advantage in the hands of the new arrivals.

In tech we call it "disruption" and celebrate it.. but of course it's different as it's tech, and doesn't involve human populations being decimated.

Just interesting to consider the similarities in nature and the similar effects on who gets the spoils.

> You design a revolutionary new programming language: who cares? ... humans don’t write or even read the code anymore. The programming language is an implementation detail the humans no longer have to care about.

Not sure what this author is smoking..

Python 3.15 just came out with hard-won improvements to free threading and a ton of minor speedups; and thanks to incremental work over the past few releases strongly typed python is more possible than ever, letting legacy codebases be progressively retrofit. Meanwhile Rust and Go are innovating at a crazy pace to give humans AND robots amazing leverage - whole categories of risk are pre-mitigated by work done in the language - chief among them memory safety and sensible async primitives. Innovative decisions coded into the language from all three camps let things get built faster than ever before, and guess what - it's all being done by .. humans! In communities! Sharing ideas!

Lots of this stuff is being written about on blogs! By humans! Writing with keyboards for other humans.

AI hasn't affected this work; AI is a sideshow. You can do app dev with an agent, maybe, with a lot of work - but guess what, the most effective ways to use AI will only emerge from humans sharing ideas with one another.. on blogs.. in communities!

This author is an island, living in a constructed reality designed to suit their false narrative. Their article is chock full of wilful failure to see what's really happening. It is simply wrong, and loudly proudly so, and that's a real shame, because there is actually SO much good happening in software in 2026.

There's so much to glean from this for anyone who wants to create!

If you're working with an AI to make your thing, know that even though its first output looks great, it's nowhere near as far as you can go if you keep going, use it as a tool, and still put in the same time and energy you'd have done if you weren't augmented.

People can tell if you've vibed something and they can tell if you've cared enough to instead build it fast but still sweated the details.

The steps described are legally required before you can file a petition to hire a foreign worker.

Vance simply exploited the fact most people don't know that.

These people aren't holding up the Internet. The software they built is doing that.

It's like how your house isn't its blueprint.

Writing software isn't the same as running it. Software maintenance work doesn't scale with the size of the user base.

For software you don't operate yourself, the scaling factors have nothing to do with where it runs or why or when. You're isolated from the number of installs, since you don't operate those installs, you don't keep them updated or bring them back when they go down. That burden is borne by vendors.

Sensationalist takes are great for clicks, but there is no "there" there. If this was a true economic failure it'd have been fixed long ago.

My bet is these maintainers aren't in it for the money. After all they're not currently dead of starvation, so by definition they're okay. Not billionaires, but okay.

And that's okay! Not everyone wants to be a billionaire. Some people just wanna be that person in Nebraska thanklessly keeping the planet on axis.

Such creativity warms my heart.

It's a clever hack, which as hackers you should expect and respect.

Your beef is with the law, not tax-minimizing entities acting within it. Companies won't voluntarily pay more tax than legally required.

Wrangling agentic code is more like an industrial process than craftspersonship.

You need to treat the code that arrived in response to your prompt as a raw material: just like some truckload of unrefined and unmolded "stuff", it just arrived at your factory door, and it is still in need of banging into shape.

Your job is to build and operate the machinery (which is also software) that turns raw slop into worthy output.

Determine the tolerances (eg. quality tolerance: how bug-free must this be to ship?) and then work out how you'll scale up and remove yourself from as much of the attainment of those tolerances as possible, given constraints.

Consider by way of analogy an industrial food factory that mass produces, say, pickled fish. The fish are caught and brought to the factory - that's your agent's first-draft output. Next the fish are put by hand on a belt to be dipped, fired, cured, etc. - that mix of by-hand steps and automated steps is where you'd be crafting the custom linting steps, making the code compiled and having the agent write tests that prove quality, playing with the work (testing the feature, bugfix etc. by hand) and bringing it further from slop to the level of quality that's needed. The amount of this you do is determined by constraints - cost, time, etc.

Now that you're sizing your efforts against constraints and using tolerances as a tool, you're making the transition in your role, from craftsperson to engineer.

It's sort of like a factory. But it's never going to be a lights-out factory. You should design, supervise and improve the process in line with constraints and tolerances provided by stakeholders. And leave perfectionism at the door. You can chase your own definition of perfection in your hobby time, or if you want to get paid for perfection, start your own business where you call the shots.

> a huge change that doesn't diminish utility at all.

It does though, by redirecting funds & human effort that could be used for more lighting.

Here's a counter-anecdote though, to give you hope.

In Bondi in Sydney there's a lighting upgrade underway on the beach promenade; the plan was circulated so I got to take a look. As you hoped they indeed did deeply consider throw angles and what's lit & unlit, and ran a competitive tender to choose among many vendors, settling on one they lauded for their flexibility in offering products that can throw at specific angles and leave specific areas unlit. In the plan published for public comments they specify the exact angles needed with diagrams and everything. So it's possible! It requires public servants to go above and beyond, which does happen, just not all the time and not in all the places. But it does happen!

I think actually we're forgetting how economically valuable light is and how much lighting at night improves human lives.

While also - paradoxically - forgetting that there are plenty of dark places just outside your city if you're willing to drive to them.

Perhaps it's in service of baiting you to click.

Life is complex. You probably don't optimize your life in a way that some other person would be shocked you're not doing.

And they probably optimize their life in a way that's invisible to you.

My house has smart light switches. They use ancient technologies of injection molding and metal wires. Their firmware has not needed an update and I haven't had to reboot them; they also recover instantly from power failures. I haven't measure their boot time but it's impressive. Probably 1/50 seconds is my guess, based on the AC switching frequency. Their data collection is primitive, and the retention is only a few hundred nanoseconds. They're quite good about not selling data about when I'm home to insurance companies or advertisers. And they use ancient principles of physics to destroy data securely. Their registration and account creation process didn't store my home address in some vendor's questionably secured database, so it can't be sold when these vendors fold in 5-10 years. It doesn't stop working while I'm traveling and never leaves my family without working lights, in fact I've yet to get one tech support call for light switches.

Home Assistant won't read the state of them so they're lacking in integration, but the plus was that caused me to not set up Home Assistant. Not making a fragile computer a single point of failure and not tying my whole house to one shared-fate system with all the failure modes of a small scale full stack SaaS gave me a couple more nines of reliability.

One downside is the commute to the wall if you want to use them. Commuting to a light switch is actually good though, I'm pleased to report it's a nice stroll. It takes around the same time it takes to unlock my phone and discover my phone can't switch on the lights in my house. It goes past a window so it has views.

When I read some vendors' stuff was gonna stop working I went home and turned off and on all my light switches to simulate a reboot in solidarity for those whose hobby is to make their home less usable. Hopefully you folks are able to spend a few hundred bucks more to refresh your stuff for this decade. If you need me, I'll be over at the wall panel.

Very cool innovation in steering - but a lot of introspection only emerges at the highest weight classes - this research would be fascinating to run on bigger models.

If you're struggling to explain why a vibecoded app isn't a complete app, you could do worse than the analogy of a movie set.

A movie set only needs to look convincing enough to film one good take from one angle. After that it is usually (but not always) torn down.

It only needs to look real enough to fool the audience - whom have no control over camera or what happens on screen..

And if you really want to stretch the analogy, anything wrong with a movie set can be "fixed" in post.

To build the real thing the movie is faking is an endeavor with almost no crossover.

Holy crap that was dope. The music, the colors, the pacing.. just an absolute work of art by 2swap - and from the sounds of the end of the video, many many collaborators!

Thanks for sharing!