Exactly. We don't worry about kids who torture dolls because we're concerned for the dolls.
3,179 karma · joined April 28, 2014
Consider how trivial it would be to write a program that would destroy every computer on the world. Any competent developer could write this in a hour. An LLM in 5 minutes. The trick is actually getting the code onto every computer in the world and running it! Until someone grants it agency and runs it IRL, the artifact has zero impact.
The models can produce whatever scary text they want. But if a human acts on it, connects that with their email or their drone weapon, it's the human who bears total responsibility. We desperately need some legislation to enforce this; otherwise I see a future where almost any accountability can be avoided by AI-washing the problem.
Instead of prompting, I hack on code in my editor.
My harness gather everything it needs from the local context - my git diff, my open editor buffers, etc. to assess what I've been doing. No chat. This is fed into phase 2 which tries to guess my intent. Then phase 3, it presents a plan to complete the work. The only user interaction is reviewing the plan and typing yes or no.
The quality of the plan of course depends on the quality of my uncommitted ideas. As it should be. If the plan goes off the rails, it's my fault. Do not chat your way to a solution! Abort the session and continue fleshing out the idea in source code/markdown.
The reason I like this is it forces me to at least take a stab at the work. I treat the AI like a relief pitcher to come in and close out the game.
I can't say this is the way to hyper productivity. I'm still slow. But at least I'm spending exactly 0 hours a day arguing with an LLM!
"The Strange Last Voyage of Donald Crowhurst" - This story is about the first round-the-world solo yacht race, and the one contestant who died at sea. His story is really strange - an anti-hero tragedy with a touch of mental illness. The author does a great job bringing the reader on this descent into madness.
"Protocols" - Andrew Huberman's recent book. Like his podcast, it's full of solid advice backed by published science - but also with some repetitive filler and product name drops. The sections on exercise and diet were not very informative to me, mainly because I've already incorporated most of that into my life already. The sections on light exposure and its relationship to sleep, melatonin and cortisol is the most fascinating new-to-me info. I've started morning/evening walking without my sunglasses to get those sweet sweet photons.
> Some 2.3 billion people are current drinkers. Alcohol is consumed by more than half of the population in only three WHO regions – the Americas, Europe and Western Pacific.
Globally, the majority of people are non-drinkers. It's not even close. The maps obviously tell a different story on a per-country basis. And if you live in an area with a super-majority of drinkers - yeah it's going to feel like everyone in the world drinks. But the global data says otherwise.
So to answer the original question, "Where does all this supposed productivity go?": Sitting in a github repo somewhere, undeployed.
Needing alcohol to socialize is not normal (though it has been normalized by some cultures). If you need alcohol to have fun, that's a sign of a serious problem that requires professional help.
It used to be a universal social lubricant, but now people stay home - the only people drinking in public are drunkards. Not a great look.
It used to be considered "healthy" to drink a bit. Twenty-one drinks a week was considered normal. Now we have decades of research meta-analysis concluding, in no uncertain terms, that the only safe amount of alcohol is zero. https://hsph.harvard.edu/news/alcohol-is-the-root-of-62-dise...
Alcohol increases anxiety. Given the zeitgeist and state of the world, it's not surprising consumers want to avoid spiking their cortisol in the evenings.
Alcohol is expensive. The cost of living is going up and wages are not. The budget has to adjust somehow - alcohol is an obvious place to make cuts.
Alcohol prevents the body from metabolizing real food. Fat burning stops while the metabolism shifts to rid the body of the toxins you just dumped in. Aside from the health impacts, you're gonna see this directly in your gut circumference via visceral fat. Considering the GLP-1 trend, beer gut is out of fashion.
So for me personally, and anyone else paying attention, there is no reason to touch alcohol. It's strictly poison and has no valid role in a human diet. Sorry farmers - maybe switch to growing something that provides nutrition and isn't toxic?
I'm trying to say that making such claims in the first place is invalid.
There is no language that is objectively more or less readable. Yes, I've read the studies. No, none of them come close to adequately addressing the confounding factor of prior exposure/education. A randomized controlled trial starting from childhood could establish such truths, but such an experiment has not been done. Until then, small-n studies that don't address this flaw in any way yet continue to make broad claims - "pop" science is perhaps not dismissive enough.
Cynically, this might be the real reason managers and investors love AI. It diffuses responsibility. No one is accountable. "Oops the AI messed it up" is a convenient excuse for bad management.
Difficulty is, by definition, relative to one's skill. You cannot so quickly discount the fact that 99% of programming is taught in Java/C style language syntax. If you've had lifelong exposure to Lisp, you might feel exactly the opposite. The author does a poor job of justifying why these pop-cognitive-psych theories should have more weight than prior exposure.
Personally, as someone with decades of exposure to both styles, I look at the factorial example and see everything I love about Lisp syntax - consistent, no magic keywords and syntax to memorize, it represents a tree just like my mental model of code, there's no way to fall through and forget an else, expressions instead of statements, no early returns ... literally everything about the Lisp example is more readable to me. YMMV.
I've heard that one before. This one is the last pedal, I swear :-)
What about a slightly different angle: a custom language runtime? The language itself stays true to the original but you build your own custom compiler and dev tooling: an expanded stdlib, LSP, linting, formatting rules, build systems, test runners, package manger, host extension system, etc.
This is becoming somewhat of a reality in the Clojure world. https://clojure.cc/dialects/ lists ~30 languages which are recognized as "Clojure" but have radically different host environments. Of course Clojure has macros too so nothings stopping you from going overboard on the DSL weirdness.
I could see the following scenario: Your company picks Typescript. You evaluate Bun and Deno and others but nothing really works. You take the most promising one, build a little test suite to make sure it stays consistent with the language spec, fork it, and add the runtime features you need. Your dev team still writes Typescript but you have full control over the tooling and how that code works at runtime.
In the worst case bugs, systems can hum along for years with silent consistency problems. Database columns that are assumed to be unique but aren't - the truth only shakes out when you CREATE INDEX. Then once you fix that, you've got to find why the app was doing it in the first place! Generally its better for the app to crash than to silently corrupt the database as it had been doing all along.
Your comments re: database state are spot on. DDL can fail in subtle ways. It's not even enough to take a snapshot of the current state and validate; things can change under your feet.
Take adding a unique index on a column: a simple CREATE UNIQUE INDEX statement, right? But you realize it will fail if the values aren't unique already, so you run a SELECT query to confirm. Yep, all unique. Deploy the app which runs the migration on startup - fail. A non-unique key arrived in the time between your queries.
Even more fun if you CREATE UNIQUE INDEX CONCURRENTLY and a non-unique key arrives in the middle of the DDL execution.
It's not just obvious who's responsible, it's obvious who benefits from pedaling the "rogue AI" narrative.
Perhaps because they are dumber, they produce better results? IMO an excellent well-tuned harness combined with a "frontier minus x" model produces the highest quality result. DS4.1 and Qwen3.8, far from being a compromise, legit give me better results. For my personal definition of "better".
Github has reported a 14x increase in code. Where did that 14x increase go? Certainly not to providing economic value - as you explained, we've basically flatlined there. It doesn't show up in revenue or profit margins. It doesn't show up in the app store metrics. It doesn't show up in speed or quality or security. It doesn't lead to new innovations or breakthroughs in software problems. All AI coding has done is to create more code.
The disconnect between 14xing code and barely 1.0xing software value is stark. It's not really an indicment on AI coding though. I think it reveals something interesting about the software industry. Specifically, that it was never about the code at all. It's about making that code do tricks in the real world - that's what constitutes almost the entirety of the value proposition. Code is cheap and getting cheaper. Software remains hard and is getting harder.
Not to mention the asphalt roads we drive them on! Though that cost is shared by all drivers, it has to be accounted for somewhere.
Even if we have a 100% electric vehicle fleet, we'd still need oil to make roads and tires. Plus the 88g from the study. Since EVs are being sold as a way to "solve climate change", this is a big problem. Most people assume a transition to EVs would mean stopping oil. They are wrong.
What about when repetitive typing is no longer a constraint? Do we need to pay for those abstractions? An LLM can scour the codebase and repeat patterns without getting tired. A simple-but-repetitive codebase might be ideal for an LLM.
This is one place I see AI coding changing the definition of code quality itself. I'm sure there are more...
Code is just sitting in a git repo somewhere, not necessarily running. That's a big distinction for me. Consider that code volume has increased 14x on github this year, but we see nowhere near that increase in the actual usable software. Code is cheap and getting cheaper. Running software ain't.
Another way to put it: I'm only interested in code so far as the value it provides. That value, not the code itself, is the source of pride. If I can provide similar value without any code at all, I'd gladly do so.
My reason: code can be checked objectively. I can run it and confirm it works. I don't get attached to it. I don't feel pride in it (even when I write it by hand). Code just is. It's lifeless, inert, and entirely replaceable.
How do I do the equivalent for prose? How can I tell if my words "work"? Do they clearly convey my ideas to the intended audience? There's an element of subjectivity here forces me to identify personally with the prose.
Code has no such personality. I don't tie my identity or ego to code the same way I would an essay.
Finally, responsible journalism.