YHK - yet another hacker news clone

Cool story bro, but it sounds like you are trying to convince yourself.

See you back here in a month, or so.

I haven't touched a line of code in over 6 months, didn't even look at it, but the work I do is faster and higher quality than ever.

Not sure how I feel about it

- Try to prove the concept early on. When you start spinning in circles, then it might be time for a different approach

- I’ve had a lot of success with projects that have not started out as LLM codebases. And adding features one by one and testing really carefully by hand

- UI generated by LLMs are overly verbose. The key lies in the iteration. What you can cut, move or approach differently. But you have to have taste.

This is pretty spot on. But it's important to note that the resulting code, is still often significantly better than what those developers used to produce. Especially under tight deadlines and pressure to complete CRUD like interfaces. We simply can't afford spending a whole day tweaking a sortable table on the web anymore.

- Why should I even start vibing up my desired tools and projects when I can just wait it out so a better clanker emerges 6 months down the line?

- What guarantee is there that Rust is the best programming language for vibe coding when the models could improve to directly output assembly? What incentive do you have in vibing up something?

The environmental aspects can be solved.

Quality is a measure of the person driving the process, models can write code better than the average engineer.

The impact on learning, seniority etc is the one big unknown - but I believe can/will also be solved.

I work for a FAANG, we have unlimited AI credits to be used on any model we want, with basically any tooling we want.

My whole team massively uses AI, I barely do. I use not even 10% of the tokens that they do. Yet I have not only the highest rate of PRs raised, but also the highest rate of PRs I reviewed, and the second lowest number of revisions per PR.

AI is all about fake productivity. You feel productive while you're achieving the same, if not less (or even often, nothing at all).

Long term it's pretty obvious that people heavily using AI will become a lot less productive, if not simply unable to work as software engineers. My colleagues are already scared.

In fact, as a challenge to anyone who says they "haven't written a line of code by hand in X amount of time", go do a test. Try to write any code without any LLM use for just 30 minutes.

We now have people commonly running multiple concurrent programming tasks, while automatically and interactively testing them, reviewing them, and validating them. They're then handed for human review, where little is found because the quality regime is now set so high. We're talking weeks of development + testing/validating time brought down into the realm of days.

So this is pure "my horse & cart are just as good as the motorcar"-levels of argument. If I bury my head in the sand long enough, and deep enough, this will all just blow over, and we'll go back to doing stuff significantly slower/worse any day now...

The quality of stuff I am shipping today is the highest it has ever been, and tested/validated significantly more comprehensively. Why? Because time/resources was the limitation, not willingness or skill. I always wanted to test for every single possible edge case/scenario, but it wasn't in the time-budget (and diminish returns). Now I don't need to make that trade-off.

At work I still read through every line in the diff (unless it's a personal util/non-production facing/critical etc). Even with this I feel less 'in touch' with my work code than I do with my home code.

LLMs are tools like motor vehicles. They allow you to get somewhere, quickly but brutishly. You won't be able to appreciate the beautiful scenery much, as it rushes by. You're going to have to compensate for the exercise you would have gotten, had you walked, or it's going to be unhealthy. It's undeniably bad for the environment (electric or not) and has other negative externalities. The politics of many of the countries supplying most motor vehicle fuel to the world are beyond fascist by modern leftist standards. Yet, we depend on them, we have constructed a lot of the infrastructure we need to live around their existence, and even if every city became walkable over night, the last mile for most transport would be covered by some motor vehicle. There are many "what ifs" surrounding motor vehicles where the answer boils down to: "Well, then we're fucked". But we're not fucked today, and so on it goes.

Of course you can try not to partake in motor vehicles, and insofar as you don't make a moral grandstand about it, that's totally fine with me. But the rest of the world is in it for the ride. The first cars weren't very good either.

If you're afraid of being called a luddite, you've already lost.

You shouldn't value the opinion of those who would insult you for not wanting to offload something you find intellectually fulfilling to a machine.

You can still manually type code, it can be a great hobby, but it is unlikely that someone will pay you a full time salary for doing it

The only arguments against AI itself are about cognitive decline and “I like coding” While there is nothing wrong with those preferences, it’s akin to saying “I like building furniture with only hand tools, and not doing it by hand leads to decline in your ability.”

The post sounds more like a programmer at heart, and that's totally fine.

To me, being an entrepreneur is more about building products, solving problems, and creating business value. LLMs help to deliver bug fixes, new features, and entirely new products faster.

Do I write less and less code myself? Yes. Do I read less and less code? Maybe. Does it always matter? No. Sometimes, the agent writes code in a project or section of code that is sensitive and deserves a thorough review. Sometimes, it's in a tool that you would never have had the time to write if you didn't have an agent. Or it can be in a project that is so well covered by TDD that you can be confident that it won't break. There could be performance regressions or duplication / code creep, but once you realize it, you can ask an agent to simplify the code and find ways to prevent that from happening again.

There are cleanup / technical-debt-reduction tasks that I thought would never be done in my lifetime due to lack of time and other priorities. But now, I'm blasting through them with Claude. I cannot imagine the lack of productivity I would suffer if deprived from an agent.

That said, I agree with most points the author makes. It becomes harder to grok the code, just as it becomes hard to navigate without GPS navigation once you are used to it. It is maybe necessary to force oneself to do a development task once in a while without the help of an agent in order to stay sharp.

But stop using LLMs? that sounds counterproductive.

There is no going back to writing code the old way. I don't see that happening in the industry, ever.

The grievances listed here are 100% real, but not fundamental - with some practice you can prompt an LLM to great effect without it taking the reins from you. If you let it write the code, it does become a management job more than an engineering job. That said, I do believe that there must be a way to structure your workflow such that the LLM becomes your read-only pair programmer/rubber duck while you yourself get to do the fun parts of coding.

I got a quick prototype of a rabbet box generator:

https://studio.flet.dev/run/bicQWc3Aul/

but the geometry was all wrong and ultimately required that I describe the calculation of the position and dimensions of each element in exhaustive detail in order to get a usable DXF (the wireframe preview is still all messed up) which eventually worked to make a box:

https://community.carbide3d.com/uploads/default/original/3X/...

I was also drawing the design up in an interactive CAD/CAM program, and modeling it in BlockSCAD/OpenSCAD (and dropping into BelfrySCAD to figure out rotation):

https://www.blockscad3d.com/community/projects/2169316

(and speaking of rotation, at this time, the sliding and hinged lid options open on opposite sides in the preview...)

Since then, I've made a very simple generator which simple does a box decorated with lines, arcs, and circles on each side --- my intention is to dis-assemble that and use it to write a Domain Specific Language implementation for creating such designs and previewing them and writing out DXFs --- which is a long-winded way of saying I've been finding this tooling frustrating to use, but potentially useful to make rough drafts of programming templates which can then be used as a starting point for more traditional coding.

Eventually all of this will go into:

https://github.com/WillAdams/gcodepreview

I wholeheartedly believe that if the general trend in SWE goes without hiring new grads/self-taught people (who can still use AI-assistance in their work), we're gonna be seeing huge outsourcing to AIaaS companies and generally become tremendously dependent on them when existing pool of seasoned programmers goes extinct and the work doesn't intrigue training new people.

Although there is cheap, open source models running locally or available at scale on reasonable prices without compromising programmer integrity.

The brittle slop happens when you tack feature on top of feature because it’s fast and easy now.

The higher quality stuff happens when you dive into the details of what you/the agent implemented (looking at code, with agent, or not) trying to make the current feature set work better.

IMO there will be outsized competitive advantage to the latter. Features are now cheap. But quality remains rare.