Hacker News Reader

Comment by KillerRAK | Hacker News Reader

Without experience, programming with AI (vibe coding, I guess) can be compared to being a rat in maze... You work your way through a project, but the dead-ends exact a high cost in terms time, attention, and ultimately cost.

With experience, you see these dead ends before they have a chance to take hold and you know when and how to adjust course. It's literally like one poster said: coding with some buddies without ego and without the need to constantly talk people out of using the latest and greatest shiny objects/tools/frameworks.

I've really enjoyed going back a revisiting old ideas and projects with the help of AI. As the OP stated -- it has restored my energy and drive.

View on HN

I have always had ADHD and as a consequence have a decades long backlog of things that I want to do “some day”, and Claude just removes all the friction from going from idea to execution. I am also a software engineer, so basically for me it is like having a team of developers available 24 hours a day to build anything I want to design.

I have built and thrown away a half dozen projects ideas and gotten one into production at work in just the last few months.

I can build a POC for something in the time it would take me to explain to my coworkers what I even want. An MVP takes as long as what a POC used to take.

The thing that really unlocks stuff for me is how fast it is to make a cli/tui/web ui for things.

are you trying openclaw?

As a fellow ADHD’er who is also old and out of coding a decade, after decade and a half coding, wholeheartedly agreed. It’s great to just shit done and abandon if needed. Feels much better than spend 6 months and abandon

Claude Code has killed my ADHD and turned me into an always-on hyper-focused machine.

I am getting 20x done. This is a literal superpower.

I am not using it in agentic mode yet. I am telling it everything I want it to do. I will tell it where I want the files, what I want structs to be named, how I want the SQL queries to join, etc. I then review every line and make edits (typically with Claude first).

I haven't tried the agentic stuff yet, but I probably will at some point soon. I'm anxious about losing control over the architecture and data model, which is something I feel gives me my speed with Claude Code and that I know is important for my engineering work and quality.

I won't be writing code by hand ever again. This is the future. We'll look back at the old way as horse carriages.

Claude is also really freaking good at Rust, and the fact that it emits proper Rust with tests makes me even more confident of my changes.

We are literally living in the future now. Twenty years of SaaS and smartphone incrementalism and now we have jet packs.

Instead of engineers inventing 50 different frameworks and conventions for any given language or platform, maybe that energy will be directed to creating better AI tools.

Edit: I'll also reiterate what others are saying in that I think this is a tool best leveraged by engineers who know what they're doing and that care about code quality. The results you get back will also depend on your repo/project's code quality. If your project is poorly structured or has a lot of cruft, Claude will see that and spit it right back out. Keeping your code clean and low on tech debt is going to matter tremendously.

100% feel the same way, and had the same starting point lol

Careful though: a lot of people are getting the feeling of getting 20x done. Do you have objective measurements?

Ha ha, for some of us the "feeling" is good enough.

this is the most self-aware, honest answer IMO

IDK just yesterday I got a complete slide / powerpoint-lite editor in Qt Quick that is sufficient for the use case I have in two prompts, roughly 7 minutes. How long would it take you to write, on your best day, using your favourite programming language ?

For me the evidence is I have completed side projects I never would have before. I also recently started building a game that I had put off for years. At work I am closing more features than historically and at the end of the day not as fatigued. It’s only my experience and everyone’s is going to be different.

In my case is nearer to ∞x. I have developed an opensource Android app which has already ~200 users that I would never ever written in my whole life. Zero experience with mobile development and zero time in my free time to focus on this appropriately to at least try to learn how to do it. I know myself, I would have given up before getting the first dummy APK on my phone. And while it's totally vibe-coded, in the sense that I just prompted CC and not written a single line of Kotlin code, I put a ton of effort on it anyway, on how I want it to behave, how it looks like, squashing all the usual subtle bugs that CC leaves here and there.

Did you prompt it step by step or let it do its thing for a bit?

I've been working on it since December. The first working prototype was a oneshot of a biggish enough prompt (for Opus 4.5). After that well, I didn't save the prompts (my bad) but I have probably prompted what... 2-3k 80-columns lines of English in Claude Code? Yeah, I guess we are in that ballpark. Sometimes it nails the new feature at first attempt, sometimes it takes a few attempts and corrections (and in that case it can definitely be frustrating)

What does it do?

(I'm going to break the rule I had to keep my HN identity separated from my real one with this post but here it goes...)

It's a frontend to the vehicle data served by TeslaMate (a local logger for your EV data), as a more mobile-friendly alternative to Grafana.

How do you even begin to define objective measurements of software engineering productivity? You could use DORA metrics [1] which are about how effectively software is delivered. Or you could use the SPACE Framework [2] which is more about the developer experience.

1. https://cloud.google.com/blog/products/devops-sre/using-the-...

2. https://space-framework.com/

I don't have time for that mysticism. I just know.

>Instead of engineers inventing 50 different frameworks and conventions for any given language or platform, maybe that energy will be directed to creating better AI tools.

I think this will happen since one of the reason for new frameworks and languages was improving the human experience of coding, but now that friction goes away and AI doesn't feel that.

Although we might need to study which language AI is best at, and possibly invent new ones to maximize that.

> Claude Code has killed my ADHD and turned me into an always-on hyper-focused machine.

> I am getting 20x done. This is a literal superpower.

Adding this comment to favourites to revisit in half a decade.

I've already "made fun" of your exaggerated hype comments, so I'll use this opportunity to say that I hope you remain sane and grounded in your discoveries. You wouldn't be the first to go psychotic after interacting with these stochastic parrots.

Don't you have anything better to do with your time?

I told you people back in 2019 that these models would replace Hollywood and you and others have been calling me all kinds of names, and every step of the way calling me an idiot. I'm a filmmaker - I know what I'm talking about. And now we're almost here. We have million dollar VFX services at our disposal for pennies.

Claude Code is doing the exact same thing for software engineering. I've been a senior software engineer for a good while - these capabilities are otherworldly and they can generalize to all new unseen problems. You're not paying attention.

I'd be more worried about whether or not you have a job in 5 years than whether I have or have not created a business or whatever criteria you want to use to thumb your nose at me.

You know how you can quickly ideate software plans for some large scale idea? Architecture, infrastructure, data models, etc., but the implementation takes longer? Claude Code short circuits that last bit. You need to hold your nose so you stop smelling whatever you're smelling and just try the damn tool.

I wish I could slap sense into you grumpy folks. You're so stiff in your beliefs. This is a train headed your way. Pay attention.

> I told you people back in 2019 that these models would replace Hollywood

What kind of alternate reality are you living in?

I wish you would disclose your credentials (though I admit privacy is an inalienable right of yours) so I could place the biggest AI hype-man on this forum. Actually, there is hype, and there is being completely gone with hubris and you’re towards the latter end of the spectrum, given your doomsday calls on other comments that software engineering is done for and that you believe AI is close to ‘putting all the HN engineers out of work’ (https://news.ycombinator.com/item?id=47185284)

> I wish I could slap sense into you grumpy folks. You're so stiff in your beliefs. This is a train headed your way. Pay attention.

Lay off the violent thoughts and get some rest, man. Sounds like you need it.

part of my adhd/bipolar is reacting like the guy you're replying to and I was thinking the same. Comment reminded me of when I'm in the "YES THIS IS IT" mode which usually isn't far off from hitting the wall. Hopefully just projection on my part, though, and this guy is really doing well. When I start talking like him though I usually have to take a step back and it'll be a topic in therapy next session.

Fully agree: I believe my decades of software engineering experience definitely help me fly LLM tools better than less experienced folks.

But the much more interesting question to me: as LLM coding becomes the norm, does it drive the cost of self or small-company generated software to 0?

Like many SW architects/engineers my not-so-developed work-in-retirement plan is to assemble a small team of people I’ve loved working with over the years, start an LLC, and try to make a reasonable (not posh) living doing what we love: making software to solve problems.

On the one hand, it’s clear LLM coding can accelerate and amplify our efforts, but alternately there’s many people claiming there’s no possibility of a moat, your solution/innovation can be cloned in a matter of days … ie. the value of your software is exactly 0.

Not sure which future will be closer to reality. A backup plan that seems reasonable in the 0-value case is to focus our effort on creating actual physical gadgets and systems in the embedded realm, which conceivably can be designed and prototyped by a small team… It seems like these would still be valuable.

Reply on news.ycombinator.com