Cleaning up after AI rockstar developers • Hacker News Board

Re: Cleaning up after AI rockstar developers

#351

Post by toobulkeh » Wed, Jun 10, 2026, 2:21 AM UTC

gizmo686 wrote: post #200

nrmitchi wrote: post #157

> Craftsmanship will always be in our hands, it's one thing we can never outsource to a machine. I'm right there with you, but this last sentence concerned me a bit. In my most other "industries", craftsmanship is not _dead_, but it's been pushed to the wayside for (significantly) cheaper and more available alternatives. You can still get hand-made leather shoes, but very few want to pay $1000+ for them. You can stil…

Craftsmanship is not dead in other industries in the same way it is being talked about for software. Sure, that cheap desk that arrived in a flat box and got assembled by me and a screwdriver was mass produced in a factory. But it's design had way more expert craftsmanship put into it than would ever be feasible for a bespoke product. High upfront design cost, then mass produced at a low marginal cost. That had been…

This is a great argument for software factories. Writing the systems that build the system.

Re: Cleaning up after AI rockstar developers

#352

Post by dyauspitr » Wed, Jun 10, 2026, 2:22 AM UTC

bjertoref wrote: post #301

dyauspitr wrote: post #282

Earlier quoted context omitted.

Yeah someone should train these LLMs on straight compiled machine code/binaries and the output so they can start writing directly in binary.

That is like running Photoshop reliably on an encrypted image.

Not necessarily. There is going to be a direct deterministic mapping to functionality in unencrypted raw binary.

Re: Cleaning up after AI rockstar developers

#353

Post by wgd » Wed, Jun 10, 2026, 2:22 AM UTC

this_user wrote: post #71

gyanchawdhary wrote: post #10

[flagged]

Compilers are deterministic and they actually possess domain knowledge of what they are trying to do. AI models are non-deterministic, have no real domain knowledge due to lack of an underlying world model, and their way of "writing" software is to spew out something that looks like something that they have been trained on, then iterate on it long enough until it has reached the level of being barely runnable.

People say "determinism" but I don't think that's actually the property we care about. For instance you could imagine a compiler that makes heavy use of superoptimization with random search and it would still have the ineffable quality that LLM codegen lacks. I think what we're actually trying to say is that the compiler preserves the formal semantics of the source language in its output, whereas English text doesn't have any such formal semantics to preserve.

Re: Cleaning up after AI rockstar developers

#354

Post by johnsmith1840 » Wed, Jun 10, 2026, 2:30 AM UTC

godelski wrote: post #345

pmontra wrote: post #341

Earlier quoted context omitted.

I probably had that kind of 10x moment yesterday. I spent 2.5 hours analysing a feature request and I eventually advised my customer to think very carefully about it: it might solve a problem but the implementation time is going to be long, even with an AI, because of actual development and the time we will spend in testing it. Furthermore the new infrastructure will have recurring costs higher than what it saves. My…

> 2.5 hours vs 25 or more. 10x. My point is that these 10x moments aren't sustained. When we're talking about a "10xer" we're talking about someone over a longer period of time. For "in the moment"s and short term projects, 10xers certainly exist. There's definitely things I can do in a week that would take juniors more than 10. But for bigger projects? Can I do in a month what a junior cannot in 10 months? Can I do…

You can but not the way you're saying.

Once was part of a team where a mid level guy spent a year building/heavy maintenance of some catastrophe of a solution. They literally had the end users copy and pasting hundreds of commands from a generated excel sheet of a command per core (shudders). I spent ~2wks on a different architecture that was a massive improvement we abandoned their code base. He quit like 4 months later to join some faang. Granted that 2wks of work was on top of a distributed cloud infra that took me 6 months to build.

So yes, a skilled dev might skip entire months of work someone else would make.

What you seem to be describing is a companies skilled engineer designs something and passes down the spec. The guy making the spec is the 10x guy. For large projects it's even more pronounced. The article literally described someone who wasn't skilled they simply knew how to smooze the MBA's and a company with poor engineering leadership.

Re: Cleaning up after AI rockstar developers

#355

Post by CodingWithJesse » Wed, Jun 10, 2026, 2:47 AM UTC

allknowingfrog wrote: post #63

Ithildin wrote: post #57

Like many other critiques along this line, I think the answer is: Yes, for now. For example, in my own code base, I have an agent dedicated to maintaining a sane data model. This agent reviews the plan in advance and again during code review. And then I review the code myself. This seems to be sufficient for my work, at least. YMMV. It was a good post though. I enjoyed the take and the writing.

If you're reviewing the code yourself, then I don't think this article is about you.

Exactly right.

Re: Cleaning up after AI rockstar developers

#356

Post by kaydub » Wed, Jun 10, 2026, 2:49 AM UTC

munksbeer wrote: post #324

kaydub wrote: post #272

Earlier quoted context omitted.

Not sure what I'm going to link to in my professional life, it's all private repos. We have a large microservice architecture in AWS. We've made major changes and the LLMs have helped every step of the way. Migrated our whole auth system from being regional, resulting in customers being limited to specific regions or managing their multiple accounts across regions, to a global authentication with data residency built…

I enjoyed hearing about the stuff you've done. Would you mind talking about your experience making the 2 godot games? Which agent and model did you use? I've been wanting to dabble here and interested to hear stories.

Got distracted and didn't touch on the godot specifics.

I start the project in godot, then I talk to an agent to define my game and give details. I tell it to have a conversation with the end goal to be building a CLAUDE/AGENTS/GEMINI markdown file for future agents to help build the game.

I made a [single player top down naval game](https://www.youtube.com/watch?v=FQUTbeQ8-oA) and an online multiplayer game (that I'd like to keep to myself for now).

I mostly used free assets for the artwork. I made some myself in aseprite. I utilized LLMs a tiny amount to generate some assets, but typically those are more of a jumping off point than game ready art assets. This has been what's slowed me down the most in making games.

Re: Cleaning up after AI rockstar developers

#357

Post by CodingWithJesse » Wed, Jun 10, 2026, 2:53 AM UTC

PeterStuer wrote: post #284

This article feels written by AI. This cascades of short sentences are a giveaway.

100% wasn't. I use Derek Sivers' approach, search "writing a sentence per line", and I try to keep my sentences short intentionally. If you view source on the HTML you can see the line breaks.

Re: Cleaning up after AI rockstar developers

#358

Post by CodingWithJesse » Wed, Jun 10, 2026, 3:02 AM UTC

rafram wrote: post #140

erichocean wrote: post #132

> Half the code was written in a language you didn't understand. The other half was written using libraries you never heard of. The author already describes himself as "not a rockstar developer", but if this is the definition of "rockstar" I need to recalibrate. Being able to learn new languages and libraries, to me, is completely normal. (Also: how funny is it to suggest rewriting code you self-admit you can't even…

The point is that developers like this purposefully introduce new languages and frameworks that only they (and maybe not even they) fully understand, so they make themselves harder to replace and ensure that they appear ahead of the curve. (Then they get bored and leave anyway.) It's totally valid to introduce new technologies when the problem calls for them, but doing it in a way that's purely detrimental to the res…

Exactly my point. Learning stuff is fun. But if you're going to introduce new technologies or programming languages into a system, you need to make sure it suits the skills and experience of the rest of the team. Rewriting the company's admin dashboard in Elm, not letting anyone else work on it, and then quitting - that's just cruel. (Elm's cool, but you know what I mean..)

Re: Cleaning up after AI rockstar developers

#359

Post by lelandfe » Wed, Jun 10, 2026, 3:09 AM UTC

bjt wrote: post #323

lelandfe wrote: post #269

Earlier quoted context omitted.

Going to be awfully hard to market rote machine maintenance as craftsmanship, but there are some suckers out there

If you've ever watched an episode of "How It's Made" and seen how incredibly customized these machines are, it won't be surprising that the people who build them are proud of their work.

My skepticism of the marketability of this idea deepens

Re: Cleaning up after AI rockstar developers

#360

Post by ytoawwhra92 » Wed, Jun 10, 2026, 3:13 AM UTC

godelski wrote: post #345

pmontra wrote: post #341

Earlier quoted context omitted.

I probably had that kind of 10x moment yesterday. I spent 2.5 hours analysing a feature request and I eventually advised my customer to think very carefully about it: it might solve a problem but the implementation time is going to be long, even with an AI, because of actual development and the time we will spend in testing it. Furthermore the new infrastructure will have recurring costs higher than what it saves. My…

> 2.5 hours vs 25 or more. 10x. My point is that these 10x moments aren't sustained. When we're talking about a "10xer" we're talking about someone over a longer period of time. For "in the moment"s and short term projects, 10xers certainly exist. There's definitely things I can do in a week that would take juniors more than 10. But for bigger projects? Can I do in a month what a junior cannot in 10 months? Can I do…

> Can I do in a month what a junior cannot in 10 months? Can I do in 6mo what a junior cannot do in 5 years?

I briefly worked at an organisation where I was consistently and sustainably able to ship in two month blocks what other teams of 3-6 engineers at that organisation could not successfully deliver at all. I would consider myself around 75th percentile productive compared to the industry, but in that specific organisation I was at least 10x as productive as the median engineer - regardless of their seniority.

I think engineers tend to form clusters where everyone is roughly as intelligent/competent/productive as each other. Outliers tend not to join the cluster, or they leave quickly. I've seen this happen at the level of a company, but also in larger engineering orgs at the level of a team or group. High performers don't stick around when their median colleague is a low performer and vice versa.

Perhaps you've had the good fortune to work mainly in organisations with a high competency floor. Looking around, you may not see anyone who's 10x as productive as anyone else, but maybe you're ignoring that everyone is in the 90th percentile of the industry.