Ask HN: Is looking at the code slowing us down?

[rstagi]

I know this scares many people, but you can control things with QA, evals, ship quality gates and so on... I know we might have more bugs in the short term, but is not looking at the code the right direction we should go for the long term?

Several important developers switched to this new idea, the ones who work in production grade stuff (one example is Antirez, who just wrote some tweets that inspired this post)

[dlcarrier]

I work in hardware development, and without the ability to release updates after something has literally shipped, embedded software has to follow the the opposite philosophy.

I've found that the forced do-it-once-and-do-it-right philosophy requires so much less total development time than the break-things-and-release-often philosophy that I've adopted it on software that I could update after the fact.

The biggest impediment is that you may be relying on a software stack adopting a different philosophy. No matter how good your software is, if the underlying OS or libraries are buggy, there may still be poor reliability.

Using standard C libraries is a pretty safe bet, whereas relying on Python or Node.js libraries means you'll need to issue constant patches, even if your software works fine, and the fixed libraries might even break compatibility.

Given the option, running your software on OpenBSD can give extremely high uptime as critical updates are rare. They have a philosophy of not only looking at what they write, but looking at it again later, even when they haven't changed anything, just in case they made a mistake.

[mikgp]

Slowing us down from what?

In the context of many other measures I think speed is an important measure. Maybe even the key measure. I’ve even written a blog post entitled “reading code is an anti-pattern.”

But if the metric is, slowing us down from having more lines in production then the answer is unequivocally yes.

If the answer is providing customer value, or having a sustainable engineering culture. Then I’m not so sure. I’m not saying I believe the opposite, but it feels like excessively optimizing on the wrong thing.

Ultimately I think this asks the wrong question, I think most other surrounding questions is the right one which is - how do you safely and quickly deploy the right code to production that delivers customer value.

I think that will continue going forward involve doing so in an automated fashion, but then the right question isn’t, “should we stop reading code”. But something like “what is the right way to ship intent to production”

Because if you do the first without the second…

[rstagi]

Yes I meant delivering value ofc, and I love the framing you propose, so my question is: are we already at the point where we could go safely faster, if we didn’t look at the code?

[mikgp]

I think it’s a maturity model question, and an individual ability question. And maybe an industry tooling question.

Individual ability? Newer engineers should read and write more code than experienced engineers.

Maturity model? Teams with better access to objective specifications can read less code than teams without it. Teams with more experience with building loops and validation frameworks the same.

Industry tooling? I think we need more work on the spec driven development front. Some set of standards as to how we define contracts that LLMs can operate against that are more deterministic.

[verdverm]

I see significant decline in product quality this year among those companies most pilled (Ant, Goog, Msft, etc...). I had been commenting that Atlassian turned a page and Jira was running quite well, but this year... their internal push on Rovo has shown in their product, and their new `twg` CLI is some real slop where help text doesn't match the code.

[theandrewbailey]

Looking both ways before crossing the street slows you down too, but it's done for good reasons. Ignore reading code at your own peril.

[rstagi]

Great example! You usually don’t look both ways if you have a semaphor though, so maybe all we need is good semaphors. I hope the metaphor holds :)

[rvz]

> I know we might have more bugs in the short term, but is not looking at the code the right direction we should go for the long term?

Would you drive without looking at the road (or blindfolded) because we now have autonomous vehicles?

Not looking at where you are driving and also never looking or understanding what the code does, makes no sense in the long term. Especially in the event of a disaster.

Both have something called "liability" and it gets expensive when it goes wrong.

> Several important developers switched to this new idea, the ones who work in production grade stuff (one example is Antirez, who just wrote some tweets that inspired this post)

Does that mean we should blindly follow them? Just because someone else is doing it does not mean you should too.

A better questions is...have we lost the ability to think for ourselves because we have LLMs doing all the thinking instead?

[al_borland]

Speed has never been my number one goal. I’m not sure why I’d start to optimize for that now.

[rstagi]

Speeding up delivering value to the users seem a pretty good goal to me tbh

[al_borland]

You mention in the original post having more bugs. When I use something for the first time and it has a bunch of bugs that prevent me from using it, I never bother using it again. That loss of trust is hard to win back and it erases the potential value. I want to deliver a good experience to the user, first and foremost, so they actually come back.

As a user, I’d rather the developers release later with a solid product/feature.

As a developer, I’d rather release something that won’t be a support nightmare.

[rstagi]

“More bugs” is different from “a bunch of bugs”, and I agree on the support nightmare. My statement was implying that without looking at the code we might still be able to release something good quality that doesn’t fire back :)

[AnimalMuppet]

How perfect is your QA?

I'll answer for myself: At no organization I have ever been at was it perfect enough that devs didn't have to worry about the quality of what they produced. So I think you're asking QA to bear a burden that it has not been able to bear at any place I have ever been.

Second question: What is your time frame? Are you building for the next six months, or the next 20 years?

It is an open question whether AI will produce stuff that is maintainable in the long term. (I know, you have to survive the next six months in order to get to the long term...) But if you don't look, you don't know what kind of headaches are being written.

[rstagi]

Fair point on the QA, I might have some strong take there but it’s a chapter I’d rather not open now, but I guess the point of my post is on the last sentence of yours: is looking at the code really necessary to prevent headaches?

I have started delegating chunks of code without looking, and many times when I do look for one reason or another (I still want to know the big picture!) they looked good to me. So my idea it’s just that it could work, but I’m still experimenting as most of us :)

[verdverm]

The issue is long-term code quality. These things are not good at it. They locally optimize, reimplement (inconsistently), and are overly defensive. You can only observe this by looking at the code they produce.

[swatcoder]

With respect, you may not have the experience to recognize the cost tradeoff you're suggesting here.

If you're choosing to rely on a strict, exhaustive validation process to approve your work product without concern for how it's produced, you shifting focus to several critical cost centers:

1. Exhaustive validation is not free and its cost grow super-linearly as your work product gains complexity -- both on its tested surfaces and in what inefficiencies it develops internally. As you invite more frequent commits of more uncoordinated code, you're running your QA process ever more often on ever more heavy code, covering ever more work surface area. This is non-trivial cost and potentially catastrophic risk.

2. AI costs scale with token requirements and project complexity. As you allow uncoordinated code to accumulate within your product, you super-linearly increase the token consumption required by AI agents attempting future commits and analyses as they digest the increasingly chaotic/uncoordinated code base, both in raw input tokens and the reasoning tokens used to strategize plans. Even if you believe that your AI agents will be able to keep up with this expanding burden indefinitely (a questionable assumption!), you're incurring another growing, super-linear cost.

3. If you're relying on frontier models to make this workflow feasible at all, you're at risk of uncontrolled cost ratcheting if and when your code grows so speghettified that only those models can maintain it. You may have economics that work today, but the model providers you use are positioned to squeeze your margin aggressively once you trap yourself within their ecosystem.

Now, it's possible that an adept team can figure out how remediate some of these concerns by appropriately tasking agents to reduce project complexity and efficiency independent of new feature development. But the overall workflow here remains uncharted territory, where few people are going to be getting it right, and many will find themselves being painted into a very expensive dead end.

TLDR; do what you want, but all the traditional concerns of technical debt management and vendor exploitation risk are at play here. They've just shifted around in a way that makes it easy for inexperienced or myopic people to think they've somehow escaped them because the underlying technology itself is so novel and unfamiliar.

In general, what you're talking about here is something that businesses have been able to do for decades by delegating their tasks to outsourced suppliers. The lesson, then and now, is that it works until it doesn't and that once it stops working, you're screwed.

So be careful -- the revolution you've promised is not what you imagine it to be!

[rstagi]

Thank you for your thoughtful answer!

Just to be clear: I haven’t made up my mind yet either :) that’s why I asked a question and avoided delivering truths or anything

Everything you said are all doubts I have, and especially the “it works until it doesn’t” is my biggest fear.

I’m probably less afraid of the “once it stops you’re screwed” part, I have been working in places where taking up codebases and products full of tech debt and renewing them was the norm, so I believe in the worst case that’s still doable.

So I’m not saying it’s an easy and straightforward revolution, yet I’m not so negative about it :)