logoalt Hacker News

aleqs • yesterday at 7:18 PM • 9 replies • view on HN

The degradation in software quality across the board (Spotify, Google Maps, GitHub, YouTube - to name a few) has been very noticeable for me, and I suspect AI has a lot to do with this.

The thing is - I also think AI can be used to build very high quality software. The problem, I think, is that the typical big tech/VC tech/MBA tech type company (who control most proprietary software at this point) have very little incentive to actually care about quality, and that was quite obvious even before AI.


Replies

pkulak • yesterday at 7:20 PM

I remember before I started using AI (I was pretty late to the game), I assumed all the bugs and instability were because people were writing software with AI. Now that I've seen how good the code that AI writes can be (CAN be), I'm moving more to the thinking that it's from AI being used to find vulnerabilities.

➕ show 2 replies
sakopov • yesterday at 7:40 PM

I think majority of companies don't know how to build safely with AI. It's just kind of handed off to engineering and mostly used as one-off chat interaction to get stuff done. Very few are building any kind of harnesses around SDLC that focus on architecture, testing, security and etc. I think we'll get there at some point, but in my experience the smaller tech players are just now on-boarding to use AI in general in any kind of capacity. We're not talking about best practices yet.

ozozozd • yesterday at 9:13 PM

Wondering if this has anything to do with the QA function being folded into the dev team in the last 15 years or so.

I always defended this, but it appears maybe we don’t have enough developers that take pride in the work they ship. LLM coding made this pretty obvious.

➕ show 2 replies
DIE_DIE_DIE • yesterday at 8:00 PM

> I also think AI can be used to build very high quality software.

I don't, and I think your comment is rhetorically useless because you've been bullied into making this ridiculous concession.

➕ show 1 reply
gtowey • yesterday at 7:29 PM

IMHO the only way to build good things is to pay attention to the details. And coding agents, while greatly increasing the appearance of productivity, they are causing people to pay less attention to what they're actually producing. This is the inevitable and obvious result.

Agentic coding could be a world-changing benefit, but the current way it's being sold and used as as a cost cutting measure -- produce software faster than ever before with fewer software engineers! Get rid of those pesky experts!

pwillia7 • yesterday at 7:51 PM

We're in the v1 machines screws stage where everything is worse but with a clear path to get better .... I hope

lotharcable • yesterday at 7:32 PM

It isn't a software or tool problem.

It is a C-level executive problem. With the FAA it is the agency directorship problem. Same issue.

These people, very literally, are trying to run the world with power point presentations and executive summaries drafted by sycophants. 70 years ago these people had actual experience doing actual work and trying to grow businesses and doing stuff at the ground level. Now they are given their positions through a combination of credentials, connections, and political maneuvering. None of which actually relates to how well they can or cannot run a organization.

When reality disagrees with what they get in their emails and presentations then it reality that is wrong. It is the messengers that are the problem, not their competency.

This is nothing new. But the failures in leadership compound over time, like interest to a unpaid debt. They start off with healthy growing companies and thanks to government protections and bailouts it takes decades to for these organizations to decay into total dysfunction. In the past these sorts of people would be removed and their assets reallocated to competent management through bankruptcy and competition. However that natural process is not allowed anymore.

All AI is doing is handing them more foot-guns.

They think they can do better/faster/stronger/cheaper with AI and if AI causes problems then it can just be fixed by paying for better AI. Which they assume is just right around the corner, always.

Just wait until they start to realize that it is

➕ show 2 replies
matt3210 • yesterday at 8:49 PM

broken windows

dmitrygr • yesterday at 7:20 PM

Part of being sane is accepting that some tools, while being very powerful, are too easily misused to be widely distributed to those lacking proper training. This is why you cannot just go and buy 5 kilos of azithromycin

➕ show 2 replies