Nothing Works and Everyone Is Euphoric: Why Software Quality Is Declining Despite AI Hype

The Symptom: Everyday Software Is Getting Worse

Software quality is deteriorating across common consumer products. The author reports that his banking app needs three FaceID attempts before the 3D Secure view appears, Slack on macOS steals focus and sends unintended commands, an LG fridge warranty form fails silently, and a car infotainment system reboots during drives and misroutes taps. These examples show that updates now inspire dread rather than excitement.

Root Cause: Incentives Prioritize Speed Over Stability

The primary reason for the decline is that companies reward rapid feature shipping, not bug fixing. Management KPIs celebrate new features, while a quarter focused solely on bug fixes looks unexciting and is rarely pursued. Comments note that "the focus on speed (and shipping to prod) over quality has also led to the increasing number of bugs" and that "the top priority has been development speed, not product quality for 20 years). When stability work does not affect measurable metrics, it is deprioritized.

Contributing Factor: Rising Complexity Makes Systems Fragile

Software has become more complex through new abstractions, frontend frameworks, and infrastructure, while the bar for user experience keeps rising. This added complexity makes systems increasingly fragile, so that an OS update or app change is a source of dread. One comment observes that "software may have been better back in the day, but that was mainly because it was much simpler" and another argues that we should not exceed the level of complexity at which we can make software reliable.

AI’s Role: Accelerating Code Without Improving Correctness

AI coding assistants increase the speed at which developers can produce code, but they do not automatically improve correctness. The barrier to writing code is lower, yet the resulting software still contains bugs because LLMs are trained on buggy code and cannot guarantee deep system understanding. A comment states that "using LLMs for speed… neither the structures, architectures or decision makings changed" and another notes that "AI can extrapolate and interpolate what’s already been done… but I need to see something new to really believe." Thus, AI amplifies output without solving the underlying quality problem.

Community Insights: Specific Pain Points Highlighted

Commenters reinforce the author’s observations with concrete grievances: focus‑stealing windows that interrupt work (KDE Plasma offers a global setting to prevent this), security teams imposing excessive authentication steps that hinder usability, and UI changes that move buttons arbitrarily each release. One user describes Slack stealing focus and sending a git pull command to a group chat, while another notes that banking apps require multiple FaceID attempts before proceeding. These remarks show that the problems are widespread and not isolated.

Opportunity: Individual Developers Can Build Better Software

Despite industry‑wide trends, the author sees hope for individual developers who can leverage AI’s productivity to create software that would previously have been beyond their reach. By focusing on robustness, quality, and user experience rather than chasing fleeting features, small teams or solo efforts can counteract the broader decay. The author concludes that everyday software may improve as a reaction to current frustration, citing acts of rebellion against macOS and Windows as early signs.