Taste: The Human Gatekeeper in the Age of AI Slop

The author argues that LLMs have made it easier to write code but harder to stand out because they cannot provide 'taste.' This has led to an influx of low-quality, derivative applications that the tech community largely views as noise or 'slop.' Ultimately, human discernment remains the essential barrier to creating successful and resonant software.

Key Points

• LLMs have lowered the technical barrier to building apps but have not lowered the barrier of taste.
• The 'magic quadrant' of software consists of skill and taste, with taste being the harder element to automate or learn.
• The current tech scene is being flooded with 'slop'—derivative, poorly crafted apps that lack original utility or aesthetic value.
• Taste is relative to the audience, but there is a minimal universal standard that creators must clear before sharing their work.
• LLMs amplify the importance of taste because they make it easier for people with poor discernment to produce and share unoriginal ideas.

Sentiment

The community is genuinely split on the article. While many commenters agree that the flood of low-effort AI-generated apps is a real problem and that taste matters, a significant and vocal contingent pushes back against what they perceive as gatekeeping and elitism. The philosophical debate about what taste even means reveals fundamental disagreements. The overall tone leans slightly sympathetic to the article's concern about quality declining, but there is strong resistance to the article's framing and a widespread belief that building things for personal use is inherently good regardless of external standards.

In Agreement

• Show HN has been flooded with hastily produced, low-effort vibe-coded apps that demonstrate neither skill nor consideration for the audience.
• Taste is not purely subjective but can be measured through community reception and demographic consensus, making it a meaningful quality standard.
• Spending excessive time with generative AI creates a slot-machine habituation effect that degrades one's ability to recognize quality.
• Taste is fundamentally the core skill of product management, and developers are now wrestling with something that entire careers have been built around.
• AI tools disproportionately benefit those who already possess skill, meaning the barrier to creating truly good software has arguably never been higher.
• Slop has negative value because it buries genuinely good work in a sea of noise, as demonstrated by app store discoverability problems.
• The real challenge in software development is data management and architecture, not code generation, which vibe coding does nothing to address.
• Good ideas come from the struggle and process of creating, so skipping the journey through AI shortcuts often produces mediocre results.

Opposed

• The article amounts to gatekeeping, with the author appointing himself as arbiter of what is acceptable to share, which is not his role.
• Building apps for personal use is inherently valuable regardless of whether they meet someone else's standards of taste, and having many options is a feature, not a bug.
• The concept of taste is too vaguely defined in the article to serve as a meaningful differentiator between good and bad software.
• Taste is deeply intertwined with class, status, and cultural capital, and invoking it as a standard often serves to reinforce existing hierarchies rather than identify genuine quality.
• AI has enabled people who were previously locked out of software creation, including a blind developer and young children, to build genuinely useful and personally meaningful tools.
• The article is itself derivative, essentially the hundredth AI think-piece with no original thought, making its complaint about derivative work ironic.
• A lower barrier to skill could bring people from other disciplines with excellent taste in design, UX, or domain knowledge into software creation for the first time.
• People should be free to post their work and receive feedback as part of their learning process, rather than being expected to pre-filter based on someone else's quality standards.