How the AI bubble ate Y Combinator

Original Article ↗ Hacker News Discussion ↗

AI Hype, Bubble, and Actual Usefulness

  • Many commenters see AI—especially LLMs—as massively overhyped, repeating earlier crypto/web3/“blockchain everywhere” cycles.
  • Others argue there is real value: fast prototyping, translation, some developer productivity, and niche tools, even if 90% of “AI startups” are thin ChatGPT wrappers.
  • Distinction is made between “using AI” as a component vs being fundamentally an “AI company”; critics say counting every startup that mentions AI as an “AI startup” inflates bubble stats.
  • Some describe AI as a bubble built on investors’ FOMO and marketing, not on clear paths to profitability; others counter that bubbles can still form around genuinely useful tech.

Impact on YC and Venture Capital

  • Multiple commenters cite the stat that ~90% of recent YC startups are tagged AI, reading this as YC being “eaten” by the hype and churning out “AI slop” and wrappers.
  • Others say YC is just following incentives: many VCs reportedly fund “AI only,” so founders reframe anything as AI to get money.
  • Concern that YC now funds many overlapping/competing AI companies, even with licensing/ethics issues (e.g., the PearAI forking incident), creating a “tragedy of the commons.”
  • One view: the real bubble is venture capital itself—AI erodes software moats and makes defensibility hard to invest in.

HN, Discourse, and Tech Culture

  • Strong AI fatigue: users note AI “eating” the HN front page and corporate meetings, crowding out topics like FOSS, Linux, and niche tech.
  • Some lament that HN used to “make stories” and incubate deep debates (e.g., about FOSS), whereas now it mostly amplifies mainstream hype and avoids high-energy contentious topics.
  • Others push back, saying skepticism is healthy and that AI is legitimately the biggest current tech story, just as blockchain once was.

Developers, Work, and Products

  • Observations that most “AI work” is API wrapping because few devs have the skills or compute to work on core models.
  • Anxiety that this is the first major hype cycle where management openly dreams of replacing developers rather than empowering them.
  • Counter-argument: AI’s nondeterminism, hallucinations, and UX limits mean it won’t simply dissolve menu-driven, deterministic software.

Open Source, Centralization, and Society

  • Several threads contrast AI’s centralizing tendency (cloud models, closed data) with open source’s liberating potential, lamenting the decline of serious FOSS discussion and funding.
  • Some describe AI and social media as degrading learning, research habits, and communication (students and workers over-delegating thought to LLMs; rise of “bossware”).

Media, Paywalls, and Scraping

  • Frustration over the article’s paywall leads to a side-discussion: paywalls both fund journalism and act as a defense against AI scraping, but restrict public access to critical information.