What's working for YC companies since the AI boom

Original Article ↗ Hacker News Discussion ↗

YC’s AI Focus & Batch Composition

  • Notable absence of consumer products is seen by some as YC being too narrow; others say it simply reflects macroeconomy and AI’s current stage.
  • Several commenters argue YC is heavily skewing toward AI, shaping who applies and gets in, rather than “just picking the best founders.”
  • There’s concern YC has become insular (B2B, often B2–YC), optimizing for selling to other YC/Valley companies rather than the broader economy.

Consumer vs B2B AI

  • Multiple explanations for “0 consumer”:
    • Easier/cheaper for incumbents to bolt AI onto existing consumer products than for a new entrant to build brand + pay inference costs.
    • Consumer AI often needs huge capital to subsidize usage (like free ChatGPT), which early startups can’t match.
    • B2C norms of “free” push startups into ad/shady models or lottery-style “hit” dynamics.
  • Others counter that consumer AI is already vibrant (e.g. search, music, multimedia apps) and may even be healthier than enterprise AI, where many projects don’t justify their spend.

AI Startup Viability & Moats

  • Pattern described: “ChatGPT but for X” gets funded, then the platform providers ship a better built-in version, erasing the startup’s wedge.
  • View 1: “AI startups” as a category are fragile; general models and incumbents quickly absorb successful ideas.
  • View 2: Moats live in vertical UX, integration, data, and deterministic workflows with AI as an assistant, not the control loop. Document understanding/IDP is cited as a large, enduring space where specialized players can thrive.

Metrics: Series A vs Real Traction

  • Many argue Series A count is a poor proxy for “what’s working”:
    • Post-ZIRP, more startups push for early revenue and even cash-flow positivity, delaying or skipping A rounds.
    • Some companies reportedly have multi‑million ARR on just seed money.
    • Better metrics suggested: non‑YC customer growth and churn.

Tooling, Evaluation & Infra

  • Absence of LLM evaluation/observability/tooling in the Series‑A list is seen as natural: patterns are immature and it’s hard to pick winners.
  • Confusion over what “tooling” means (infra like local model runners vs dev tools vs runtime monitoring).

Hardware & VC

  • Zero hardware in the Series‑A data resonates with hardware engineers who say traditional VC timelines and expectations don’t fit long, capital‑intensive hardware cycles.
  • Some see this as healthy: bootstrapping, strategic customers, and slower growth may be better aligned than mainstream VC.

AI Hype vs Reality

  • One camp claims YC is going all‑in on AI with unproven business value, partly due to its stake in foundational players.
  • Another counters that seed capital is supposed to underwrite exactly this kind of technology/market risk; lack of quick Series As doesn’t imply lack of long‑term economic impact.