The GPT Store

Original Article ↗ Hacker News Discussion ↗

OpenAI’s launch of the GPT Store, a marketplace for custom ChatGPT-based “apps,” is drawing both excitement and skepticism. Commenters debate whether simple prompt-based GPTs have real value, how users will discover quality amid millions of low-effort entries, and whether OpenAI’s vague “engagement-based” revenue sharing and tight policy controls (including bans on romantic companions) give third‑party developers enough incentive or protection. Others question the strategic focus on an app store, raise concerns about privacy and spam, and note uneven quality and reliability in the current GPT ecosystem.

Role and Purpose of the Official GPT Store

  • Many expect the official store to dominate simply through distribution; most users don’t know third‑party catalogs exist.
  • Some see it as a way for OpenAI to gather usage data and revenue, and to acclimate people to AI “agents.”
  • Others argue running a two‑sided marketplace distracts from the stated AGI mission.

Monetization, Incentives, and Copying Concerns

  • Revenue program promises US builders payment based on “user engagement,” but details are vague and not trusted.
  • Developers worry about hosting their own backends without clear ROI and fear OpenAI will clone popular GPTs as “official” versions.
  • Some frame this as standard capitalism; others see it as exploitative but historically common.
  • Several predict microtransactions or stronger monetization later; currently there’s “no moat” for prompt‑only GPTs.

Barrier to Entry, Discovery, and Quality

  • Creation barrier is extremely low (millions of GPTs already), raising fears of a flood of mediocre prompt wrappers.
  • Debate over whether “low barrier” is inherently bad: the real issue is discovery and ranking.
  • OpenAI’s past plugin store had poor discovery; current store ranking is reportedly simplistic.
  • Example: niche GPTs (e.g., trail finder) sometimes perform worse than plain ChatGPT, often due to restrictive instructions or narrow data sources.
  • Some expect eventual algorithmic discovery and virality, similar to YouTube/SoundCloud.

What Custom GPTs Actually Are

  • Components mentioned:
    • Behavior: detailed system instructions.
    • Knowledge: uploaded files / RAG via embeddings.
    • Capabilities: browsing, code, image generation, plus custom “actions” hitting external APIs.
  • Interesting GPTs usually require real backend services and careful prompt design, not just a single prompt.

Developer Experience and Limitations

  • Actions are described as brittle: strict checks prevent hallucinated parameters but produce user‑visible failures.
  • Users can see requests but not action responses, reducing transparency and trust.
  • Loss of plugin‑style response inspection and lack of a native “fork” button frustrate tinkerers, though many GPTs will reveal their instructions if asked.

Access, Pricing, and Policy

  • Store is gated behind a paid subscription, with minimal feature descriptions; users dislike paying without being able to meaningfully “try before buy.”
  • Some argue $20/month is a bargain; others note free alternatives and company restrictions that limit work use.
  • OpenAI’s updated policies ban profanity in names, graphic violence, romantic companionship, and regulated activities; rationale for banning romance is debated, with references to prior controversies.
  • Concerns raised about data sharing with third‑party GPTs and potential for spam, malware, and prompt‑injection attacks.