Siri is so bad at everything
Siri’s new AI-powered incarnation is drawing heavy criticism from Apple users who say it still fails at basic tasks like setting timers, placing calls, controlling smart-home devices, and remembering preferences, often with added latency and unpredictability. Commenters contrast this with more capable LLM-based assistants such as ChatGPT or Grok, which feel smarter but lack deep OS integration, and with older, more deterministic voice-command systems that were more reliable for simple actions. Many see Siri’s shortcomings as a product and management failure rather than a privacy or on-device-compute limitation, and some question whether Apple’s broader home and automation ecosystem can be trusted given these persistent issues.
Overall sentiment on Siri
- Broad consensus that Siri is unreliable, slow, and often fails at basic tasks.
- Many describe increasing frustration and “rage,” especially because Siri is deeply embedded in the Apple ecosystem.
- A few report improvements with the latest Siri AI, but say they still treat it as “dumb” and limited.
Specific Failures and Regressions
- Core tasks routinely fail or behave inconsistently: setting timers, alarms, reminders, calling contacts, playing media, and controlling HomeKit scenes.
- Same command phrasing works one day and fails the next; Siri often requires multiple attempts.
- Alarming misfires: calling the wrong contact with similar name, mishearing “call X” as adding songs, or refusing to place calls until repeated.
- Apple Watch examples: Siri unable to set timers due to “missing Clock app,” extremely slow or no responses, and stale or inconsistent complication data.
- Home control and scenes are slow, flaky, and syntax-sensitive; users rename devices or scenes as a workaround.
LLMs vs Classic Siri Design
- Many believe moving Siri toward an LLM-style system worsened determinism and latency for simple commands.
- Some argue Siri should use small, robust models or fixed grammars for common actions, reserving LLMs for complex requests.
- Complaints that Siri lacks working memory for multi-step tasks and conversational corrections.
Comparisons with Other Assistants
- ChatGPT Voice and other LLM-based tools are praised as far more capable conversationally but lack OS integration.
- Google/Gemini and Alexa are criticized too, especially after “Gemini upgrades,” but are seen as better at some basics (e.g., timers, adding to timers).
- A few mention alternative ecosystems (Home Assistant, car-specific assistants) working more reliably than Apple’s stack.
Privacy, Strategy, and Product Direction
- Some dismiss “privacy” as an excuse for poor quality, arguing the failures are product and engineering issues, not inherent to on-device processing.
- Perception that Apple does not acknowledge Siri’s weaknesses and keeps shipping regressions.
- Criticism of Home/HomeKit onboarding, background update policies, and Siri’s data-learning settings as confusing or dark-pattern-like.
Use Cases, Skepticism, and Workarounds
- Defenders of assistants cite driving, cooking, and disability/hands-busy scenarios; skeptics say buttons are more reliable and that assistants mainly reduce “friction,” not meaningfully increase convenience.
- Some users disable Siri AI, revert to “classic” Siri, or rely on Shortcuts directly, ChatGPT/Claude apps, or third-party home automation instead.
- Meta-discussion questions why a short Reddit gripe post drives large HN engagement, with worries about brigading and declining discussion quality.