IdeaWell

Project ideas from Hacker News discussions.

Apple's Siri AI Can Be Swapped Out for Claude, ChatGPT, Code Shows

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

1. EU DMA pressure for interoperable AI choice
- “Apple has chosen not to release the new Siri AI in the EU because our Digital Markets Act requires them to allow interoperability. I.e. that users be allowed to choose another AI provider.” – Gareth321
- “Apple proposed a solution to the EU … The EU said ‘build it then we’ll tell you’.” – MBCook

2. Flexible, local‑first routing architecture
- “flexible routing with easy tasks running locally and only the more complex ones hitting the cloud” – PunchyHamster
- “Handle the query on device. If the device cannot handle the query, push the query to Private Cloud, If Private Cloud cannot handle the query, push the query to a third party of the users choosing.” – spiralpolitik

3. Skepticism about Apple’s true motives and desire for control
- “Apple had enough experience relying on and being screwed by third parties that they have a built‑in allergy to it. They do not like being boxed in and always want to be in control of their destiny.” – MBCook
- “this was not Apple's plan A … they are able to massively fail at developing their own models and still end up in a good position.” – freedomben
- “This looks more like a participation medal, if at all.” – bel8

🚀 Project Ideas

SiriSwap: Interoperable AI Backend Selector for iOS

Summary

  • Allows iOS users to choose any AI provider (local model, cloud API, or Apple’s Private Cloud) as the backend for Siri, satisfying EU DMA interoperability requirements.
  • Core value: restores user choice and privacy while keeping the familiar Siri UI and system‑wide integration.

Details

Key Value
Target Audience Privacy‑conscious iOS users in the EU, developers who want to test alternative models with Siri
Core Feature System‑wide toggle in Settings → Siri & Search → AI Provider that routes all Siri voice and text requests through a secure local proxy to the selected model; supports MCP for tool calls
Tech Stack Swift, NetworkExtension framework, Apple’s AppIntents, gRPC/WebSocket proxy, optional CoreML/MLX for local models
Difficulty Medium (requires entitlements and careful handling of audio session)
Monetization Revenue-ready: Freemium (free basic routing, $2.99/month for advanced rules, custom model uploads, and priority support)

Notes

  • HN commenters emphasized the need for true interchangeability: “They want Siri to be replaced by Gemini/etc completely.” – Gareth321
  • “Apple proposed a solution … they want Siri to be replaced by Gemini/etc completely.” – MBCook
  • Enables the “local AI hub” desire expressed by syntaxing: “I would pay a good chunk of money if Apple released a local AI hub to coordinate all AI usage locally.”
  • Provides a concrete path for DMA compliance while preserving Apple’s user experience, likely sparking discussion about interoperability vs. platform control.

LocalAI Hub: On‑Device Model Router for Hybrid AI Workloads

Summary

  • Daemon that automatically routes simple queries to an on‑device LLM (e.g., Phi‑3, Mistral) and complex or privacy‑sensitive requests to a user‑chosen cloud model, with fallback to Apple’s Private Cloud.
  • Core value: gives users low‑latency, private responses for everyday tasks while still accessing powerful cloud AI when needed.

Details

Key Value
Target Audience Power users, developers, and privacy‑advocates on iOS/macOS who want control over where their AI workloads run
Core Feature Configurable routing rules (based on query length, keywords, latency thresholds, or privacy tags) that decide whether to call the local engine, a custom cloud endpoint, or PCC
Tech Stack Swift/XPC service, MLX or Llama.cpp for on‑device inference, gRPC/HTTP client for cloud, Settings bundle for rule configuration
Difficulty High (balancing model loading, power management, and secure inter‑process communication)
Monetization Hobby

Notes

  • Reflects the comment: “flexible routing with easy tasks running locally and only the more complex ones hitting the cloud” – PunchyHamster
  • Aligns with the fm command examples showing local vs. PCC model selection: “Use the larger Private Cloud Compute model” – runjake
  • Supports MBCook’s vision: “make Siri / your mac the central access point for all AI related workflows and make models interchangeable.”
  • Would likely generate discussion on optimal routing strategies and the trade‑offs between on‑device quality and cloud capability.

SiriMCP Bridge: Tool‑Invocation Framework for Voice Assistants

Summary

  • Exposes iOS app intents, HomeKit actions, and custom shortcuts as MCP (Model Context Protocol) tools, enabling any AI model (including Siri) to perform app‑level tasks via voice.
  • Core value: turns Siri (or any assistant) into a universal voice‑driven orchestrator for apps and smart home without waiting for Apple to add each integration.

Details

Key Value
Target Audience Developers building voice‑first experiences, advanced users who want to control their phone/desktop with any LLM, HomeKit enthusiasts
Core Feature MCP server that registers available intents/actions as tools; translates incoming tool calls from a model into SiriKit/AppIntents or HomeKit requests, returning results as MCP responses
Tech Stack Swift (AppIntents, HomeKit) + a lightweight MCP server (Python or Rust) communicating via local Unix socket or XPC; optional Docker‑compatible container for easy deployment
Difficulty Medium (requires understanding of MCP spec and bridging to Apple’s intent system)
Monetization Hobby

Notes

  • Directly addresses sajithdilshan: “Create a platform where app actions can be invoked with a permission model by any AI model.”
  • Supports theturtletalks’ excitement: “Once Siri gets MCP support, it’ll be my main orchestrator.”
  • Enables the use case described by stego‑tech: “Now let me point it to an OpenAI‑compatible API endpoint of my choosing …”
  • Would likely spark conversation about standardizing voice‑tool interfaces and reducing reliance on vendor‑specific skills.

Read Later