John Giannandrea to retire from Apple

📝 Discussion Summary (Click to expand)

The discussion revolves heavily around the perceived failings of Apple's Siri. Here are the three most prevalent themes, supported by user quotes:

1. Siri's Extreme and Persistent Lack of Utility / Functionality Regression

A major theme is that Siri has remained stagnant or, worse, has actively degraded in usefulness over its many years of existence, with users primarily relying on it only for the most basic functions like timers and alarms.

Quotation: "I remember buying the iPhone 4S in 2011, and it being the first iPhone to ship with Siri. It's 2025, and Siri is still fundamentally useless." - "pharos92"
Quotation: "Honestly, my experience with Siri is that it works worse than it did 10 years ago." - "ianferrel"
Quotation: "Uses for Siri: 1. Checking the current temp or weather 2. Setting an alarm, timer, or reminder 3. Skipping a music track or stopping the music altogether..." - "98codes"

2. Inconsistent Failures and Lack of Transparency (Silent Failures)

Users expressed significant frustration with Siri's inconsistency—commands that work one day fail the next, often without explanation or meaningful feedback. This lack of transparency contrasts sharply with what users expect from technology.

Quotation: "The silent failure modes everywhere gets really frustrating." - "highwaylights"
Quotation: "I'm amazed 'set a reminder for x when I leave this location' still doesn't get the 'when I leave this location'." - "npunt"
Quotation: "How is it even possible to build a computer system that behaves like this? ... I'm sorry, I don't know a device called 'bedroom lights'." - "perryizgr8" (Describing Alexa's non-determinism, which is seen as endemic to the voice assistant category.)

3. Organizational Missteps and Executive Direction Blamed for AI Stagnation

There is speculation that the leadership and organizational structure within Apple, particularly concerning AI development, stifled progress, with specific focus on the tenure of John Giannandrea.

Quotation: "Nah, ksec had a great post in a thread a few months ago about a number of senior execs Tim Cook has promoted to positions to which they're ill-suited. John Giannandrea was just one of many such people..." - "biglyburrito"
Quotation: "He did nothing, or net-negative. Same as Google; it was totally unclear how this guy got promoted to the level of power he had based on his history." - "dekhn"
Quotation: "I worked on systems for evaluating the quality of models over time... It was difficult to get Siri to use these tools that were outside of their org." - "bradly" (Citing organizational friction impacting quality evaluation.)

🚀 Project Ideas

Context-Aware Command Verification Service (CACVS)

Summary

A backend service that performs post-interpretation "reality checks" on high-consequence voice or text commands before execution, specifically targeting failures identified in assistants like Siri (e.g., playing the wrong song, calling the wrong number, executing unrelated actions).
Core value proposition: Drastically reduce unintended and undesirable actions caused by LLM/NLP failures, enhancing user trust and safety, especially for critical functions.

Details

Key	Value
Target Audience	Developers integrating or relying on voice command interpretation layers (e.g., smart home platforms, automotive software, third-party iOS/Android assistants).
Core Feature	A queryable API that receives a parsed "Intended Action" (Intent, Slots) and the User/Context metadata, returning a Confidence Score and a suggested "Action Plan" (e.g., confirm, execute, or cancel).
Tech Stack	Python/FastAPI, PostgreSQL for storing user history/preferences, potentially leveraging lightweight local models or RAG against user data for personalization.
Difficulty	Medium (Building robust context profiling requires careful data handling, but the core execution logic is service-based).
Monetization	Hobby

Notes

"What I've seen more of the latter [contextual errors] than the former [hallucinations]" (gedy). Users explicitly call for this: "I'm amazed more AI tools don't have reality checks as part of the command flow." (npunt).
This solves the issue of Siri performing completely unrelated actions like calling the Tanzanian embassy instead of opening the garage door, by checking historical norms.

Siri/Voice UI Discoverability & Capability Explorer Tool

Summary

A lightweight desktop/mobile tool that interfaces with Apple's Shortcuts/Siri Intents documentation (or simulates interaction) to provide an always-available, searchable interface showing exactly what commands Siri can currently process for integrated services.
Core value proposition: Solve the "hunt the verb" problem inherent in natural language interfaces lacking explicit documentation, providing discoverability without forcing the user to search external web pages.

Details

Key	Value
Target Audience	Frustrated Siri power users and casual users who want to maximize utility but don't know the "magical incantation."
Core Feature	A live, filtered search interface that dynamically suggests command structures based on installed apps and system capabilities (e.g., showing specific location-based reminder syntax).
Tech Stack	Swift/SwiftUI (for native feel and deep OS integration), structured data parsing (JSON/XML) for officially documented Intents, potentially using local LLMs for summarizing capabilities rather than executing.
Difficulty	Medium (Requires clever parsing of Apple's documentation/intent structures, but avoids the complexity of backend API integration).
Monetization	Hobby

Notes

Commenters noted: "it never communicated what that menu was" (epistasis) and "the problem really boils down to the fact that... a natural language interface looks like it should support 'basically any query' but in practice does not... Those limitations are not obvious" (nneonneo).
This tool directly addresses the mismatch between user expectation (human conversational interface) and technical reality (narrow command set).

Cross-Platform Voice-to-Action Synchronization & Debug Interface

Summary

A service focused on standardizing and logging voice command attempts across different devices (iPhone, HomePod, CarPlay) and mapping them to successful executions via Shortcuts or Home Assistant, providing transparency into failure modes.
Core value proposition: Provide consistency and detailed debugging information for voice commands that fail differently across various Apple hardware, allowing users to unify and troubleshoot their voice interactions.

Details

Key	Value
Target Audience	Users heavily invested in HomeKit/Home Assistant who rely on Siri across multiple devices, and those frustrated by inconsistent execution ("Siri inconsistency still stands").
Core Feature	A central dashboard showing the raw voice input, the successful execution path (e.g., routed via Home Assistant, failed at HomePod server), and historical success rates for specific phrases.
Tech Stack	Self-hosted backend (Go/Node.js), heavy use of logging/event streaming, potentially leveraging Home Assistant integration APIs.
Difficulty	High (Requires deep integration with local device logs or leveraging outputs from intermediary smart home systems).
Monetization	Hobby

Notes

Users complained about context loss and device inconsistency: "why is it possible for commands that used to work... to just stop working?" (quenix). And, "I suspect that [Siri] doesn't have access to the data [when locked]" (VogonPoetry).
This acknowledges that much of the utility users want (smart home control, reliable execution) is currently happening outside of first-party Siri logic (e.g., via Home Assistant). This bridges the gap by logging the attempt regardless of where it ultimately resolves.