OpenAI Working on AI Smartphone with MediaTek/Qualcomm Chips; Mass Production Target 2028
Ming-Chi Kuo's latest supply chain checks reveal OpenAI is working on a smartphone, reversing earlier reports that the company had no phone market plans. Key details:
- Chip partners: MediaTek and Qualcomm are chosen, with exact specs and additional suppliers expected by late 2026 or Q1 2027.
- Manufacturer: Luxshare Precision Industry is the exclusive manufacturing partner.
- Timeline: Mass production scheduled for 2028; first hardware announcement (likely a smart speaker) expected H2 2026 with launch around early 2027.
- Rationale: Kuo argues smartphones are uniquely positioned for AI agents because they capture user's full real-time state (location, activity, communication, context) — the most important input for real-time AI agent inference. Full control of OS and hardware is required for a comprehensive AI agent service.
- Business model: Subscription-bundled model could enable a developer ecosystem around agents.
- Context: This contrasts with OpenAI's prior hardware strategy focused on non-phone form factors (smart speaker, smart glasses, smart lamp, earbuds) developed with Jony Ive's io Products (acquired for $6.5B).
OpenAI CEO Sam Altman posted on X the same day, saying it feels like a good time to seriously rethink how operating systems and user interfaces are designed.
📖 Read the full source: HN AI Agents
👀 See Also
A 50-Dev Low-Code Shop Vaporized in 12 Months: The Dependency Trap of AI Coding Agents
A 50-person low-code shop lost all clients in 12 months because "low-code + AI" beats pure low-code and full-stack. Meanwhile, a solo developer dependent on Claude Max faces session caps and rising costs. Both illustrate the same dilemma: adapt or depend.
Talkie: A 13B LLM Trained Exclusively on Pre-1931 Text, Using Claude as a Judge in RL Training
Researchers released Talkie, a 13B LLM trained only on text published before 1931 (no internet, no WWII data). Claude Sonnet 4.6 was used as the judge in its online DPO reinforcement learning pipeline, and Claude Opus 4.4 generated synthetic multi-turn conversations for fine-tuning. The model can write Python code from a few in-context examples despite zero modern code in training.
Claude-Code v2.1.94 adds Mantle support, fixes critical bugs
Claude-Code v2.1.94 introduces Amazon Bedrock support via Mantle with the CLAUDE_CODE_USE_MANTLE=1 environment variable, changes default effort level to high for most users, and fixes 15+ bugs including rate-limit handling, macOS login issues, and plugin system problems.
Tokenmaxxing Is the New Stopwatch: Why Your AI Policy Needs to Be Coherent
Brian Meeker argues against vanity metrics like tokenmaxxing and shares his team's four-point AI policy: no mandate, understand generated code, survive without AI tools, care about teammates and customers.