Claude and ChatGPT need a datacenter. This runs on my phone
Details
- External ID
- 49526848
- Source
- HN
- Company
- —
- Product
- Claude and ChatGPT need a datacenter. This runs on my phone
- Website domain
- pages.dev
- Launched
- Sept. 1, 2026
- Cohort
- —
- Upvotes
- 18
- Upvotes percentile
- 0.7216905901116427
- Tags
- —
- Fetched at
- Sept. 10, 2026, 5:31 a.m.
- Updated at
- Sept. 10, 2026, 5:31 a.m.
Enrichment
- Theme
- niche social and community platforms
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Hobby / open-source project
- Normalized one-liner
- run llms locally on mobile devices
- Manually corrected
- False
Could you build this?
No Running local LLMs efficiently on mobile devices without internet requires deep expertise in low-level model quantization, Apple Metal / Qualcomm NPU hardware acceleration kernels (e.g. llama.cpp or MLX porting), and mobile memory optimization. It is far beyond the reach of vibe coding.
What it would actually take: The app requires compiling and optimizing C++ inference engines like llama.cpp or MLC-LLM for iOS/Android, linking against Metal Performance Shaders (MPS) or NNAPI/Hexagon DSPs. Models must be quantized to 2-4 bit formats (e.g., GGUF, AWQ) with strict memory bounds so the OS watchdog doesn't kill the app due to RAM limits. This requires specialized expertise in low-level ML systems engineering, hardware-accelerated computing, and mobile OS memory management.
Discussion
7 comments analyzed.
Competitors mentioned: PocketPal AI, Google AI Edge Gallery, MLC Chat
Concerns raised: Performance significantly slower than competitors (AI Edge, MLC, PocketPal), Exaggerated claims about running on 'almost any phone' - actual performance underwhelming, In-app performance indicator misleading ('run comfortably' claim doesn't match reality), Direct APK download from website lacks professional presentation, security concerns
Feature requests: Optional online lookup mode while keeping offline as default, Tool integration for models to perform various tasks via harness
Competitors
Other products that read as similar to this one — 96 launches clear the similarity bar, closest 8 shown.
Attention rank: #34 of 97 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 293 days after the earliest competitor.
- I got 50% of my traffic from ChatGPT instead of Google · hn · 2025-12-09 · 16 upvotes · similarity 0.49
- tunneldock · github · 2026-09-16 · 38 upvotes · similarity 0.48
- chatgpt-prism2api · github · 2026-09-24 · 15 upvotes · similarity 0.47
- Claude-codex-proxy · hn · 2026-04-19 · 7 upvotes · similarity 0.45
- ChatGPT-i18n · github · 2026-09-09 · 59 upvotes · similarity 0.44
- ChatGPT Work · ph · 2026-07-11 · 249 upvotes · similarity 0.41
- Shellular · hn · 2026-07-07 · 33 upvotes · similarity 0.41
- chatgpt-local-connector · github · 2026-09-19 · 11 upvotes · similarity 0.40
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.