Pocket Intel
An offline-first AI research companion that runs on a phone or USB stick so journalists, aid workers, and traveling researchers can summarize documents, transcribe interviews, and translate conversations without ever going online.
freelance investigative journalists and field researchers who work on sensitive stories or in low-connectivity regions
- Local document Q&A: drop in PDFs and notes, ask questions, get cited answers with no cloud call
- On-device interview transcription that never leaves the phone
- Multi-language translation for documents and recorded conversations, fully offline
- USB-boot option that runs from a thumb drive on any laptop and leaves no trace
BMASS-style offline AI OS projects are trending on GitHub because source protection and connectivity gaps are acute — reporters working on whistleblowers, aid teams in disaster zones, and researchers in restrictive jurisdictions need AI that never touches the internet.
Multiple independent sources confirm demand: Freedom.press recommends offline AI for journalists, the ITU AI for Good program has humanitarian field AI projects (0-LA, GBViE LOAIS), a Gemma hackathon entry specifically targets humanitarian field workers, and dedicated products like OfflineGPT and Viska already serve this user base.Using AI safely as a journalist: Stand-alone AI tools ↗Zero Latency Assistant (0-LA) - AI for Good ↗Offline AI for Journalists | Secure Reporting Tool | OfflineGPT ↗
Space is crowded with overlapping solutions: OfflineGPT (iPhone offline AI agent), Viska (on-device Whisper + local LLM), Wavery's WhisperScript (offline transcription), GPT4All/Ollama/AnythingLLM (general offline LLMs), plus specialized humanitarian tools like GBViE LOAIS and 0-LA — direct overlap on document summary, transcription, and translation.OfflineGPT - Offline AI Agent for iPhone ↗Viska - Private AI Voice Notes | Offline Transcription & Local AI Chat ↗Why We Built WhisperScript - Offline, Private AI Transcription ↗
OfflineGPT has a published pricing page, indicating willingness to charge; however, free open-source alternatives (GPT4All, Ollama, AnythingLLM, whisper.cpp) dominate the space, and freelance journalists are notoriously budget-constrained — monetization is possible but constrained.Pricing | OfflineGPT ↗Local AI for Journalists: Protect Sources With Offline AI ↗
BMASS confirms the offline-AI trend (bootable Linux + llama.cpp + quantized Qwen3 on USB stick, profiled on Korben), edge/local AI infrastructure is improving rapidly with smaller capable models (Phi-4 Mini, Gemma 3 4B), and source-protection concerns are intensifying globally — long-term tailwinds are strong.BMASS - Boot your PC directly into a local AI - Korben ↗Local Voice Assistant Whisper + LLM Phone 2026 ↗
Stack is well-documented and proven: Whisper.cpp + llama.cpp + Ollama + Piper TTS runs locally on phones (WhisperKit + LLM Farm on iOS); multiple shipped products (OfflineGPT, Viska, BMASS on 8GB USB) demonstrate the pipeline works end-to-end on commodity hardware.Local Voice Assistant 2026: Whisper + LLM + Piper TTS ↗Run Whisper Locally 2026: Free Offline Speech-to-Text Setup ↗