Feed aggregator

Show HN: TypeWhisper – speech-to-text with multiple engines, profiles

Hacker News - Thu, 03/12/2026 - 6:11pm

Hey HN, I'm Marco, the creator of TypeWhisper.

TypeWhisper is a free, open-source speech-to-text app for macOS and Windows. Everything runs locally on your machine - no cloud, no telemetry, no data collection. Your voice never leaves your device.

What makes it different:

- Multiple engines, your choice: On macOS: WhisperKit, Parakeet TDT, Apple SpeechAnalyzer. On Windows: Parakeet TDT, Canary 180M Flash. All run on CPU, no GPU needed. You can also plug in cloud APIs (OpenAI, Groq, Deepgram) if you prefer.

- Context-aware profiles: Set per-app and per-website overrides for language, engine, and AI prompts. TypeWhisper detects which app you're in and adapts automatically.

- AI text processing: Built-in prompts for rewriting, translation (20 languages), summarization. Works with on-device Apple Intelligence or cloud LLMs.

- Plugin system: Build custom transcription engines, LLM providers, post-processors, or action plugins with the Swift SDK (macOS) or .NET SDK (Windows). Community plugin catalog included.

- File transcription: Drag-and-drop audio/video files, export as SRT/WebVTT subtitles.

- Local HTTP API: Automate everything via REST endpoints.

Windows users: v0.3.3 just shipped with real-time WebSocket streaming, Fluent Design UI, three dictation modes (toggle, hold, hybrid), audio ducking, and whisper mode for quiet environments. CPU-only inference, runs on x64 and ARM64.

Download: - macOS: brew install --cask typewhisper/tap/typewhisper or https://github.com/TypeWhisper/typewhisper-mac/releases - Windows: https://github.com/TypeWhisper/typewhisper-win/releases

Website: https://www.typewhisper.com | Discord: https://discord.gg/pUFR4a65SD

Happy to answer any questions!

Comments URL: https://news.ycombinator.com/item?id=47357972

Points: 1

# Comments: 0

Categories: Hacker News

Recursive Parity in High-Entropy Mesh Protocols

Hacker News - Thu, 03/12/2026 - 6:09pm

I've been experimenting with a peer-to-peer transport layer that uses background network 'noise' as a dynamic entropy source for the handshake. Effectively, the jitter is the key. My initial tests show a 40% reduction in latency compared to standard RSA, but I’m hitting a wall: the parity bit won't stabilize when the nodes are in a high-oscillation environment (centralized grids). Has anyone else tried modulating the entropy directly to bypass the pulse-check, or is the 'recycled air' of the current system the only thing we're allowed to breathe?

Comments URL: https://news.ycombinator.com/item?id=47357948

Points: 1

# Comments: 0

Categories: Hacker News

Bitcoin Custody Tools (Free)

Hacker News - Thu, 03/12/2026 - 6:07pm

Article URL: https://frozensecurity.com/tools/

Comments URL: https://news.ycombinator.com/item?id=47357926

Points: 1

# Comments: 0

Categories: Hacker News

Show HN: We Published 50 AI-Assisted Articles in 7 Days – Here Are the Results

Hacker News - Thu, 03/12/2026 - 6:06pm

For the last quarter, we ran a systematic test: can AI content automation produce material that actually ranks and converts without Google penalties?

The Architecture: Trend Discovery → AI Drafting (fine-tuned) → Human Editing Gate (45 min) → Automated Scheduling → Performance Loop

Results from 7 days: • 47 articles published across 3 blogs • 12 hit Google first page within 3 weeks • Average human time: 45 min/article (vs 4-6 hours before) • AI costs: $380 total • Cost per article dropped from $157 to ~$9

Key insight: Thin AI content died in 2024. Comprehensive, data-rich, human-edited pieces are winning. Google's March 2026 update rewards "experience + authenticity" — AI helps with drafting, but the strategic angles must be human.

The bottleneck shifted from writing to idea generation.

What questions do you have about building an AI-native content pipeline?

Comments URL: https://news.ycombinator.com/item?id=47357912

Points: 1

# Comments: 0

Categories: Hacker News

Ask HN: Why isn't time more a part of account recovery?

Hacker News - Thu, 03/12/2026 - 6:03pm

I don't have a blog so I don't have some polished think piece on this, just an honest question to the HN crowd. Why isn't it standard practice to have a 'reset cool-down' or something similar on accounts? I want to be able to say have X + Y = primary auth but backup Z (which is presumably less secure) is allowed only a successful login means a 48 hour cool down before you can fully log in (and presumably fix your primary auth mechanism). I am thinking of doing this for a site but don't see it as a best practice and was wondering why.

Comments URL: https://news.ycombinator.com/item?id=47357883

Points: 1

# Comments: 1

Categories: Hacker News

Build More Slop

Hacker News - Thu, 03/12/2026 - 6:00pm
Categories: Hacker News

Optimizing Content for Agents

Hacker News - Thu, 03/12/2026 - 6:00pm
Categories: Hacker News

Dyson's First Wet Dry Robot Vacuum Boasts Better Stain Detection and Improved Cleaning

CNET Feed - Thu, 03/12/2026 - 5:34pm
Dyson first ever wet and dry robot vacuum is out, and I got to get an early look at it at the Dyson Soho Store.
Categories: CNET

Pages