I use voice dictation every single day. Wispr Flow is great, but $15/month adds up, and I kept thinking: everything it does could run locally on my Mac. So this week I built it. Meet OpenFlow. 🎙️ Hold a key, talk, release, and your words appear wherever your cursor is. Any app. Editors, Slack, the browser, your terminal. The best part: it runs 100% on your Mac. No cloud. No subscription. No account. Your voice never leaves your machine. What's under the hood: 🔹 NVIDIA's Parakeet model transcribes at 35x realtime (it even spells "kubectl" correctly) 🔹 Whisper handles other languages 🔹 Apple Intelligence cleans up your "ums" and false starts, fully on device 🔹 Custom dictionary, searchable history, lives in your menu bar Here's the thing I want you to take away: I built the entire app with Claude Code. Native Swift, two speech engines, an on-device cleanup pipeline, a landing page, and it even generated its own app icon. A year ago this would have been a month of work with a Swift developer. This week it was a weekend. It's free and open source (MIT), so you can download it, read every line, or fork it: 🌐 Website: robj1925.github.io/open-flow 💻 Code: github.com/Robj1925/open-flow I'll be posting a breakdown of how I built it if there's interest. Question for you: if typing was never the bottleneck, what would you build or ship faster? Drop it below. 👇

Hello everyone, I'm currently experimenting with using Claude Code to edit my YouTube videos, and I'm hoping to connect with others who might have tried something similar. My workflow involves recording a talking-head video, then using Claude Code with OpenAI's Whisper models to generate a transcript for editing. However, I've hit a major bottleneck: the transcription is frequently inaccurate, and worse, it often cuts off mid-sentence. This leads to poor edits and a lot of manual cleanup that defeats the purpose of automation. I'm looking for any advice from the community on: 1. Better free/open-source transcription models – Are there alternatives to OpenAI Whisper that are more accurate and handle long-form video better? 2. Tools that solve the "cutting off" problem – Has anyone found a solution that reliably transcribes an entire video without truncation? 3. Claude Code integration – If you've built a similar workflow, what was your approach? Any tools, skills, or MCP servers you'd recommend? I'm specifically looking for free/open-source solutions that can run locally. Any guidance, lessons learned, or resources would be greatly appreciated. Thanks in advance for any help!

I know claude was known for being pretty natural and I can confirm it from my experience. But I wonder if GLM or some other Fable 5 Competitors could be better at that. Also how does a model become good at copy? Is it the most trained on more quality data? Explain this to me cause I don't get it as much

I am currently speaking with leads and the service I get asked for the most is website builds. I am trying to figure out the best way to package and price this, and I would love input from those of you who are actively selling. Right now I am considering two models: Option 1: One and Done. I charge a one time fee for the build. I walk the client through getting their own domain. They handle renewals and management after that. No ongoing work for me. Option 2: Recurring (Monthly Only)No upfront fee, only a monthly charge. I purchase and hold the domain, or manage it for them. I handle renewals, hosting, and any small updates. It brings recurring revenue but adds management complexity. For those of you who have actually sold website builds to small businesses, which model has worked better for you? Did you start with one and switch to the other? Any pricing or retention insights you can share would be very appreciated. Thank you, just trying to learn from people who have been in the trenches.

Hopefully this is a learning lesson for everyone, recently swapped from zapmail to an inbox provider that offered a much cheaper model (subscription + domains, inboxes are free). I've come to learn that over the last 4-5 weeks, the deliverability issues my campaigns have experienced is entirely on them, because the sending IP is burned (verified by checking the reason emails bounced from instantly). Let this be a lesson to not be cheap and let it be more expensive.