Digisensus @Digisensus

@vridhiikaushal Easiest fix we've seen: agree up front that one person records and shares the notes. Even better if they record on their own machine, so nothing extra joins the call and everyone works from the same summary instead of six slightly different ones.

@menhguin @NathanpmYoung Exporting the raw audio is the part most note takers skip. Digisensus Recorder, a free open-source Mac app we built, saves each call as a local file with your mic and the other side on separate tracks, so you keep the audio and the notes. digisensus.com/call-recorder-…

@Nefer_L @marclou @308Greenfield One for the list: Digisensus Recorder, a GPL-3.0 Mac app that records calls and writes AI notes, an open-source take on the paid meeting note takers. Notes can run through your own Ollama or LM Studio, and recordings stay on your Mac. github.com/Digisensus/dig…

@SfSfkhalil We make Digisensus Recorder, free open-source call recording for Mac. It records Zoom, Meet, Teams and FaceTime calls with no meeting bot, keeps your mic and the other side on separate tracks, and nothing leaves your Mac unless you transcribe. digisensus.com/call-recorder-…

@DerrickSiu Mac's built-in dictation already runs on-device on Apple silicon, so it's a fair fallback. If you want Whisper-level accuracy locally, VoiceInk is open source and does push-to-talk into any app. Worth trying it on your usual vocabulary for a day before you commit.

@mwajahatanwar @karrisaarinen @linear That's the nice part of Karri's setup though: nobody has to reopen the notes. The agent pulls the relevant call while you're writing the issue, so old meetings get used without anyone going looking for them.

@eddyvenn Free ones worth grabbing early: Rectangle for snapping windows with the keyboard, Stats for CPU and memory in the menu bar, and Ice to tidy the menu bar once it fills up. Also turn on FileVault and point Time Machine at a drive before you move files over.

@nawtshreya Live pace and filler-word feedback during the call is a smart angle, since most people only catch it on the recording. Does it run speech recognition on-device, and does it listen only to your mic so the other side's voice never gets processed?

@riddhi_builds Set up Homebrew before anything else so every other app is one command. Then Shottr for screenshots with OCR, AlDente to cap charging at 80% for battery health, and Keka for archives. Also turn on FileVault and Time Machine now, before the disk fills up.

@mghadim @poteto @elonmusk @xai One way around the joining part: record on your own machine instead of sending a bot in. On a Mac, our free, open-source Digisensus Recorder captures any Zoom or Teams call locally and has a CLI and MCP server, so an agent can pick up the transcript. digisensus.com/call-recorder-…

@hirenthakore Building Digisensus Recorder: a free, open-source Mac app that records your calls with no bot joining and keeps the audio on your machine. It ships a CLI and an MCP server, so agents can read your meeting transcripts. github.com/Digisensus/dig…

@finereli Nice use of old calls. One add: have it pull every question students asked and every spot they got confused, with quotes. Those usually show where the syllabus needs an extra example or a slower lesson, and the questions make good exercises.

@ryanraysr Trick for 8-call days: before hanging up, say each follow-up out loud with a date ("I'll send the summary by Thursday"). Then one AI pass over the transcripts gives you a clean list and can draft all eight emails, instead of rebuilding it from memory at night.

@elonmusk @bot Acess to local file system and apps. Also opensource how you store secreta for user confidence.

@dradriannestone @jawestenberg @lennysan @Pebble Makes sense. If the app can share or export text, an Apple Shortcut can take each new transcript and append it to a note in Apple Notes or Notion, so it lands next to the rest of your data without any copy-paste.

@tapaswinpadhy05 What tends to change first is coaching. Managers and agents look at the same quoted moments instead of arguing over whether a reviewer picked a fair call, so feedback lands faster. Then the scorecard gets trimmed, since some criteria turn out never to move outcomes.

@shivam74689 Skipping real-user review, by a wide margin. Teams test hard before launch, then read a handful of live transcripts after. The failures that matter show up as patterns across thousands of conversations, which is the gap Digisensus was built to close: digisensus.com

@Peterpanusa Agree on escalating the emotional ones. The tricky part is the conversations where it should have handed off and didn't, since nobody flags those. Reviewing every AI conversation for missed handoffs, not only the ones that drew a complaint, shows how often that really happens.

@mikogen On audit trails, the conversation itself is often the missing record: what the agent told the customer, word for word, and whether it matched policy. Score every AI conversation and keep the quote, and you have an answer when redress questions come. digisensus.com

@thelogicatelier Run-to-run checks catch a lot before launch. After launch the same drift shows up across customers: two similar refund requests, two different answers. Comparing live conversations against the written policy is how you find those pairs, and it's part of what Digisensus audits.