A roundup · ~8 min read
The ambient capture stack: Limitless, Plaud, Granola, and where they all forget you
In two years the ambient-AI category went from one weird Kickstarter to a whole shelf of devices and apps. Pendants, clip-on recorders, meeting-notes copilots. Each one is genuinely useful for something. None of them solves the deeper problem: the captured context is locked to whichever tool captured it.
The three shapes
- Limitless (pendant). Always-on wearable. Captures the day — meetings, hallway conversations, the things you mutter at your desk. Cloud-backed transcripts you can export from settings. Best for: people who want a record of life, not just meetings.
- Plaud (clip-on recorder). Pocket-sized recorder with strong transcription on intentional capture. More meeting-shaped than life-shaped. Best for: clinicians, consultants, anyone who wants clean structured transcripts of named sessions.
- Granola (software). Lives on your laptop, runs during meetings, produces structured notes in real time. No hardware. Best for: knowledge workers who live in back-to-back video calls.
Honorable mentions: Friend, Rabbit, Tab, Bee, the various AI badges. Most are early. Most are in the same lock-in boat described below.
The shared problem
Each of these tools captures something real — sometimes something irreplaceable — and stores it in their cloud, searchable inside their app. None of them solve what happens when you:
- Want to switch tools (you can export the file, but the AI layer doesn't transfer).
- Want to use the captured context inside ChatGPT or Claude (not the in-app assistant).
- Watch the company pivot, sunset, or quietly degrade the free tier.
- Just stop using the device but don't want to lose two years of context.
The capture is the easy part now. The persistence layer — the thing that takes what you've captured and keeps it useful across tools, models, and years — doesn't ship with any of them. That's a separate purchase, and it has to be something you own.
The portable answer
Konshus is the vault layer underneath any of these tools. Limitless has a dedicated importer; Plaud, Granola, and others come in through the documents path. The persona that comes out the other side is provider-agnostic — paste it into ChatGPT, Claude, Gemini, whatever's next. When the pendant breaks or the company pivots, the persona doesn't care.
See also how to back up your Limitless pendant transcripts for the device-specific playbook, and the hidden cost of switching AI models for why portability matters across the AI side too.