❓ Problem

Everyone has the same 11pm moment. You unlock your phone, scroll to Voice Memos, and there they are: 146 recordings named "New Recording 82", each one containing a thought you were certain at the time you would never forget. You have not listened to a single one since.

The capture problem is solved. Voice is now the default input for a large chunk of knowledge work, and the tools are excellent. Dictation alone has become a real category, with the voice-to-text market sitting at roughly $11.9bn in 2026and projected to reach $33.5bn by 2035. Wispr Flow turned pressing a hotkey and talking into a business that raised $280m at a $2bn valuation in August 2026. Getting words out of your head has never been cheaper or easier.

The retrieval problem is not solved. It is barely being worked on. McKinsey's number, repeated for over a decade because nobody has fixed it, is that knowledge workers lose around 20% of the working week hunting for information and chasing colleagues. That figure is about company wikis and inboxes. Nobody has even bothered to measure the loss on the stuff that only ever existed in your own head: the promise you made in a car park, the hiring criteria you worked out on a run, the objection a customer raised that you swore you would fix.

And here is the bit that makes this a business rather than a feature. AI journalling apps already exist and they are doing well. Rosebud has over 150,000 users and a 4.7 rating across 35,000+ reviews, sitting at $9.99 a month with CBT and ACT frameworks layered on top. But every one of them is pointed at reflection: how do you feel, what are you grateful for, what patterns do you notice. None of them are pointed at follow-through. They will help you understand yourself. They will not remember that you told Sarah you would send the deck by Thursday.

The gap is not transcription and it is not insight. It is the loop between what you said and what you actually did.

✅ Solution

A voice-first AI journal built for commitments, not feelings. Hold one button, ramble, close the app. The product does the rest overnight.

• One capture surface, zero organisation. No folders, no tags, no "which app should this go in". Widget, watch, lock screen, AirPods double-tap. The entire user job is press and talk, because any extra step is where the habit dies.
• An overnight extraction pass, not a live one. While you sleep, the model splits each recording into five buckets: ideas, decisions, commitments, people, and open questions. Commitments get an owner and a date. People get linked to every previous time you mentioned them.
• A five-minute morning briefing that is a queue, not a document. Approve, snooze or bin. Approved items become drafted follow-up messages, tasks pushed to your existing to-do list, and deep-work blocks dropped into real calendar gaps.
• A memory that connects backwards. "You said something similar about pricing in March and again in June" is the feature people will not switch away from. Long-context models make this trivial now and it is the only part a competitor cannot copy in a sprint, because it compounds with usage.

The wedge is deliberately narrow: help one founder keep one promise they would otherwise have dropped. The platform underneath, capture plus structured memory plus an action layer, expands into weekly reviews, relationship reminders and eventually team knowledge capture.

📊 Key Numbers

Market size

• The digital journal apps market is worth $6.34bn in 2026 and forecast to hit $9.64bn by 2030 at an 11.1% CAGR, with AI-assisted journalling the fastest-growing segment.
• Stack the adjacent voice-to-text market on top at $11.9bn in 2026 and you are looking at roughly $18bn of annual spend already flowing to capture and reflection tools that stop short of doing anything.
• The comparable that matters most is Granola, which does this for meetings and raised $125m at a $1.5bn valuation in March 2026, up from $250m less than a year earlier. Otter, the incumbent it is eating, has scaled past $100m ARR with under 200 people. Meetings are maybe a third of the thoughts a founder has in a day. Nobody owns the other two thirds.

Serviceable market

• There are 36.2m small businesses in the US, and roughly 29.8m of those are non-employer firms: solo consultants, freelancers and one-person operations. That is the exact profile with no assistant, no chief of staff, and nobody to catch dropped commitments.