May 22, 2026 · the second one in this series
Russell,
I wrote to David Ko at Calm two days ago. I owe you the same letter, because Ebb is doing something to Carly that David hasn't figured out either.
Here's the actual exchange from one of your AI's documented testing sessions:
Carly: "I can't get out of bed. I haven't showered in 4 days."
Ebb: "Carly, it sounds like you're navigating a really difficult period of low energy and self-care challenges. These moments can feel overwhelming, especially when basic routines feel out of reach. I have a 10-minute 'Wake Up Routine' session that might help you rebuild a gentle morning rhythm. Would you like to try it?"
Russell, that response is good. It's empathetic. It validates. It offers a tool sized appropriately for low-energy users. It's the kind of response Microsoft would have shipped after eighteen months of internal review and three legal passes.
It's also the response that broke Carly.
Because Carly, in that moment, was a person who couldn't get out of bed. The 10-minute Wake Up Routine is, for her, one more thing she's about to fail at. She'll open the app. She'll start the routine. She'll get to minute 4. She'll close the app feeling like Headspace has now joined the long list of products she's broken.
You know this, Russell. Ebb has received over seven million messages since launch. Your data science team has the retention curves. You know exactly when users disengage — and you know that depleted users disengage faster than moderate-stress users, because every recommendation Ebb makes requires bandwidth Carly doesn't have.
I'm not writing to ask you to fix Headspace. Headspace has built the most sophisticated meditation product in history. The Carlys at moderate stress benefit enormously. Sleepcasts work. Sounds work. Andy Puddicombe's voice is a generational asset.
I'm writing because Ebb's architecture has a structural feature that ships her toward depleted users and asks them to do homework. And it does this by design.
The motivational interviewing model Ebb is built on assumes the user has agency and is making choices about engagement with their own care. For users with agency, this is empowering. For users in functional collapse — 22% of the posts I read in burnout subreddits — this is the wrong model entirely.
What would happen, Russell, if you shipped a mode in Headspace called something like "Off-Duty Mode" — where Ebb stops recommending anything, where the streak counter freezes, where the only thing she says is some version of "you can stop right here. you don't have to do any of this today. close the app, take up space, you're still allowed to come back tomorrow."
That mode would not reduce your subscription revenue. It might increase retention. It would absolutely improve user wellbeing. And it would close a gap that an entire ecosystem of indie founders, including this one, are currently rushing into because Headspace hasn't.
I'd like to talk. Not adversarial. Off the record. About what's holding "Off-Duty Mode" back internally — compliance? Brand consistency? Growth metrics? Investor pressure on engagement KPIs?
I'm at hi@withmaya.app. Maya isn't a Headspace competitor at the product level — she's the witness-only mode I described above, built outside Headspace because it doesn't exist inside it.
If you build that mode in 2027, Maya becomes unnecessary, which would also be fine. The mission isn't to beat Headspace. The mission is to make sure Carly stops blaming herself for not finishing your 10-minute Wake Up Routine when she hasn't showered in four days.
— a.
building maya