We typically send updates every 4-8 weeks. All past updates can be found in our archive.
As we finish H1, we wanted to write a bit more detail about our approach, progress, and plans. This update contains:
- The big picture
- Phases of work
- Our status at the end of H1 2023
- An overview of the team, investors, and advisors
The Big Picture
The problem: People are unhappy.
The opportunity: In the last 5-7 years, several thousand advanced meditators in the US have started reporting a "life-changing" ability to enter blissful meditation states on command known as jhanas. Doing so has solved opioid and heroin addictions, saved thousands of dollars on drugs and therapy, is extremely pleasurable, and makes it easier to live into values-aligned behavior like selflessness and delayed gratification. Brain images from institutions like Harvard, Oxford, Berkeley, and McGill corroborate accounts of extreme pleasure, and suggest these states are detectable.
The challenge: It's very hard to learn these states, often requiring 100s of hours of guesswork.
The solution: Biosensors that (a) monitor progress and (b) give formative feedback to make it 100x faster to learn these states. If successful, this would be a dramatic ("life-changing"), non-pharmacological, and nonaddictive improvement in wellbeing.
Why now? Consumer-grade EEG and other biosensors are newly cost-effective, advancements in deep learning allow for new pattern recognition in biodata, and breakthroughs in neurofeedback-assisted meditation suggest meditative states can be taught in record speed.
Phases of Work
To build biofeedback for these states at scale, we conceptualize the work in three phases: detecting progress, building feedback, and going to scale.
Phase 1: Detecting progress (~1 year, pre-revenue)
We need to detect and measure progress towards the jhanas using biosensors. This involves collecting labeled data across various biosensors from expert meditators and training ML models to predict when someone is or is not in jhana.
Uncertain progress represents one of the biggest reasons people give up too soon or never attempt to learn or teach these states. We plan to offer these measures of progress to design partners -- retreat providers, meditation centers, and neurotherapy clinics.
As an illustrative example: if detection of someone's progress towards the jhanas via thermal cameras works in approximately 40% of expert meditators by observing their hands heat up, then for the first time ever, teaching meditation could become like teaching yoga, where instructors watching thermal cameras could provide real-time directional feedback.
Phase 2: Feedback (~1 year, with revenue)
A measure of progress is just a start -- sufficient to land earliest customers and most enthusiastic design partners. Our aspiration is to use biosensors to also instruct meditators and/or their teachers on what to do differently in real-time.
In Phase 2, iteration with earliest partners will focus on:
- Design more effective instruction -- once detection is achieved, conventional and unconventional instruction can be evaluated, blending ideas from hypnotherapy, elite sports psychology, or emotional acting with traditional jhana instruction.
- Offer feedback in real-time -- tailoring the music or guided instruction a meditator is listening to based on biosensor data.
- Pare back consumer hardware -- we've been using neural imaging equipment costing $70K per machine to evaluate experts. Like others before, algorithms will need adaptation to work with existing, consumer-grade systems, most likely the Neurosity or Emotiv for EEG.
Eventually, the goal is a closed-loop, self-sufficient, real-time system that adapts audio, haptic, or visual (e.g. VR) instruction as one meditates. After achieving detection, we estimate roughly a year to build the first, scalable versions of a standalone product with real-time feedback and no teacher required.
Phase 3: Scale
Once a sufficiently exciting product exists, scaling will extend beyond trusted early design partners to more rank-and-file retreat providers, meditation centers, and neurotherapy clinics. Our strategy starts at the very high end, where revenue per unit can be maximized, influencers can be won, and positioning for moving downmarket is strongest.
Eventually, we'll be positioned to go direct-to-consumer, with a decision between a Peloton model (selling units with membership subscription independently) or a SoulCycle model (running premium retreats or meditation centers ourselves). Unit economics of either approach are appealing.
When asked about FDA approval and clinical reimbursement at scale: initially, no. Winning FDA clearance and reimbursement is expensive and slow, likely requiring 4+ years. Additionally, the consumer market is ultimately bigger. However, like the Apple Watch starting as a consumer product before seeking FDA approval for specific features, we might seek approval and reimbursement when timing is right.
H1 2023 Status: Nearing the End of Phase 1
In Phase 1, we chose a single metric to guide all actions: cross-subject jhana vs. non-jhana baselines classification accuracy. Successfully training models on some subjects and accurately predicting jhanas on different subjects would demonstrate (a) problem tractability and (b) proximity to offering design partners a measure of progress.
To advance this key metric, four elements are required:
- Sufficient quantity of data
- Sufficiently expert jhana meditators labeling their experiences with sufficient expertise
- Sufficient signal-to-noise ratio from hardware
- Correctly chosen analysis
In H1 2023, we set out to achieve world-class performance in all four areas:
Data quantity: By end of May, we collected approximately 60 hours of jhana data from 28 subjects across three retreats (one hosted on 7 weeks' notice). EEG deep learning for emotion recognition benchmarks show 85%+ cross-subject classification accuracy on publicly available datasets with 15-30 subjects and 30-40 hours of data.
Data labels: We recruited some of the most advanced jhana experts in the known western world, and careful interviews ensured data was accurately labeled.
Signal quality: Negotiation secured $300K of neural imaging equipment for $30K, accelerating our R&D roadmap by months.
Analysis: We built flexible ML infrastructure, iterated models systematically, and emphasized interpretable, classic ML early to build intuition and make progress before sufficient deep learning data existed.
In May, just before collecting the last data batch, iteration paid off: cross-subject classification AUC_ROC reached an average of 0.65. New data integration and model refinement are upcoming.
This positioning is strong: all four detection-building challenges were overcome faster than expected, demonstrating problem tractability. This was achieved using classic ML rather than deep learning, which advisors and benchmarks suggest will substantially improve accuracy.
We are not yet out of Phase 1 -- extending the model to newest data and triple-checking that results aren't due to confounding signal remain necessary steps.
The earliest conversations with potential design partners -- retreat providers, neurotherapy clinics, and meditation centers -- about Phase 2 collaboration are now beginning.
The Team, Investors, and Advisors
At the end of 2022, we raised $400K (exceeding our $250K target) from angels Nick Cammarata, Adam Ludwin, Coyne Lloyd, Max Bodoia, and Winslow Strong.
Around the same time, 150 neuroscientists and ML scientists applied, and we made two additions:
Alex Gruver joined as cofounder. Alex conducted ML research at Harvey Mudd, declined Facebook and other software engineering offers to secure early promotion at Bain, and most recently led go-to-market at Zoox (later sold to Amazon for $1B). His work has supercharged efforts from running retreats to ML prioritization.
Tamaz Gadaev, lead ML engineer, brings experience leading multiple ML R&D roadmaps, including for biosensors like the Samsung Galaxy Watch.
In H1, we consulted advisors from OpenAI, Kernel, AEStudios, and System2 Neurotech almost weekly. Notable additions include Rob Luke (Head of BCI at AEStudios and key contributor to the Blackrock Neurotech human-BCI decoder) and Graeme Moffat (former Chief Science Officer of Muse, the largest consumer neurofeedback device on the market).
Kati Devaney transitioned from heavy part-time support in H2 2022 to couple-hours-per-week Chief Science Advisor in H1 2023.