Technology should give us more of our lives back

Ambient AI stays present in the room, hears the context of a day, and helps without needing a prompt.

Each era of computing brought machines closer to people: point and click, tap and swipe, ask and answer, and now presence. Speech understanding and on-device processing finally make presence practical, at the moment agents are ready to act but lack real-life context.

That gap is not felt equally. For most people, a forgotten promise is an inconvenience. For people with ADHD, it is a pattern with a real cost, paid in missed bills, dropped commitments, and lost opportunities.

More than half of American adults with ADHD were diagnosed as adults, often mid-career and after years of building their own workarounds. They already pay for planners, focus apps, and coaching because going without help costs more. Yet every one of those tools depends on the user remembering to open it. Legend is the first that hears the day and follows through on its own.

Our first milestone is a million people with ADHD. In the US alone, 15.5 million adults have a diagnosis^[1], and 6.2 million of them live in households that can comfortably afford Legend^[1].

Starting with ADHD also answers the stigma question, because the need is visible and the pendant reads as assistive technology.

What works for the hardest case tends to work for everyone. OXO Good Grips began as a peeler for hands with arthritis, and closed captions were built for deaf viewers before becoming the way millions watch TV. From ADHD, we expand to everyone who struggles with executive function, then to overloaded knowledge workers, and then to everyone.

Legend is $499 per year, flat. The device is free, usage is unlimited, and hardware upgrades are free. The subscription pays for an ever-present chief of staff for your life. The price point is premium to match the impact it has. Context retention and proactive task execution are how people unlock that value.

Premium is a product decision, not a pricing tactic. People pay $300 for Oura or Whoop for sleep and strain. Legend holds commitments, remembers what matters for professional and social life, and takes work off the plate. Locking in at premium lets us deliver the best possible product, without rationing quality. Zach ran this experiment at Fieldy: $99 upfront and $179 a year. Staying profitable meant using worse models and components that were just good enough, and the experience suffered.

The device BOM is approaching $100 and total AI cost per user is higher than $100. At $499, the target 60 percent gross margin is a scale story: AI costs fall by orders of magnitude as they cross thresholds, and when they do, we become more profitable. iPhone entered at a multiple, proved it needed to exist, and phones are $1,000 forever.

At 1 million subscribers at $499, Legend is earning $499 million per year. Renewal is earned by ensuring losing Legend feels worse than paying for it. As an extra incentive, they get the latest hardware, and can donate the old one to a friend or family member.

What we will not do is subsidize the subscription with commerce or ads. This type of revenue would make a sensitive product feel manipulated, with your most personal asset, your conversations, as the inventory. The subscription is the business model, and trust is what keeps the pendant out of the drawer.

Even a wearer who pays $0 is valuable because wearing drives word of mouth and the network effect. The goal is ubiquity: something for everyone, fashion forward.

The form is jewelry. Objects of value. Something people already choose to wear all day with no new behavior to learn. The first product is a pendant. Sitting at the heart of where the conversation happens.

Ultimately, form will be up to personal preference and we’ll continue to expand our product line to meet the culture’s demands.

The retention mechanism behind the wear time is to make something that keeps being worn. Power users wear it the entire day, because they don’t have to remember to pull it out of a drawer.

The distribution mechanism comes from word of mouth, driven by a fashion statement.

This only works if the trust model is structural.

Legend runs on a chip certified to PSA Level 3, the highest security tier for a wearable. Your data at rest is protected by AES-256 envelope encryption, so deletion is clean and complete. Audio is not stored after processing.

Private zones are excluded before anything is ever indexed. Every agent gets access only through OAuth, with explicit scopes that must be approved. Every grant can be revoked at any time. Every query is written to an audit log that can be inspected.

Model providers process requests under zero data retention, and every request is anonymized before it leaves your device. People can sign in through Apple private relay, so even their account cannot be traced back to them.

For us, privacy is a design decision, not a policy page.

Follow-through compounds. Every day adds context, every kept promise makes the next day easier, and starting over somewhere else means starting with no history.

Legend is in a controlled, paid closed alpha designed to maximize our surface area for learning and retention. 170 people have paid to join the program. In the last two weeks, trusted testers have begun receiving and are wearing the device. Our strongest early users are executive women with ADHD. It is helping them remember conversations, organize what is on their mind, and notice what needs attention. Together we’re developing an assistant that can help them do the work, too: preparing next steps, checking in when a decision is theirs to make, and taking things off their plate rather than adding to it.

Every recording technology since the 1888 Kodak has challenged social norms. Society tends to settle these things quickly: in the early 2000s, someone talking into a Bluetooth headset looked like they were talking to themselves, and within a few years no one looked twice. The same thing has happened more recently with AI notetakers in professional environments.

Hardware, models, and raw transcripts commoditize quickly, and none of them create a network effect on their own. Trust is the asset that compounds.

That is also why incumbents are unlikely to fight for this position directly. A platform that profits from attention cannot credibly promise a person that their private life will stay theirs, and we expect incumbents would rather an independent company carry this responsibility than absorb the regulatory and social backlash themselves.

We are repeat founders building at the intersection of voice AI and consumer scale. Saba was on the founding team of Orai, an AI speech coach that reached 300,000+ users and raised $3.3M, previously led the startup pipeline at Techstars, and co-led an $8M pre-seed fund. Zach co-founded Fieldy, a hands-free wearable AI notetaker that reached $5M in annualized revenue, and was a growth hacker at NGL, which reached 300M+ installs. AK exited an award-winning mobile app development studio, has sold ~$5M in luxury jewelry on Shopify with his wife and held senior roles at Google and Meta, where he authored a patent and leveraged AI to drive revenue and customer growth.

All three of us have ADHD. We are building for ourselves and the community around us first.

We are raising $4M, far less than the $30M+ many pre-launch hardware companies raise, because we’ve spent years in R&D to get our supply chain ready for mass production. Our North Star is 100,000 subscribers, about $50M in ARR, within 18–24 months.

The capital goes to three things: a full-time hardware team in Shenzhen to keep quality high as production scales, a small ML team that continuously improves our models as usage grows, and an expanded trusted tester program alongside the ADHD communities that are already looking for Legend. During this time, we will prove retention through careful refinement of both hardware and software ahead of a major launch funded by the next priced round. This is the ramp that will get Legend into the hands of over 1M people within the next 3 years.