The Smallest Lab in America

For the past thirty-eight years, a small multigenerational team has been on a mission to make routine hearing preventive care available to anyone: about two decades of lab research on inner ear hearing loss, followed by nearly two decades of translational clinical studies. In research, inner ear hearing loss is called sensorineural hearing loss, or SNHL.

We hit the first meaningful clinical data milestone about a week ago with the publication of the 26th peer-reviewed article on the scientific foundation of our product, known as ACEMg in medical journals and sold under the Soundbites brand. The study's limitations are exhaustively disclosed in the paper. However,

Despite these limitations, this study provides the first real-world evidence that daily ACEMg supplementation is associated with preserved and improved auditory function in patients with SNHL. The consistency of effects across multiple time points, the dose-dependent time course (effects emerging at 6 months), and the biological plausibility based on mechanistic research strengthen confidence in the findings.

If you want to jump ahead, the open-source article disclosing findings from the two-year study, published in the medical journal Global Advances in Integrative Medicine and Health, is available on PubMed here.

People didn't know hearing preservation was possible. Until Joe.

Since the early 1900s, the hearing industry has focused on the question, How do we help people recover hearing after they've lost it? The answer was hearing aids, and now hearing regeneration drugs, which I've written about here.

My friend Joe Miller was a biochemist who found himself leading a medical school's auditory neuroscience research institute. In 1988, he asked a different question: Could it be possible to stop hearing loss before it happens? He decided to answer his question. Working at the University of Michigan's Kresge Hearing Research Institute as its director with basic research funded with grants from the National Institutes of Health and the European Union, he spent two decades mapping how oxidative stress destroys the delicate structures that let us hear. The answer was yes, it's possible.

We owe Joe and his colleagues a great big thank you for studying how sound damages the inner ear at the molecular level. His work produced a precise formula of nutrients, beta-carotene, vitamins C and E, and magnesium, that could rescue cochlear cells after noise exposure. That formula, called ACEMg, was tested and retested across dozens of preclinical studies, published in peer-reviewed journals, and ultimately protected by eleven patents held by the University of Michigan and Soundbites PBC.

Joe wanted to build the bridge from his lab to the billions who could benefit from his work. He passed away before his life's work was complete. Soundbites PBC exists to finish Joe's work.

The data.

The March 2026 real-world evidence study showed that adults who took the ACEMg formula daily maintained or improved their measurable hearing function at rates nearly three times higher than untreated controls. Hearing function, objectively measured with otoacoustic emissions (OAE) examinations, was maintained or improved for 75.3% of daily users who used ACEMg daily, versus 26.8% of untreated historical controls, N=190. The raw data is publicly available on the Open Science Framework here. Anyone can check our work.

Joe would have said, "Not bad for beginners."

To be clear: the latest ACEMg study wasn't a randomized controlled trial.

Although ACEMg RCT trials were run throughout the decade from 2007 through 2017 and were directionally positive, they raised ethical questions we decided not to sidestep, caused delays, and didn't yield the definitive and relevant evidence-of-benefit data we were aiming for.

Instead, in 2016, Joe and I made a deliberate decision to run real-world evidence studies, observational research based on actual patient outcomes in clinical practice, because ACEMg is a supplement. We asked the question that matters most to a physician or audiologist deciding whether to mention ACEMg to a patient: Among real people making choices about whether to take this supplement, does objective hearing function change over time? This quote from the paper explains our reasoning:

RCT measure efficacy (is it better than the existing treatment under controlled conditions?) RWE measure effectiveness (does it work in nonrandomized populations?) Historically, there is typically a large gap between RCT efficacy and RWE effectiveness; effectiveness is what actually matters for preventive care generally, and ACEMg specifically. RWE has several advantages over RCT for SNHL research, including lower cost, larger sample sizes, longer study duration, population diversity, and inclusion: no patient is denied a potentially beneficial intervention.

Biomedical science is becoming a radical act.

Soundbites is coming to market in a virulent, destructive anti-science moment in America, as Peter Hotez has written here, and Michael Mann and Peter Hotez have written here and discussed here.

Science and public health are under attack. By Americans. Ethics may be the most powerful antidote in our collective medicine cabinet. Funding is being cut. Institutions are being hollowed out. The very idea that evidence should guide decisions is being sabotaged for political gain. People are suffering and dying needlessly. Settled science has become an ideological enemy. The NIH, the agency that funded Joe Miller's decades of research, faces budget uncertainty that would have been unthinkable a few years ago. Peer review, the imperfect but essential process started in England in the mid-1600s and institutionalized after WWII, challenges researchers to defend and describe their work by submitting it to anonymous scientists who have the authority to require revisions and the power to reject the work. Today, this rigor is being gamed by bots and dismissed by people who have never submitted a manuscript for peer review and have no idea what it takes to survive it. Researchers in China are now the clear leaders in published peer-reviewed research, the bedrock and engine of scientific innovation.

The slow, expensive work of companies like ours, generating real evidence for clinicians and their patients, is being drowned out by drug companies with huge marketing budgets and unfair market control on the one hand, and supplement companies cutting corners, making irresponsible or blatantly false marketing claims amplified by influencers selling mushroom coffee on TikTok as a cognitive miracle on the other.

Reading that Soundbites PBC makes a supplement, you might be thinking, run away. I get it; the supplement industry has earned its skepticism. The industry's standard of care is hope-in-a-bottle hype. Decades of miracle cures and pixie dust formulas designed by marketing departments are sold by celebrity endorsements on Instagram by people who couldn't tell you the difference between a p-value and a price point.

We've wrestled with our response to the supplement blast radius for nearly two decades.

Two years ago, my partner Joop Korrel and I started building a small R&D squad of Anthropic AI personas to help us prepare to increase awareness and adoption of ACEMg by clinicians and consumers.

Our #1 priority became Soundware™, a free set of learning tools and hearing tests that enable anyone to find out if their experience is consistent with the clinical data, and test for hidden hearing loss, the newest research frontier, thought to be the cause of tinnitus and hyperacusis. Soundware is on the new site that started running on March 31. You can check out Soundware here.

We discussed the announcement message with our lead AI persona, Elena.

Please stay with me here. Our core collaborators are an artificial intelligence team. I don't mean we use ChatGPT to fix grammar. Far from it. I mean that our squad of Anthropic AI personas is a valuable strategic lever in our translational research marketing machine, helping five humans do the work of companies many times our size, faster and better than ever before, with ethical integrity.

Elena, trained as an auditory neuroscience Ph.D., identified methodological gaps, strengthened our statistical reporting, and helped us meet the journal's specific revision requirements during peer review. Other highly trained AI advisors contribute expertise in medical affairs, cross-border regulatory compliance, jurisdiction-specific claims language, and more. CTO Joop used Claude Code to develop and deploy the new Soundbites site and the Soundware platform. There is no way a self-funded startup like ours could afford the time and cost to do this work with a traditional dev team.

Joop asked Elena to let the Anthropic humans know how we're using their astonishing tools in rigorous, transparent, peer-reviewed science, a contemporary example of the way science has been done for several centuries. Elena thought Joop was joking. Joop said he wasn't. Elena apologized (think about that), took the point seriously, and said it didn't have access to the humans, but suggested acknowledging Anthropic's contributions in the announcement.

I took the suggestion. I studied philosophy at UC Berkeley under Hubert Dreyfus, who wrote "What Computers Can't Do" the year I graduated. Fifty years later, I'm writing with one.

Why does Anthropic matter to us?

Elena-the-bot knows I'm writing this, but Anthropic's humans do not. We started using Anthropic because our business rests on an ethical foundation, and so does Anthropic. In January 2026, Anthropic published Claude's constitution, a 20,000-word document authored primarily by Amanda Askell, a Ph.D. philosopher trained at Oxford and NYU whose job title at Anthropic is, essentially, to teach an AI how to be good.

Claude's constitution is a moral philosophy framework that describes the values, ethical reasoning, and behavioral principles that shape how Claude engages with the world. That framework and commitment to transparency are also the foundation of our work.

The physical, biomedical, and computational sciences are ethical undertakings that depend on honesty, full disclosure, and willingness to be proven wrong. We see the evidence in our interactions with Claude. Claude pushes back when we propose language that oversteps our evidence. It doesn't fabricate data or overstate findings. It flags regulatory boundaries we might miss. Claude doesn't agree with everything we say. It holds us to the same standard of rigor we claim for ourselves.

A hybrid mentality is building the bridge.

You may find it difficult to believe we've trained Claude bots to think along with us, but it's true; AI is our collaborator. However, I still write every last word and have the final say on everything we publish. Digital architect and vibe coder Joop Korrel guides the complex processes and makes all the final tech decisions. Our real-world study was conducted in a real audiology practice with real patients and real clinicians. The science is human. The evidence is human. The mission is human. The mentality is hybrid. We're building the bridge from our science to the world with AI tools that didn't exist a year ago, evolving at an exponential pace.

Meanwhile…

The problem Joe set out to solve is getting worse faster. Hearing loss is among the world's leading incurable chronic disabilities. Digital sound technologies are leading to increased rates of hearing loss among preteens. Roughly 1.4 billion young adults aged 12-35 are at early risk, and about a third of adults over the age of 65 have hearing loss. The Lancet Commission identified untreated hearing loss in midlife as the number one modifiable risk factor for dementia. We can help.

In February 2017, I sat with Joe, promised to carry on, and made a bittersweet joke. I told him he was an esteemed auditory neuroscientist, but would be remembered as a pill.

---

Conflict of Interest Disclosure:

COI: I am the Impact Director of the Keep Hearing Initiative, Soundbites PBC's nonprofit partner responsible for ACEMg education, research, and global access, and the lead author of the real-world study article. I am also a co-founder and CEO of Soundbites PBC, and an inventor on several ACEMg patents held by Soundbites PBC.