The arc of success digital health startups is a thoughtfully crafted wedge grows to become your moat, which then defines your right to exist and ability to have human impact at population scale.

Andrea Ippolito has built Simplifed into a national scale telehealth platform delivering lactation and post-partem support in all 50 states. She has contracts with all of the major payers and just raised a $10M series A. When she started Simplifed 6 years ago, the ACA Lactation mandate was almost a decade old, but fewer than 15 percent of lactation consultants billed insurance and less than 5% of those claims got paid. In this interview she shares the steps that she took to get here.

Clip #1 Building for Understaffed Healthcare Mandates.

One reason that I asked Andrea to be my first guest on Right to Exist is that lactation support is structurally similar to many other spaces, pediatric asthma, osteoporosis, menopause, hormone replacement therapy, and others.

What does it take to build for a population or condition that is understaffed or under-resourced by conventional enterprise healthcare?

Clip #2: Integration Friction vs. Product Perfection

One of the common questions that founders have at pre-seed and seed is how much to invest in integration rails with existing healthcare infrastructure.

In Andrea’s case, the signal from OBGYNs was that smooth integration with care flows, and more importantly, feedback loops about how referrals had been handled, would be the driver for adoption.

Clip #3 When Lactation Support Drives more than Lactation Outcomes

Part of the secret sauce for Andrea, was the realization that Simplifed’s contact area with patients was much broader than that of the intermittent visits to OBGYNs who drove the initial referrals.

This meant that they were able to catch things like post-partem depression, preeclampsia, and gestational diabetes and route the patients in need of additional care back to the clinic.

Subscribe below to watch the entire 30 minute conversation.

Read more

Right to Exist: Episode 1, Lactation Support at Population Scale

• Andrea Ippolito - Simplifed

• 30 Minute Video / Podcast

• The most common category error made by founders, investors, and operators in digital health is believing that GTM, venture design, and stakeholder alignment are distinct disciplines. The best operators execute these as an integrated whole.

• Right to Exist is a new interview series on how founders earn their project’s right to exist, from the wedge to the moat to the human impact they aspire to.

• The ACA in 2010 required health plans to cover lactation support at no cost. Despite this, when Andrea started Simplifed in 2019, fewer than 15% of lactation consultants billed insurance, and only 4% of those claims got paid. Andrea built the 50 state clinical practice to deliver a mandated benefit new moms desperately needed that almost nobody was serving.

• We get into why interoperability became her moat, the five-body problem of enterprise healthcare adoption, what she would sequence differently if she were building today, and her advice for other orphaned and understaffed clinical domains like peds asthma, osteoporosis, HRT, and others.

How to Own Risk Without Getting Pwned (Part 1 of 2)

• 20 Minute Read

• Start-ups love to put “owning risk” into their roadmaps.

• Bright Health went public at $11.2B and wound down its insurance business two years later on a 101.3% medical cost ratio.

• Meanwhile, Kinetic, an NYC wearables company, started with a buzzer on a warehouse worker’s belt and is now living the VC startup dream.

• It created a new multi-billion dollar category that it now dominates.

• Same phrase, “taking risk,” two outcomes that could not be further apart. This guide separates the four distinct things people call risk, then hands you a five element framework to determine how and when to take risk, the failure modes and the multi-billion dollar startups that fell into them, and four diagnostic instruments to place any company on the map.

• In Part 2 (live July 15th), we’ll use live examples of 4 very different companies tackling high risk pregnancy outcomes and close with a nine-question diligence memo to evaluate both your own work and that of your portcos.

Venture Funding for Fun and Profit: Part II, The Process

• 40 minute read

• What does it actually take to raise a 6/7 figure angel or pre-seed round?

  • Where do you find angels?

  • What does a first e-mail or call look like?

  • How should the overall process run?

• I provide the framework that I’ve taught for years along with detailed instructions on how to execute, measure, and course correct a raise that isn’t going as planned.

• I’ve also done original research here to help you manage both expectations and execution on your fundraise - especially if you’re outside of NYC or SF. Tabulated and scored lessons from 50 verified first-person raises, roughly 4,500 investor names worked, more than $150M closed across three continents.

• This is the seven-phase process built from it, from filing your 83(b)s to chasing the last wire that won’t land, written for founders who don’t have the network the standard advice quietly assumes.

All three are behind the paywall for premium subscribers. As I continue to experiment with the Substack format, my goal is both provide free actionable content and give folks who a taste of how electric our in-person events these past years have been.

If you know a founder, operator, or investor who would use any of these, forward it their way.

Share

Only New Mistakes.

— Vadim

Contents

1. How Kinetic created and then won a category — the worked example before the framework.

2. The framework for effectively owning risk — five conditions, three stages, five failure modes, four diagnostic instruments.

3. Value-based care economics are often bad. Yours do not have to be — owned-loss economics vs. savings-share economics.

4. Three failures, three modes triggered — Bright, Cano, Babylon.

5. CareBridge — a narrow-population exit consistent with the framework.

Part 2 ships July 15: maternity through four different concepts of risk and venture design (Fetal Pillow, Maven, Pomelo, Ouma + MedArrive), a nine-question diligence memo for evaluating risk-bearing projects, and the venture-scale path through the framework.

TLDR: Take responsibility only for losses you can define, measure, materially influence, price with credible data, and survive when the model is wrong.

1. How Kinetic created and then won a category

My favorite startup example of a both wearables and owning risk is almost never discussed in the healthtech startup ecosystem.

Kinetic Insurance is an NYC company living the VC-funded dream. It created a new multi-billion dollar category that it now dominates. It did so by providing ground truth capable of dramatically beating the relevant actuarial tables and an effector arm for delivery both human and economic impact at scale. Most healthtech companies that announce a path to risk-bearing never even come close to reaching those destinations.

The Kinetic Reflex sensor detects high-risk lifting and bending postures among warehouse workers. When it sees one, it provides haptic feedback to the worker in real time. The accompanying software shows managers where risk is clustering by department, job type, and time of day, creating something employers rarely have: ground truth about how work is being performed. The employer value proposition is therefore twofold: fewer injuries and lost workdays, plus a usable map of the behaviors likely to produce the next injury. The product is not merely a warning buzzer; it is a sensor-to-coaching-to-management system aimed at a specific workers’ compensation loss.

Kinetic launched in 2014 as a workplace-safety wearable. By November 2021 it had launched Kinetic Insurance, a workers’ compensation program sold and serviced by Kinetic, with Nationwide’s excess-and-surplus and specialty organization underwriting the policies. Unsurprisingly, Nationwide Ventures has been involved with Kinetic since its 2020 series A.

Most healthtech companies that claim a path to “taking risk” never leave the much lower-margin strata of vendor economics. Some reach outcomes-priced contracts where a performance fee can be lost while the payer still owns the medical claim. Some accept capitation and bear delegated medical-cost downside. A much smaller group administers an insurance product. “Taking risk” collapses these into one sentence; this field guide separates them.

What Kinetic did is still the rare thing: it satisfied the operating and evidentiary prerequisites in sequence, in public, before moving into a delegated insurance role. The sequence is applicable and deeply instructive even when the strategic endpoint differs.

The loss event Kinetic addressed (strain and sprain claims in industrial work) is unusually well-suited to risk-bearing: a compensable injury is observable, maps to the policy and employer, settles inside the policy period, has decades of workers’-comp actuarial baseline, and posture, lift technique, repetition, and workstation design are operationally addressable.

In Section 2, we’ll work through these properties as the five conditions for a successful risk program, how Kinetic’s venture design passes each, and what new lessons new teams can apply to their own work.

Read more

Healthtech venture is hard, a fact I’ve learned firsthand as an operator, accelerator director, and investor. I’ve watched too many good teams repeat avoidable mistakes others have already paid for. I write Healing Healthtech to distill research and the experiences of top-tier operators into actionable tactics and frameworks, so we can all aim to make only new mistakes.

• Part 1 covers what I published this month on Substack.

• Part 2 covers NYC Healthtech events that I’m hosting.

• Parts 3 covers the most impactful and informative things that I read in the past month.

• Part 4 covers other NYC healthtech events worth a spot on your calendar.

Subscribe now

Subscribe for a twice-monthly dispatch of free and paid field guides meant to help you build more deliberately toward something better in human health. And if it’s useful, please forward it to other healthtech builders in your network.

Share

Warmly,
Vadim
Only new mistakes.

✍️ Part 1 — What I Wrote This Month

• 🎬 Venture Funding for Fun and Profit - Part 1 The Narrative — Free — 20 minutes

  • The heart of this video is how to build a successful pitch deck for raising angel and preseed funding.

  • More broadly, its a workshop that I’ve given many times focusing on the meta narrative that creates a fundable company.

  • My approach to pitch decks and narrative works - 80% of my final 12-team accelerator cohort raised follow-on funding within 3 months of graduation. One got into YC and another has already exited. Historically, we had 300+ teams raise 2.7B into 44 exits.

• 📊 Join an Accelerator or Nah? — Free — 10 minute read

  • When a program is worth the equity and when it isn’t; the comet distribution of accelerator outcomes, the hedging-accelerator model, two diligence checks for founder, and my advice from getting the most out of a program that you decide to participate in.

• 💰 Venture Funding for Fun and Profit - Part 2 The Process — Paid — 45 minute read

  • Deep dive into the process of running an effective raise.

  • How to find, qualify, engage with, and close angels and early funds.

  • How to measure and re-calibrate your raise against benchmarks as you go.

  • I combine my own anecdotal experiences, data from Docsend, Carta, and Angellist, along with the writings of the best venture investors of the past 20 years into a deeply actionable and practical field guide.

• 📱 SMS vs. Apps for Effective Digital Health Interventions — Paid — ~30 min

  • A field guide on when to reach for SMS versus an app in chronic care, built from 75+ clinical trials and operators who delivered at population scale.

• 🏛️ Anthropic’s Future Looks More Like Oracle Than Google — Free — ~15 min

  • Argues open weights are compressing frontier API margins, with OpenAI on a Cisco path and Anthropic on an Oracle one, and what that means for healthtech builders.

• ⚖️ Will My SMS Architecture Survive Scrutiny or Create Liability? — Free — ~5 min

  • The free, abbreviated edition of the SMS field guide: the four regulatory regimes governing healthcare SMS and six tripwires that turn a compliant program into a class action.

📅 Part 2 — My Events This Month

• 🍽️ Self-Funded Employer Founders & Funders Dinner | Tribeca, New York City

  • If you’re building or investing in angel or pre-seed self-funded employer healthcare, please apply to attend. Sponsored by Rho.

  • Historically, 70%+ of the founders at my F&F dinners have follow-on meetings after the event with 1 or 2 checks resulting.

  • 🗓️ Tuesday, July 7, 2026

• ⌚ HealthTech Watch Club | TIME+TIDE Watches NYC, 460 Broome St, New York City

  • One Health Collective x MarkOne Health x Time+Tide gather NYC’s healthcare-ecosystem watch enthusiasts for horology, no pitches. Wear your most interesting mechanical, automatic, or quartz piece (no smartwatches).

  • 🗓️ Thursday, June 25, 2026

✍️ Part 3 — The Best Things I Read in Healthtech This Month

• 📉 340B Wasn’t Stolen From Rural Hospitals. It Was Built On Their Graves. — — ~15 min

  • Follows where the 340B spread actually landed, naming five publicly traded chains and PBMs rather than the safety-net communities Congress wrote the program for, plus the dual-classification loophole that let urban hospitals qualify as rural.

• 🐄 How Farmers are Smarter than Doctors — — ~3 min

  • Uses livestock veterinary practice to show how far human medicine lags on women’s reproductive and vaginal health.

  • This one really isn’t as off-color as it might seem. A founding patriarch of reproductive endocrinology once gleefully walked me through how both the devices and procedures in his field owe more to animal husbandry than human medicine.

  • Lara is a heck of a clinician scientist and well worth the follow and read.

• 🏗️ It’s time to build (in Medicaid) — — ~13 min

  • Makes the operator case for building inside an $880B program most founders avoid, with field notes on navigating the bureaucracy.

  • Youssef isn’t an idle bystander. As healthtech investment principal at AlleyCorp, he led investments into several current rocket ships in the portfolio that work with Medicaid-heavy panels, including Altitude, Birches, and Stepful.

• 🧠 Mental Health AI Is Operating Without Clinically-Informed Safety Standards. That Should Alarm Us — — ~15 min

  • Proposes a seven-part architecture of enforceable safety standards for mental health AI, grounded in documented patient deaths and existing clinical frameworks like ASL-MH.

• 🩺 The Dangerous Practitioner: on miscalibrated confidence in medicine — Graham Walker MD — ~10 min

  • Defines the “dangerous” clinician as one whose confidence outruns competence with no outward signal of the gap, and shows why medicine’s safety systems (consults, escalation, supervision) all depend on someone recognizing where their competence ends. The calibration frame maps directly onto the question of when to trust a confident AI.

  • I’m placing this Scott and Graham’s piece above right next to Byron’s piece on purpose, as a counterpoint: the case for caution right before the case for urgency.

• 💊 Our Penicillin Moment — — ~7 min

  • Argues the evidence for medical AI’s benefit is now strong enough that continued delay carries its own moral cost.

  • Byron is CMO of Doctronic, which runs the first state-approved autonomous AI prescription program in the U.S., currently limited to prescription renewals in Utah under a graduated-autonomy model.

• 🎓 The Old World Is Dying — — ~12 min

  • Career advice for the class of 2026 entering an AI-disrupted job market, useful for anyone weighing where durable work comes from.

  • I graduated during the last recession and now teach digital health venture design to graduate students pursuing MPA, MHA, and MPH degrees. Like many, I’ve been struggling with what to advise new alums in the face of AI emergence. This piece reached me by way of Youssef (above).

• 💸 How much I’ve made angel investing — an update — — ~5 min

  • Publishes a full deal-by-deal angel track record (0.89x cash returned on $2.9M across 53 investments) and benchmarks it against Carta fund data, a rare unvarnished look at returns.

  • Halle is easily the best-connected US healthtech angel of the past 25 years, and her openness about portfolio performance (or lack thereof) is badass.

  • Get her book.

📅 Part 4 — NYC Healthtech Community Events

• 🤖 Healthcare in the Age of AI | New York City

  • UJA’s Healthcare Division hosts a panel on how AI is reshaping care across clinical, technical, and business functions.

  • 🗓️ Monday, June 15, 2026

• 🏛️ Civic Health Forum 2026 | Civic Hall, New York City

  • A half-day convening of providers, policymakers, investors, and founders built around one question: how do we keep people healthy before they enter the healthcare system? Co-hosted by Fedcap, Digital Health Hub Foundation, and StartUp Health.

  • 🗓️ Tuesday, June 16, 2026

• 🍹 NY–IL Digital Health Community Happy Hour | New York City

  • An evening bringing together Israelis and allies building healthcare software, with rapid-fire startup pitches alongside the drinks and networking.

  • 🗓️ Wednesday, June 17, 2026

• ❤️ An Evening of Courage, Connection & Purpose (KuraHealth) | New York City

  • An intimate talk on grief and resilience at the residence of economist Nouriel Roubini, featuring Kevin Hines, one of the few to survive a jump from the Golden Gate Bridge.

  • 🗓️ Thursday, June 18, 2026

• 🎨 Art in the Education Lab: “The Shape of Memory” by Maya Ciarrocchi | New York City

  • The unveiling of the 2026 Art in the Education Lab commission by artist Maya Ciarrocchi.

  • 🗓️ Thursday, June 18, 2026

• 🥂 The Art of emPOWERment: Ultimate FO Happy Hour | New York City

  • The 13th event in the Art of Empowerment series, a curated family-office happy hour where founders, allocators, and investors discuss where they put capital and energy to work. Hosted with EisnerAmper.

  • 🗓️ Thursday, June 18, 2026

• 😎 Summer Chill (HIVE) | New York City

  • A HIVE healthcare networking evening for clinicians, administrators, and healthcare leaders built around intentional conversation and connection.

  • 🗓️ Thursday, June 18, 2026

• 🧠 BrainHealth: VC Panel + GTM Bootcamp | FINN Partners, 1675 Broadway, Midtown NYC

  • A practical session on go-to-market strategy for brain-health innovation, with lessons from CEOs and institutional BrainHealth VCs. Moderated by Richard Hatzfeld of FINN Partners.

  • 🗓️ Monday, June 22, 2026

• 🚀 AI Engineers Presents: Founders to Follow | New York City

  • Intimate conversations with NYC startup founders about their journeys, including Tyler Johnson (Tennr) and Aditya Maru (Blacksmith).

  • 🗓️ Wednesday, June 24, 2026

• ⚖️ Digital Health Market NYC Symposium | Nixon Peabody, 55 W 46th St, New York City

  • A CTeL and Nixon Peabody roundtable bringing together digital health leaders, investors, and policymakers on the regulatory, data, and AI deployment dynamics shaping market growth and M&A.

  • 🗓️ Thursday, June 25, 2026

• 🌇 Pepper Community Hang: Vessel Sunset Walk + Tapas | Vessel, Hudson Yards, New York City

  • A golden-hour walk around the Vessel followed by tapas and drinks, hosted by Pepper NYC’s health/tech community.

  • 🗓️ Thursday, June 25, 2026

• 💻 Healthcare x AI Hackathon | NYC | New York City

  • A 1.5-day hands-on hackathon to build real AI tools and agents for healthcare, for engineers, PMs, and founders working in regulated spaces. Partners include Baseten, Lovable, ElevenLabs, Upfront Ventures, Frist Cressey Ventures, and Aegis Ventures.

  • 🗓️ Friday–Saturday, June 26–27, 2026

• 🏙️ Health Tech Startup Networking Rooftop Happy Hour | New York City

  • Rooftop networking for health tech startup professionals over drinks and skyline views.

  • 🗓️ Tuesday, June 30, 2026

• 🧬 NeuroNYC GP/LP Happy Hour | New York City

  • A curated evening for GPs and LPs active in or exploring neurotech, mental health, brain-computer interface, and neurodegenerative disease, held the night before the NeuroTech Summit.

  • 🗓️ Wednesday, July 15, 2026

• 🌟 Health2Tech NYC | New York City

  • An exclusive digital health gathering hosted by Bailey Morgan, with a guest talk and extended networking.

  • 🗓️ Thursday, July 16, 2026

• ⚓ Ship It 2026 (Healthcare Software Engineering Conference) | New York City

  • An invite-only Out-Of-Pocket summit for software engineers in healthcare: 80 curated guests, 12 hands-on workshops, and tactical talks over panels. Two-plus years of healthcare experience expected.

  • 🗓️ Thursday–Friday, September 17–18, 2026

  • ⏰ Deadline to apply: July 1, 2026

This deck and presentation are keyed to my core domain of digital health, but I’ve personally used this exact process with startup in many other industries.

The companion pieces for this presentation are below.

📊 Join an Accelerator or Nah? — Free — 10 minute read

• When a program is worth the equity and when it isn’t; the comet distribution of accelerator outcomes, the hedging-accelerator model, two diligence checks for founder, and my advice from getting the most out of a program that you decide to participate in.

💰 Venture Funding for Fun and Profit - Part 2 The Process — Paid — 45 minute read

• Deep dive into the process of running an effective raise.

• How to find, qualify, engage with, and close angels and early funds.

• How to measure and re-calibrate your raise against benchmarks as you go.

• I combine my own anecdotal experiences, data from Docsend, Carta, and Angellist, along with the writings of the best venture investors of the past 20 years into a deeply actionable and practical field guide.

Hi, I’m Vadim.

Healthtech venture is hard, a fact I’ve learned firsthand as an operator, accelerator director, and investor. I’ve watched too many good teams repeat avoidable mistakes others have already paid for. I write Healing Healthtech to distill research and the experiences of top-tier operators into actionable tactics and frameworks, so we can all aim to make only new mistakes.

Subscribe for a twice-monthly dispatch of free and paid field guides meant to help you build more deliberately toward something better in human health. And if it’s useful, please forward it to other healthtech builders in your network.

Drafted with support from Claude as a research assistant, with GPT and Gemini as a red-team.

Angel rounds are achievable. Anywhere in the world.

Every bubble in that chart is a real raise from a verified founder account [22]. Read it as a promise and a warning at the same time: angel rounds close at every size and on both sprint and marathon clocks, and none of them closed without sustained, disciplined effort against a wide funnel.

Angel investors aren’t magical. They’re mid-to-late-career professionals who remember what it was like to be a fresh builder, and who are now interested in and able to bet on younger founders in exchange for some of those founders’ upside. Most angels underwrite 1-3% of the deals they see, when they feel they have met a pitch that embodies their beliefs about founders, problem, and solution spaces [17]. Roughly 500,000 angels will fund some 60,000 companies around the world this year [38]. The goal of this guide is to help you find yours.

So when an angel raise fails, it fails for one of three reasons:

1. The founder never ran a disciplined process.

2. Something is fundamentally wrong with the plan or the proposed business.

3. The founder didn’t adapt the plan, the pitch, and the angel targeting as the process taught them things.

The first and third are fixable with method, and this guide is the method.

Over 15 years, my venture partner and I led the Data Future Lab at NYU, where we supported 300+ teams to $2.7B in capital raised and 44 exits; this guide exists to put that playbook in the hands of the thousands of founders I’ll never meet. It is keyed to my core domain of digital health, but I have personally coached teams running these exact strategies on thin films for solar panels, fintech SaaS platforms, and math tutoring programs for charter schools.

Part I — The map

1. The stages — the capital ladder, what each round buys, and the markup principle that runs through all of them.

2. The strategic fork: growth versus viability — know the plan at the start of the round, lock the choice with no less than 50% runway remaining, and run the next-round-investor diagnostic before you commit to growth.

3. Raising less money is not easier — the small-round trap, quantified, with the healthtech amplifier.

4. What makes a fundable healthtech company — the scale math, the operational machine, the acquirer pool, and the credibility tells.

5. Accelerator or nah? — when a program is worth the equity and when it isn’t; the comet distribution, the hedging-accelerator model, two diligence checks, and my advice from getting the most out of a program.

Part II — The process (phases run in parallel, not in sequence; a given investor sits at a different phase on any given day)

• Phase 0: Set the table — vesting, 83(b)s, IP assignments, cofounder splits, option pool, the data room.

• Phase 1: Build your list — finding 80-120 qualified names when your network is thin; the warm-intro priority stack.

• Phase 2: Before you’re raising — meeting investors months ahead; lines, not dots.

• Phase 3: Decide the round — price, instrument, size, founder salary, the procurement-cycle runway.

• Phase 4: Choose your outreach pattern — the sprint versus the marathon, the forwardable, the investor-side comparator funnels.

• Phase 5: Work the conversation — format for your first call with a new prospect, cadence, reverse diligence, reading and responding to the slow no.

• Phase 6: Diagnose and calibrate the raise — the two-stage instrument, reading the clock and the funnel, asking the angels who passed, recalibrating before you scale.

• Phase 7: Close and collect — stack checks behind your first investor, collect every yes the day it lands, the staged close, the SAFE stack, strategic checks.

Part III — The data — aggregate data sources organized by what you can dig into, with link-outs.

References

Read more

This document is a free sample. It’s part of larger three part series on

Venture Funding for Fun and Profit

(Free Subscribers) How to build a pitch deck and startup narrative
(Paid Subscribers) How to find investors and run an angel or pre-seed process

Subscribe now

TLDR: Do your best to get into a top decile accelerator. They’re genuinely helpful. Otherwise, you’re better off spending those 3-4 months focused on fundraising, product, and biz dev.

For an under-networked clinical founder, an accelerator is often pitched as the alternative path through friends-and-family and angel: a program that promises to manufacture the qualified investor funnel, the health-system intro graph, and sometimes a first pilot in one move.

The largest global study (8,580 startups, 408 accelerators) finds that participation roughly doubles the probability of raising venture capital, from a ~3% baseline to ~6.4%, and adds about $1.8M to first-year funding [1].

But subsequent matched-pairs research (~900 graduates vs ~900 non-participants) shows the first-year bump dissolves: over the three years after graduation, graduates raised about $9M less in total funding, were more often acquired in talent-driven acqui-hires than technology acquisitions, and shut down earlier and more often [2].

A long tail, not a bell curve.

Underneath both findings sits a comet distribution of accelerator outcomes. Top-tier programs delivered large gains, second-tier delivered beneficial learning, and bottom-tier programs actively inhibited venture development along some dimensions [3]. The accelerator average (+2x funding likelihood) is the wrong unit to plan against; the specific program is.

The vantage for what follows: I ran the Catalyst and Innovention Accelerators at NYU Tandon Future Labs for four years, and Steve Kuyan, now my venture partner at Mighty Middle Ventures, ran them for eleven before me. We supported 300+ teams who went on to raise $2.7B and produce 44 exits (a 15% outcome rate). I have watched multiple Catalyst and Innovention alumni go on through Techstars, Y Combinator, a16z Speedrun, Forum, ERA, and many others, which means I have seen the same companies before and after the brand-name programs and can compare what those programs added.

Empathy demands that we reject mediocre accelerators

I learned this myself. As a young founder I was considering a nationally known healthtech accelerator network, since thankfully shuttered. Before applying, I reached out to alums. Nine in ten told me their time there had been a complete waste, and I did not apply. The information was already in the community. It was sitting in the peer network rather than in the marketing copy. The community knew. The community had no way to say so to the next founder.

Personal runway is finite. Spousal tolerance is finite. The patience of friends-and-family backers is finite. A mediocre accelerator that absorbs three to four months of that scarce resource does damage that does not show up in the program’s own metrics: the founder still finished demo day, still got the certificate, still appears in the alumni count. What is missing from the count is the venture that might have raised, sold, or pivoted in those same months.

Brad Feld’s Startup Communities lays the cornerstone for how regional ecosystems work. Entrepreneurs lead. “Feeders” (universities, government, investors, service providers) support. Feeder-led programs that misallocate founder time and credibility do real damage to the community as a whole. Feld’s “give first” ethos functions as a normative filter; actors who extract more than they contribute should not become central figures. Even if Feld does not say it in those words, bad-actor accelerators need to be identified and named. A community that protects its mediocre programs out of misplaced civility is failing its founders.

My proposed framing is as follows.

If a startup accelerator fails to get at least 50% of its teams attributable follow-on funding or a substantive attributable new contract within 3 months of graduation, that accelerator is likely doing more harm than good by absorbing precious and finite founder runway.

There are 500,000 active angels around the world. If a program couldn’t shape and nurture 50% of it’s teams into a form factor that a handful deem fundable, that is an outright failure.

The kinder posture is the harder one. Empathy for founders requires the community to tell them when a program is mediocre, even when the program is run by well intentioned operators. The alternative, letting those programs absorb founder runway in the name of being supportive, is a failure of empathy dressed as collegiality.

The Hedging Accelerator

One model to screen out, and it’s a newer one gaining ground: the hedging accelerator. It admits a large cohort but only invests in, and takes meaningful equity from, a fraction of it, never more than 30% of the teams. The rest get the standardized programming and the demo day.

The bet is rational for the accelerator: three to four months of close-quarters work gives them a strong read on both your team and your customers, so they back the few they come to believe in. The risk lands on you, and it’s a signaling one. You’re far more likely to sit in the cohort’s “not funded” slice than its “funded” one, and you’ll spend real effort in your next round explaining to investors why the accelerator that watched you for a quarter chose not to write the check.

MD matters more than the brand.

The brand on the accelerator is a wrapper. The implementation is the work of the managing director, and that’s where the variance lives. The cleanest example in the literature: one accelerator ran two cohorts a year apart — one produced strongly positive effects on fundraising and employment, the other reduced participants’ web traffic during the program and showed no detectable effect on fundraising. Same brand, opposite outcomes, twelve months apart. The accelerator effect lives in program design and operator quality, not in the marquee.

What the MD controls and the brand book doesn’t: mentor recruiting, cohort selection, which VCs come to demo day, whether your time gets protected or strip-mined. I’ve mentored at six Techstars affiliates and the founder experience varies enormously across them. The diligence question isn’t “is this a well-regarded brand?” but “who’s the MD, how long have they been here, and can I talk to alums from THEIR last two cohorts?” A famous logo with an MD who arrived nine months ago is a brand-new program.

Geography Matters, but not How You’re Assuming

Location matters far less than founders assume, and the MD point is most of the reason why. For many years I felt the two best healthtech Techstars programs were LA, led by Matt Kozlov, and Fort Worth, where Trey Bowles was managing director and Jordan Warnement ran the program. I volunteered my time with the Fort Worth team and was given their all-star mentor award, so I am not neutral, but the quality of that program had nothing to do with being in Fort Worth rather than a coastal hub.

Fort Worth Techstars has since shuttered. Trey now runs 1845 Ventures, a dedicated healthtech fund and studio in Fort Worth. I would assume without hesitation that the thoughtfulness, drive, and wisdom he brought to Techstars teams is still there. The brand on the door changed; the operator did not.

The geography that does matter is your customers’, and it cuts the opposite way from intuition. Seth Weill, formerly an investor on the venture team at Mount Sinai Health System in New York City, gave my portfolio companies the consistent advice: raise in the major metros, but find your payer and provider customers elsewhere. VPs of utilization at payers and CMIOs at hospitals in major metros are desensitized to local teams pitching them the latest widget. A hospital in Louisville or Tulsa is no less relevant as an anchor customer, and is often a more willing one. Pick a program for its MD and network, not its zip code, and run your commercial motion where the buyers still pick up the phone.

Incidentally, Seth has taken a version of his own advice. As of June 2026, he has left Mount Sinai Ventures to build healthtech ventures at General Catalyst in Ohio. GC has effectively infinite money and dedicated teams in NYC and SF. Despite this, they chose to buy a health system as their innovation petri dish in the Midwest.

But they promised me contracts?

Be especially skeptical of the promised buyer access, because the health-system programs are where it underdelivers most.

Here are two anonymized examples, both from brand-name hospital accelerators and where I have direct knowledge of how things played out.

At one major east coast academic medical center, the accelerator’s venture arm was completely separate from the clinical service lines, and the service-line leaders got annoyed enough at cohort teams running customer discovery against their staff that they capped how senior a person any startup was allowed to talk to. Sell-through to the host system was poor as a result: the program could get you in the building, but not past the org chart.

The second case is subtler and more expensive. A device team went through a brand-name accelerator at a major coastal hospital system with a tool focused on Emergency Department usage. The ED’s medical director loved it, and the CFO loved it enough that the hospital invested in the company off its own balance sheet. Then purchasing ran the numbers: this category of device came as part of a vendor bundle, and dropping the incumbent’s clearly inferior ones would have repriced everything else in that contract. Even with top-to-bottom leadership support, the system bought two cartons, ever. A hospital can love your product, invest in your company, and still be unable to buy it, and the accelerator’s promised buyer access does not survive contact with procurement.

Before you join any program, run two checks.

• Talk to founders from the most recent cohort, not the alumni the program puts in front of you. Last cycle’s teams know what it delivered before a good exit rewrote the story, so ask them what happened when their raise went quiet and who opened a door.

• Treat the managing director and the program manager as two advisors you’d have to pay in equity, and ask whether you would.

A program is its MD and its program manager. Ninety percent of the value you’ll derive from an accelerator is driven by those two people, and this holds even at the YC level: the teams I’ve worked with have had vastly different experiences depending on which YC partner they drew. If those two aren’t worth advisor shares on their own merits, the brand and the demo day won’t change that.

A template for reaching a recent-cohort founder cold:

“Hi [Name],

I’m applying to [program]’s next cohort. Would it be possible for us to connect? I’d love to hear about outcomes and experiences from someone who just went through the program.

Two questions: 1. what did the program actually do for your raise when it got hard, and 2. would you do it again knowing what the equity and time cost? Fifteen minutes by phone whenever’s easy would be hugely helpful. I’m available [option 1], [option 2], or [my calendly].

Please feel free to share your calendar and and I can grab time if that would be more convenient.

Thanks!

-New Founder”

Be honest, too, about why you’re joining. If the answer is to learn how to sell B2B enterprise into healthcare, an accelerator is an expensive way to do it. Three resources will teach you more in a week of evenings:

• Julie Yoo’s go-to-market playbooks for digital health startups at a16z Bio + Health, the cleanest single resource on commercial design for digital health [4].

• Koby Conrad’s “6 Growth Channels That Actually Work”, by Rupa Health’s then-head of growth, who scaled the platform from $5M to $75M in annualized lab sales before founding his own company [5].

• A 2023 workshop I ran on selling to health plans, with Jacob Victory (former Centene executive) and Bobby Murphy (a sales lead with exclusive Medicaid-plan experience) on the panel, and a fireside chat moderated by Rohan Siddhanti, then GTM lead at the EHR Healthie and now in a healthtech GTM role at Anthropic [6].

Or wait for the B2B healthtech GTM field guide I’m publishing in late July. The equity an accelerator costs is worth it for the MD and the network behind the program. It’s rarely worth it as a place to learn enterprise sales.

Even the best accelerators don’t replace a fundraise

The dollars that you get from most programs will not replace the need to run a dedicated fundraise after demo day.

Raising from angels and pre-seed funds is entirely doable even without an accelerator on your cap table. This document is part of a series of field guides that I’ve written on the topic through the lens of having coached MANY successful founders.

(Free Subscribers) How to build a pitch deck and startup narrative
(Paid Subscribers) How to find investors and run an angel or pre-seed process

If you join, three pieces of advice

I ran these programs for years, so let me close with the advice I gave every team I took on. All three come down to the same thing: a program can only help with what you put in front of it.

Be the squeaky wheel. When you ask, a good director and their team will bend over backward to get you the intro, the mentor, the contact, whatever you need. We can’t read minds. If you don’t tell us what you’re stuck on, we can’t unstick it.

Tell us when things are going wrong. Teams break up, commercial champions change roles, roadmaps drift. We can’t help with what you don’t share, and the problems you hide are usually the ones that sink the round.

Do the work. Showing up for the PMF seminars and demo day and nothing else is a waste of everyone’s time. The best outcomes among my teams came, every time, from the founders who actually showed up and did the work between the sessions.

References

1. Valentina Assenova & Raphael Amit, “Poised for Growth: Exploring the Relationship Between Accelerator Program Design and Startup Performance,” Strategic Management Journal 45(6), 1029-1060, 2024. 8,580 startups across 408 accelerators in 176 countries (GALI data, 2013-2019). Accelerator participation roughly doubles the probability of raising venture capital (from ~3% baseline to ~6.4%) and adds about $1.8M to first-year funding. Peer-reviewed. Comparator is other applicants, not a clean experiment; the positive headline average is built on a highly unequal distribution dominated by a small cluster of strong programs. https://doi.org/10.1002/smj.3581

2. Sandy Yu, “How Do Accelerators Impact the Performance of High-Technology Ventures?” Management Science 66(2), 530-552, 2020. Matched-pairs design comparing ~900 accelerator graduates against ~900 non-participants. Three years after graduation, graduates raised about $9M less in total funding, were more often acquired in talent-driven acqui-hires than technology acquisitions, and shut down earlier and more often. Yu’s framing: accelerators “resolve uncertainty around company quality sooner, allowing founders to make funding and exit decisions accordingly.” Peer-reviewed. The first-year bump dissolves over the 3-year horizon; compressed time-to-clarity, not compounding capital, is the real product. https://doi.org/10.1287/mnsc.2018.3256

3. Benjamin L. Hallen, Susan L. Cohen & Christopher B. Bingham, “Do Accelerators Work? If So, How?” Organization Science 31(2), 378-414, 2020. Accepted-versus-almost-accepted design at a set of top US accelerators, controlling for startup quality. Some but not all programs aided venture development: top-tier produced large gains, second-tier delivered beneficial learning without the sorting/signaling premium, and one of the studied programs actively inhibited venture development along some dimensions. Mechanism credited: “broad, intensive, and paced consultation.” Peer-reviewed; accepted-vs-almost-accepted quasi-experimental design. https://doi.org/10.1287/orsc.2019.1304

4. Julie Yoo (a16z Bio + Health), “The New Go-to-Market Playbooks for Digital Health Startups.” The cleanest single resource on commercial design for digital health. Named-investor essay; fetched in full. https://a16z.com/the-new-go-to-market-playbooks-for-digital-health-startups/

5. Koby Conrad, “6 Growth Channels That Actually Work” (video). Conrad scaled Rupa Health from $5M to $75M in annualized lab sales as its head of growth before founding his own company; a masterclass in multi-channel healthtech B2B sales. Disclosure: I’ve sent this to many founders and invested in Conrad’s friends-and-family round.

1. Vadim Gordin, “Selling to Health Plans,” 2023 workshop (video). Panel with Jacob Victory (former Centene executive) and Bobby Murphy (a sales lead with exclusive Medicaid-plan experience), and a fireside chat moderated by Rohan Siddhanti (then GTM lead at the EHR Healthie, now in a healthtech GTM role at Anthropic). First-person workshop.

A mini field guide for founders building SMS-led healthcare programs who keep being told “we’ll figure out compliance later” and have not yet seen the bill.

5 Minute Read • June 8, 2026

Author’s note

Hi, I’m Vadim.

Healthtech venture is hard, a fact I’ve learned firsthand as an operator, accelerator director, and investor. I’ve watched good teams repeat avoidable mistakes others have already paid for. I write Healing Healthtech to distill research and the experiences of top-tier operators into actionable tactics and frameworks, so we can all aim to make only new mistakes.

Subscribe for a twice-monthly dispatch of free and paid field guides meant to help you build more deliberately toward something better in human health. And if it’s useful, please forward it to other healthtech builders in your network.

This is the open, abbreviated edition for free subscribers of Healing Healthtech. The full SMS vs. Apps Field Guide is available to paid subscribers. It combines deep operator experience from payer, provider, and startup CEOs, with findings of 75+ published studies into a tactical document on how, when, and why to choose SMS vs. Apps for Digital Health programs as well as how to instrument and manage their operation.

Subscribe now

TLDR

• Bucket every message by content type before the first send, and never let two buckets ride on the same wire. The mixed-content SMS is the most common TCPA enforcement vector in healthcare; $500 per message under a private right of action.

• Sign a BAA with the SMS vendor, then walk the chain down to the carrier aggregator and back up in writing. HIPAA coverage is only as strong as the weakest BAA in the path; downstream gaps surface during enforcement, not before.

• Send notifications in the SMS body; route identified clinical content through a click-out to a HIPAA-compliant surface. The SMS layer’s HIPAA posture is the click-out it lands on, not the BAA on the vendor.

• Treat consent that came with your partner’s list as covering the partner’s sends, not yours. Without your own express written consent or contractual business-associate coverage, the first message is the violation.

• Register every brand and use case with The Campaign Registry, plan capacity against the tier-3 ceiling, and build the opt-out flow to CTIA spec. The carrier layer filters before TCPA enforcement gets the chance.

The four regimes

Four regimes govern healthcare SMS, and they enforce independently on different timelines. An audit, a class action, a carrier dispute, and a filter incident each demand a different exhibit.

• TCPA (Telephone Consumer Protection Act) — federal SMS consent law. Requires prior express consent for automated SMS to mobile numbers. Enforced by class-action plaintiffs (private right of action at $500 per message) and the FCC. The FCC’s 2015 declaratory ruling carves out treatment-purpose messages from HIPAA-covered entities; the 2024 Consent Order codified cross-channel revocation.

• HIPAA — PHI handling. Governs what protected health information can flow through which channel on which consent. Enforced by HHS OCR and state AGs in parallel through HITECH. Civil penalties run from $137 to roughly $2.1M per violation. SMS is permitted under specific architectural patterns.

• 10DLC (10-digit long code) — carrier registration for application-to-person SMS. Verizon, T-Mobile, and AT&T require every brand and use case to register with The Campaign Registry before traffic flows. Unregistered traffic is filtered, throttled, or blocked. Per-violation carrier penalties reach $10,000.

• CTIA Messaging Principles — industry self-regulatory content standard. Published by the wireless trade association. Not law. Enforced by the carriers through 10DLC filtering and campaign suspension. The cost of noncompliance is filtering and lost reach.

Six tripwires, and how to avoid them

1. The mixed-content SMS converts a treatment-exempt message into marketing.

Appointment reminders, lab-result notifications, post-discharge follow-up, prescription nudges, and pre-op instructions all ride the FCC carve-out without prior express written consent. The carve-out evaporates the moment one send mixes a marketing hook in. “Your appointment is Tuesday at 2pm. Ask about our new weight management program!” reads as a reminder to the founder writing the template and as a marketing message to the FCC and plaintiff’s counsel. At $500 per message, one careless template underwrites a class action.

The fix is architectural, not editorial. Treatment-purpose, marketing, and billing each travel a separate code path with a separate consent gate. The content classifier sits upstream of the send, not downstream of legal review. If your engineer cannot point to the three flows on the system diagram, you have one flow and three liabilities.

2. The BAA chain breaks where you aren’t looking.

HIPAA permits PHI in SMS when the covered entity has a BAA with the SMS vendor and that vendor has equivalent BAAs with everyone the message touches: the carrier aggregator, the message routing service, the analytics tooling, the observability stack. The chain is only as strong as its weakest link, and the weakest link is rarely the one your counsel reviewed. Get the executed BAA before the first production message, then ask the vendor for written confirmation of downstream coverage and read the actual list of subcontractors. Vendors that have done this work hand you the list; vendors that have not give you a paragraph that begins “we maintain industry-standard.” Treat that paragraph as the answer on whether they are ready for production.

3. PHI belongs behind a click-out, not in the SMS body.

SMS is not a HIPAA-noncompliant channel. It is a channel with a specific architectural pattern: notification in the body, identified clinical content behind a click-out to a HIPAA-compliant surface. Lab values, full clinical narratives, sensitive diagnoses, medication changes, mental health, and sexual health content sit behind the click-out. The SMS announces that something is ready; the secure surface is where the patient sees it. The click-out has to authenticate the patient before displaying PHI and log access for the audit trail, or the BAA on the vendor is doing no work. The patient-initiated exception is the other path: when the patient texts in first, the practice can reply on the same channel under the 2013 Omnibus Rule preamble, after a one-time unencrypted-channel disclosure documented in the chart.

4. The patient list from your partner does not bring their consent with it.

When the startup receives a list from a payer or provider, the consent that runs with it is consent the patient gave to the partner, not to the startup. Two questions before first send. Does the partner’s notice of privacy practices cover communications by their business associates, and is the startup contracted as one? And is the content treatment-purpose? Yes to both means the inheritance plus the FCC carve-out cover the send. No to either means the startup needs its own express written consent before the first message. The 2024 TCPA Consent Order adds one more wire to trip on: cross-channel revocation, with the compliance deadline currently extended to January 31, 2027. An opt-out on any channel from any covered brand revokes consent across all channels from that brand. Build the handler to revoke globally by default; do not let the engineer ship the channel-local version because the global one is harder.

5. The carrier layer filters you before TCPA enforcement gets the chance.

Register the brand and every use case with The Campaign Registry before first send. Healthcare programs register Account Notification, Appointment Reminders, and Public Service Announcements as separate use cases. The throughput tier (1–3) caps per-second send rates; tier 3 is required for population scale and is gated by your brand trust score. Plan capacity against the tier-3 ceiling at launch; retrofitting after a campaign hits the tier-1 cap is a delay measured in weeks while clinical operations sit on their hands.

Build the opt-out and acknowledgment flow to CTIA spec, not the TCPA floor. STOP, HELP, and frequency disclosure at first message. Branded short-domain links, not bit.ly. Consistent sender ID across the use-case lifecycle. Clean content classification tagged to the registered use case. Programs that meet TCPA and miss CTIA get filtered before the recipient sees the message.

6. Retain two attorneys, not one.

A TCPA specialist for the federal SMS consent layer, the carve-out mechanics, the 2024 Consent Order, and the class-action exposure curve. A healthcare-regulatory specialist for HIPAA, state privacy, the corporate-practice-of-medicine questions, and the BAA chain. A generalist who opines across all four regimes from a single seat is the wrong choice; the deliverable is a memo that hedges everything and decides nothing. Come to each prepared, with the architecture document, the message-type buckets, and the BAA-chain status in hand. The attorney is most valuable on the discrete decisions you have already framed, not on building the frame.

Primary sources

FCC 2015 Omnibus Declaratory Ruling (FCC 15-72) • FCC 2024 TCPA Consent Order (FCC 24-24), §64.1200(a)(10); cross-channel revocation deadline Jan 31, 2027 • 45 CFR § 164.522(b) and 2013 HIPAA Omnibus Rule preamble • The Campaign Registry A2P 10DLC framework • CTIA Messaging Principles and Best Practices. Distilled from the SMS vs. Apps Field Guide (May 2026), §3.3 and Appendix A.2.

This article’s observation and prediction is threefold:

1. Both companies are generational category creators. Both will persist as significant standalone businesses for many years.

2. Neither achieves the durable-monopoly economics implied by their current valuations: monopolist pricing power relative to peers, or exclusionary network effects of the kind Google has in search or Meta has in social.

3. As frontier API premiums hollow out, the valuation gap closes through multiple compression countered by expanded revenue lines, not collapse.

OpenAI made a developer-first, platform-vendor bet: its API became the default substrate for the vertical AI gold rush of 2023-2024, and Abridge-class companies built on top. Anthropic made an enterprise-first, direct-to-buyer bet: Claude for Healthcare launched at JPM26 with HIPAA-ready BAA distribution, native CMS/ICD-10/PubMed connectors, and Banner Health’s 55,000-staff “BannerWise” deployment as anchor.² The Anthropic Partner Network composition (Accenture, Deloitte, PwC, KPMG, Salesforce, all system integrators that sell into health systems rather than into vendors) and the safety-first brand positioning point in the same direction. The customer-base split appears deliberate.

Healthcare workloads bundle into a handful of bounded language transformations: ambient documentation, prior authorization, coding, inbox triage, eligibility screening. Distilled models thrive precisely where the problem and solution spaces are discrete. That is healthcare’s shape as LLM capability has lunged into competence at an expanding range of healthcare tasks.

Two distinct mechanisms compress the frontier here, and they should be named separately.

Distillation

The first is distillation: a vendor builds a student model on Llama, fine-tunes it on the workload, and routes inference to its in-house stack. Vertical AI vendors with engineering capacity and unit-economic incentive run this play. Abridge’s 40%, Perplexity’s Sonar built on Llama 3.3 70B,³ and a May 2025 npj Digital Medicine paper showing an 8B distilled Llama beating its 70B teacher (89.3% vs 76.2%) at less than a quarter of the cost on clinical extraction⁴ are all this mechanism. Distillation has a flywheel that has been underappreciated. Every Abridge call to a frontier API on a real clinical encounter generates an input-output pair, and those pairs are the training corpus for the student model that eventually replaces the teacher. Frontier providers are paid for inference today, and that inference produces tomorrow’s labeled dataset for the customer’s in-house replacement. The teacher pays for its own student.

Provider Rotation

The second mechanism is provider rotation: builders stay on third-party APIs but route to whoever is cheapest or best on a given workload. OpenRouter is the visible aggregator. Rotation does not require a vendor to in-source; it only requires a cheaper or comparable alternative to exist on the market, which open weights guarantee. Google has priced Gemini aggressively against that floor; Anthropic and OpenAI have declined. Google and Meta can afford to compete that way because AI is not their core revenue; OpenAI and Anthropic have no such cushion.

At population scale, the OpenRouter aggregator shows the rotation mechanism playing out. OpenRouter does not capture direct API connections or in-sourced model deployments; it captures the slice of builder demand routed through a marketplace, an approximation of efficient market behavior on LLM choice. Within that slice the trajectory is sharp: Anthropic’s developer share fell from a dominant first-place position to roughly 15% in 18 months; OpenAI’s fell to 8%; the Chinese open-weight cohort rose from 1% to over 50% over the same period.⁵

Anthropic, OpenAI, Llama, and Chinese open-weight share of OpenRouter token volume, October 2024 to May 2026.

The frontier premium is earned on reasoning but does not survive cost-adjustment on production work, and the providers still charging frontier-tier prices are the ones absorbing the compression. Meta absorbed the training cost and gave away the weights. Frontier providers still have to price for theirs. The frontier cannot price against open weights. Anthropic is not structurally safe either, even with its different customer base. Its $965bn post-money valuation is built on charging frontier prices for the foreseeable future; the cap structure cannot survive that premium compressing without reconfiguration.⁶

Both leadership teams see this clearly. OpenAI’s revenue mix has shifted decisively toward ChatGPT consumer and business subscriptions over the API; Sam Altman has repeatedly framed OpenAI as a consumer company, and the “ChatGPT for Healthcare” launch is a destination product rather than a healthcare-capable API release. The strategic logic is to monetize the consumer franchise while API margins compress, with frontier-model capability as the moat that keeps the destination worth visiting. Anthropic has built around reasoning-depth differentiation and around the workloads that scale with frontier capability rather than against it: agentic systems, Claude Code, computer use, multi-step clinical reasoning, autonomous workflows the next-quarter distilled model cannot reach. Both bets are strategically coherent reads of where value will sit at the next generation. Neither bet rescues frontier-tier API margin on the workloads that are already getting in-sourced. The precedent is Cisco and Oracle.

Both companies responded coherently to the substrate threat, and neither response prevented multiple compression.

Cisco’s original business was selling proprietary networking hardware (routers and switches) with high-margin lock-in on a proprietary network operating system. When hyperscalers started building their own switches running Linux-based network operating systems on commodity hardware, Cisco moved up-stack into software and services it could still charge full price for: Webex (video conferencing, acquired in 2007), AppDynamics (application performance monitoring, 2017), ThousandEyes (internet observability, 2020), and Splunk (observability and security analytics, 2024 for $28bn, the largest acquisition in Cisco’s history).

Oracle’s original business was selling a proprietary relational database that ran on proprietary Unix and served as the system of record for most large enterprises. When Linux started eating Unix at the OS layer and open-source databases (MySQL, PostgreSQL) eroded the database moat from below, Oracle leaned into platform lock-in: Oracle Unbreakable Linux in 2006 to capture Linux support revenue against Red Hat; the Sun Microsystems acquisition in 2010 ($7.4bn) absorbing Solaris, Java, and MySQL itself; and Oracle Cloud Infrastructure (OCI) launched in 2016 to compete with AWS as the substrate moved to cloud.

Both pivots were strategically coherent. Both worked, in the sense that the companies still exist and grew revenue. Neither stopped the multiple compression. The customer base in-sourced on Linux regardless of what either vendor did to monetize what was left.⁷ The strategic response cannot escape the substrate. The pivots are real; they do not stop the in-sourcing.

What differs between OpenAI and Anthropic is timing and shape. OpenAI faces the same mechanism as Cisco: customer in-sourcing of the substrate, vertical AI vendors instead of hyperscalers, Llama instead of Linux. Every vendor that follows Abridge takes per-token revenue off the platform; the compression has been in motion for 18 months at population scale. Anthropic faces the same mechanism as Oracle: locked-in enterprise revenue at a lower multiple but durable, the multi-decade compound that comes from procurement complexity and switching costs. Health systems and pharma do not in-source; they keep paying at the contract rate for years.⁷ Both OpenAI and Anthropic are growing ARR rapidly through this. Cisco grew through its compression too: the multiple compresses, the top line keeps growing, and they are the same arc.

Anthropic’s confidential S-1 lands at a $965bn post-money valuation against ~$47bn in annualized run-rate revenue, a roughly 20x sales multiple. OpenAI’s most recent secondary marks sit at $852bn against ~$20bn in trailing revenue, closer to 42x.⁶ Cisco peaked at 39x on ~$12bn of trailing revenue in March 2000 before the multiple compressed to roughly 5-6x over the 25 years it took to reclaim the share price; Oracle trades today at roughly 12x on $57bn in revenue, the high end of a 4-to-15x range over the past decade and well above its ~5x historical median.⁷ The arc this essay predicts puts OpenAI on Cisco’s path: entering public markets near Cisco’s peak multiple, growing revenue rapidly, and watching the multiple compress as the vendor base in-sources on the open-weight substrate. Anthropic enters closer to Oracle’s mature band already, which is the easier IPO setup but the harder long-run story. The multiple has less to fall. The customer franchise has to compound durably enough to grow the top line into a public market cap built on frontier pricing assumptions the buyer base will not always honor.

OpenAI has Cisco-shaped margin risk. Anthropic has Oracle-shaped lock-in risk.

The compression has somewhere to go. Open weights are the substrate. In healthcare specifically, Llama leads the cohort. Abridge on Llama 3, Perplexity on Llama 3.3 70B, and the npj paper on Llama 3.1 are the visible entries in a larger pattern. The 2025 healthcare AI literature shows a sharp bifurcation: GPT-4 dominates evaluation papers (65.7% of model mentions across 4,609 studies per a recent Nature Medicine systematic review), but Llama dominates the build-a-new-model layer where production lives. Me-LLaMA, MMedIns-Llama 3, Med42, OpenBioLLM-70B, Llama-3-Meditron, Llama-3-radiation-oncology, Llama-3.2-Vision-glaucoma, and the local Llama-Anonymizer in NEJM AI are all 2024-25 entrants. There is no equivalent list of GPT-4 or Claude derivatives because those models are closed. Zuckerberg and LeCun saw it first. Open weights are the substrate.⁸

The second beneficiary is the vendor survivors. Abridge built its own SLIM distillation method in 2023, before any vendor offered the capability as a product.⁹ It now runs the in-house stack that captures the gross margin that used to flow through to its frontier API provider. Its gross-margin trajectory differs from non-migrators by roughly 50 percentage points: the difference between software multiples and wrapper-company multiples. The Abridge-class companies that have made the migration are the durable winners of the compression. The wrapper companies that built quickly on frontier APIs without distillation infrastructure are the casualties.

For builders: the test is whether you can move your workload off a frontier API in roughly the time it takes the frontier to ship the next generation. Call it 24 months. You need a distillation pipeline, a labeled-data flywheel feeding off your real workload, and an internal model team that owns the migration end to end. If you do not have those pieces, the next contract repricing reaches you before you can route around it.

For investors: the picks-and-shovels play has moved one layer up the stack, to orchestration, evaluation, fine-tuning, and clinical-data curation. The role this layer would play is the Databricks of clinical AI: one platform where a healthtech vendor brings its clinical data, fine-tunes a model, evaluates and deploys it, and monitors it in production, all under healthcare-grade compliance. That company does not yet exist. General-purpose ML infrastructure vendors carry HIPAA compliance and serve some healthcare customers (Fireworks AI, Snorkel, Arize, Galileo, Predibase), but none integrate the healthcare-native clinical data layer and the clinical-workflow specifics into the training, evaluation, and deployment stack. Healthcare-native data platforms (Truveta, Datavant) have the data and the compliance but do not run those loops. Nobody has stitched the two halves together. The customer is not 6,000 hospitals; it is the few hundred vertical AI vendors who need to in-source their model stack to capture margin currently flowing to their frontier API provider. The wedge has to survive a 2027 first-party Anthropic or OpenAI clinical-tuning release.

OpenAI ends like Cisco. Anthropic ends like Oracle. Meta led the open-weight bet that became the substrate both arcs run on. Abridge’s 40% is the leading indicator. The rest of the sector is the lag.

Drafted with support from Claude as a research assistant with Perplexity and ChatGPT as a red-team.

Citations

1. Shiv Rao, interview with Harry Stebbings, The Twenty Minute VC (20VC), episode “The Five Year Desert to Product Market Fit & a $5.3BN Valuation with Shiv Rao @ Abridge,” May 16, 2026,

1. The 40% in-house figure and the open-source distillation framing appear in the model-strategy section (~18:30). The pulled quote (“Own your stack to protect your P&L and UX”) is from Rao’s LinkedIn post on the episode. Abridge funding position: Marina Temkin, “Shiv Rao’s Abridge has become the AI healthcare startup to watch,” Fortune, 2025 — Series E at $5.3bn post-money, the largest single round in the vertical clinical AI category at announcement. Abridge specific Llama deployment: Brad Genereaux, “Dozens of Healthcare Companies Adopt Meta Llama 3 NIM,” NVIDIA Blog, June 2, 2024, https://blogs.nvidia.com/blog/llama-3-nim-healthcare-generative-ai/ — Abridge is named building its physician-patient encounter summarization on the Llama 3 NIM. OpenAI healthcare partner status: OpenAI, “ChatGPT for Healthcare” launch materials (2026), naming Abridge, Ambience, and EliseAI as health tech partners.

2. Anthropic, “Advancing Claude in healthcare and the life sciences,” timed to the J.P. Morgan Healthcare Conference, January 11, 2026. Coverage: Heather Landi, “JPM26: Anthropic launches Claude for Healthcare to turbocharge AI efficiency at health systems, payers,” Fierce Healthcare, January 2026. Banner Health “BannerWise” 55,000-staff enterprise deployment per Banner-Anthropic case-study coverage, Q1 2026. Anthropic Partner Network composition per Anthropic, “Anthropic invests $100 million into the Claude Partner Network,” 2026.

3. Perplexity, “Meet New Sonar,” Perplexity Hub blog, January 2025, https://www.perplexity.ai/hub/blog/meet-new-sonar. Sonar is built on Meta’s Llama 3.3 70B base, internally trained and fine-tuned for search-answering, served via Cerebras inference at ~1,200 tokens/second. Premium tiers retain access to GPT-5, Claude, and Gemini for harder workloads.

4. Elizabeth Geena Woo, Michael C. Burkhart, Emily Alsentzer, et al., “Synthetic data distillation enables the extraction of clinical information at scale,” npj Digital Medicine 8, 267 (May 10, 2025), https://www.nature.com/articles/s41746-025-01681-4. 8B-All model achieved 89.30% accuracy on annotated synthetic trial criteria vs 76.20% for the 70B-Instruct teacher (Table 2). Cost for a 10,000-patient apixaban cohort: $929 (8B) vs $4,066 (70B) on the least expensive cloud provider (Supplementary Table 5). Paper evaluates information extraction, not generative ambient documentation.

5. OpenRouter LLM Rankings, State of AI 2025 report, and Series B announcement (May 2026). Sources: https://openrouter.ai/rankings, https://openrouter.ai/state-of-ai, OpenRouter’s May 2026 funding-round press coverage. Aggregate token throughput: ~5T tokens per week (late 2025), ~25T tokens per week (May 2026, equivalent to ~100T per month), representing roughly a 5x increase over six months per OpenRouter’s May 2026 Series B press materials. Provider shares as of April 2026: Anthropic 15.4% (down from a dominant first-place position in late 2024); OpenAI 8.1% with GPT-5.4 at approximately 0.98T tokens per week and seven other models ranked above it on weekly volume. Chinese open-weight cohort (Xiaomi MiMo-V2-Pro, MiniMax M2.5, Moonshot Kimi K2.5, Alibaba Qwen 3.6 Plus, DeepSeek): 1.2% (October 2024), roughly 10% (March 2025, post-DeepSeek V3), roughly 25% (Q3 2025, post-Kimi K2 / MiniMax), 45% (April 2026), 51.2% (May 2026). The chart anchors the verified data points (Chinese cohort Oct 2024, Mar 2025, Sep 2025, Apr 2026, May 2026; Anthropic and OpenAI Apr 2026); Anthropic, OpenAI, and Llama trajectories at other timepoints are directionally interpolated from the State of AI narrative. OpenRouter is a developer-routing aggregator. It does not include direct API traffic to Anthropic, OpenAI, Google, or Meta endpoints, nor inference on customer-hosted or vendor-distilled models. Share figures reflect builder adoption within the marketplace slice rather than total enterprise inference share; the cost-compression dynamic the marketplace approximates operates in those other slices as well.

6. Anthropic Series G: $30 billion raised at $380 billion post-money, February 12, 2026, per Anthropic, “Anthropic raises $30 billion in Series G funding at $380 billion post-money valuation,” and TechCrunch coverage. Anthropic Series H: $65 billion raised at $965 billion post-money, announced May 28, 2026, per NBC News and Axios coverage; this is the post-money figure cited in body ¶7. Sacra, “Anthropic revenue, valuation & funding,” May 2026, tracks ARR through both rounds. On OpenAI: $122bn raise at $852bn post-money valuation closed March 31, 2026 per OpenAI press release “Accelerating the next phase of AI” and Forge Global secondary-market indications (~$880bn implied as of April 24, 2026); $25bn annualized revenue as of February 2026, ~$24bn run-rate as of March 2026 per Sacra OpenAI tracking page. The multiples paragraph uses OpenAI’s ~$20bn trailing-twelve-month revenue figure (TTM lags annualized run-rate by 4-6 months in a fast-growing revenue line) per TradingKey, “Anthropic Pre-IPO Valuation of 965 Billion Surpasses OpenAI for the First Time,” 2026, https://www.tradingkey.com/analysis/stocks/us-stocks/261935293-anthropic-ipo-openai-claude-code-tradingkey. Anthropic Series H ARR figure of ~$47bn is per CNBC, “Anthropic tops OpenAI as most valuable AI startup, nears $1 trillion valuation in latest round,” May 28, 2026, https://www.cnbc.com/2026/05/28/anthropic-open-ai-startup-value.html, and TradingKey. Anthropic confidential S-1 filing per Jessica Mathews, “Anthropic confidentially files for IPO after raising $65 billion in a funding round at a $965 billion valuation,” Fortune, June 1, 2026, https://fortune.com/2026/06/01/anthropic-confidentially-files-ipo-965-billion-valuation/. Anthropic’s $965bn post-money exceeded OpenAI’s $852bn for the first time in May 2026 per the same CNBC and Fortune coverage. OpenAI revenue mix per Sacra (May 2026): ChatGPT consumer and business subscriptions (Plus, Pro, Team, Enterprise, Business, Edu) account for approximately 65% of total revenue, API access ~25%, and partnerships ~10%; enterprise customers are on track to reach parity with consumer revenue by end of 2026. Sam Altman has explicitly framed OpenAI as a consumer tech company; see Ben Thompson, “An Interview with OpenAI CEO Sam Altman About Building a Consumer Tech Company,” Stratechery, March 20, 2025, https://stratechery.com/2025/an-interview-with-openai-ceo-sam-altman-about-building-a-consumer-tech-company/. See also Alex Kantrowitz, “Sam Altman on OpenAI’s Plan to Win, AI Personalization, Infrastructure Math, and The Inevitable IPO,” Big Technology, 2025, https://www.bigtechnology.com/p/sam-altman-on-openais-plan-to-win.

7. Cisco Systems market capitalisation peaked above $555bn on March 27, 2000, briefly making it the world’s most valuable publicly traded company at ~39x trailing revenue. The 39x figure anchors to Cisco’s FY1999 revenue of $12.2bn (the most recently reported fiscal year as of the March 27, 2000 peak); annualized revenue was ramping toward FY2000’s $18.9bn at the time, so trailing-quarter multiples differ from headline trailing-year multiples by a meaningful gap. The stock then traded sideways for over two decades through multiple compression; revenue grew from ~$22bn (FY2001) to $56.7bn (FY2025). The share price did not reclaim its March 2000 split-adjusted high until December 10, 2025, by which point market cap stood at ~$317bn, implying a roughly 5-6x trailing P/S. Sources for Cisco: Cisco 8-K filings (revenue); CNBC, “Cisco’s stock closes at record for first time since dot-com peak in 2000,” December 10, 2025, https://www.cnbc.com/2025/12/10/ciscos-stock-closes-at-record-for-first-time-since-dot-com-peak-2000.html; Bilello tweet citing YCharts on 2000 peak P/S ratios; public market data aggregators (Macrotrends, public.com) for market-cap snapshots. Oracle comparator: Oracle FY2025 total revenue $57.4bn per Oracle Corp, “Oracle Announces Fiscal 2025 Fourth Quarter and Fiscal Full Year Financial Results,” June 11, 2025, https://investor.oracle.com/investor-news/news-details/2025/Oracle-Announces-Fiscal-2025-Fourth-Quarter-and-Fiscal-Full-Year-Financial-Results/default.aspx. Market cap ~$717bn as of June 2026 per TradingEconomics, implying ~12.5x trailing P/S. Oracle’s 10-year P/S range is 3.78 to 15.06 with a median of ~5.27 per GuruFocus and Macrotrends historical data; the current ~12x sits near the high end of that band, reflecting re-rating on cloud and AI infrastructure narratives, while the long-run median ~5x is the more relevant anchor for a mature-state endpoint. Cisco’s strategic response to Linux/white-box in-sourcing, by acquisition: Webex ($3.2bn, 2007), AppDynamics ($3.7bn, 2017), ThousandEyes (~$1bn, 2020), Splunk ($28bn, closed March 18, 2024 — Cisco’s largest deal in four decades, per the Cisco-Splunk close press release at https://www.splunk.com/en_us/newsroom/press-releases/2024/cisco-completes-acquisition-of-splunk.html). Oracle’s strategic response: Oracle Unbreakable Linux launched October 2006 as a Red Hat-compatible distribution with $99/system support pricing per Oracle’s October 2006 OpenWorld announcement (see Oracle Linux history, https://en.wikipedia.org/wiki/Oracle_Linux); Sun Microsystems acquired for $7.4bn, closed January 2010 (delivering Solaris, Java, and MySQL); OCI launched 2016 and now constitutes a material and growing portion of Oracle’s revenue mix per Oracle 8-K filings. Both companies’ pivots were strategically coherent but the multiple compressed regardless, which is the precedent the essay’s body paragraph anchors to.

8. 2025 healthcare AI literature bibliometric scan, May 2026. Macro evidence: Nature Medicine LLM-assisted systematic review of 4,609 clinical-medicine LLM studies through September 2025 (OpenAI/ChatGPT 65.7% of evaluated models; Gemini/Bard 13.1%; Llama family ~7%), https://www.nature.com/articles/s41591-026-04229-5. Llama-derivative medical foundation models named: Me-LLaMA (Xie et al., npj Digital Medicine 2025, Llama 2 base), MMedIns-Llama 3 (Wu et al., npj Digital Medicine 2025), Med42 (Llama 3 base), OpenBioLLM-70B (Llama 3 base; 86.06% average across nine biomedical benchmarks), Llama-3-Meditron, Llama-3-radiation-oncology (peer-reviewed 2025), Llama-3.1-biomedical-QA (medRxiv 2025), Llama-3.2-Vision-glaucoma (peer-reviewed 2025), LLM-Anonymizer (NEJM AI 2025).

9. Mrigank Raman, Pranav Mani, Davis Liang, Zachary Lipton, “For Distillation, Tokens Are Not All You Need,” NeurIPS 2023 Workshop on Instruction Tuning and Instruction Following, OpenReview, November 2023, https://openreview.net/forum?id=2fc5GOPYip. Three of four authors (Mani, Liang, Lipton) are Abridge-affiliated. Introduces SLIM (Sparse Logit Infused Modeling): distillation using the top-5% highest teacher logits with dynamic weighting of KL divergence and cross-entropy loss.

2.1 Five clinical truths that persist in the research literature
2.2 Where Apps vs. SMS wins
2.3 LLMs in the care delivery stack

2.1. Five Clinical Truths

Five claims hold across the strongest studies. Most operator decisions follow from them; trial detail lives in the appendix.

First: SMS improves medication adherence at chronic-disease scale.

Thakkar 2016 [16] pooled 16 RCTs (n=2,742) across five chronic conditions: SMS roughly doubled adherence odds (OR 2.11, 95% CI 1.52-2.93). The effect generalizes; design choices (personalization, frequency, bidirectionality, content) didn’t separate in subgroup analysis. Configuration is the operator’s call.

Second: program adherence predicts outcomes; app engagement does not; small SMS-driven adherence gains in short trials do not predict endpoint changes.

Read more

3.1. How to instrument and measure an effective program.

Builder TLDR: Instrumentation answers two questions: did the prescribed action happen, and does what you measure tie back to “effective.” Bottleneck pattern names the action layer; engagement metrics answer neither.

Effective as used in our field guides means something very specific. A program that survives its unit economics, hits criteria that enterprise healthcare buyers purchase and renew on, and most importantly, reduces human suffering.

A note on scale. What follows describes instrumentation for a program past first design. A team running its first 150 patients only needs to know which layer the mechanism sits in and instrument that layer; the full panel comes with scale.

Lens 1: Did the prescribed action happen?

The first instrumentation question is whether the prescribed action actually happens. Three bottleneck patterns recur, each pointing at a different action layer.

Titration-bottlenecked conditions (hypertension, polypharmacy, oncology dose adjustment, anticoagulation) make Layer 4 (clinical process and escalation) the operator’s primary leading indicator.

Track titration rate, escalations closed, time-to-human-review, false-positive burden. Bressman BP Pal demonstrates the pattern at the mechanism layer: program-driven data flow produced higher titration in the SMS arm, and titration is what moves the outcome the buyer pays on. (The trial was null on its primary six-month SBP endpoint with high attrition; the titration-rate signal is exploratory and operator-relevant, detailed in Section 2.1, Truth 4.)

Behavior-bottlenecked conditions (weight loss, glycemic management via diet and exercise, smoking cessation, physical activity) make Layer 3 (program adherence) the primary leading indicator. Track compliance ≥80% or the condition-specific equivalent: smoke-free days, food-log compliance rate, step counts above threshold. A patient who logs meals and consequently eats less is moving the outcome directly.

Reach-bottlenecked programs (Medicaid outreach, post-discharge, public-health campaigns where the binding constraint is delivery rather than action) make Layer 1 (reachability) the primary leading indicator. Track currently reachable, dormant-but-reachable, hard unreachable, re-engagement attempt success.

Figure 10. Engagement vs. program adherence as leading indicators for clinical outcomes.

Engagement metrics don’t answer the action-completion question.

Read more

1.1. The False Binary

Figure 1. The reach-vs-response job-allocation frame.

“Should we build an app or send texts?” frames SMS and app as competing technologies. They are channels for different jobs. A chronic care management program runs many tasks: enrollment, reminders, vital capture, symptom escalation, education, trend tracking. Each has an optimal channel. Sending all of them through a single channel leaves performance on the table. The default single channel is the app, because apps are the venture-backed default.

The operative distinction is reach vs. response. SMS is the channel when reach is the goal; app is the channel when response is the goal. Reach jobs (medication prompts, post-discharge check-ins, appointment confirmations) succeed when the message arrives and the patient acts outside the channel. Response jobs (logging vitals, multi-drug regimen entry, behavior-change modules) succeed only when the patient produces structured input inside the channel. Most chronic care programs contain both kinds of job.

The productive question is which jobs are reach jobs and which are response jobs, not which channel wins.
We’ll go into detail on this distinction in section 2.2, but it’s important to establish here upfront.

1.2. The Operating Stack

Every digital chronic care program runs the same causal chain from message delivered to economic outcome. Each layer is its own measurement object, its own design decision, and its own failure mode.

Figure 2. The Operating Stack — the six-layer causal chain every digital chronic care program runs.

Read more

This document is meant to be an actionable artifact for builders.

Primary (numbered) references are cited inline in the body; Supplemental references in each specialty section are additional reading.

These are organized by specialty rather than alphabetical order or the sequence in which they appear in the text.

Many of the references below are open-access and freely available. Others sit behind academic subscription paywalls. For paywalled papers, Google Scholar is the best place to start; it often surfaces a PDF or preprint via an institutional repository or author’s homepage. PubMed Central (PMC) holds full text for NIH-funded work after the 12-month embargo. When all that fails, sci-hub.ru remains a working option for readers who choose to use it.

Read more

• Allocate by reach vs. response.

  • SMS for reach jobs (action outside the channel);

  • Apps for response jobs (structured input inside it).

  • Most chronic care programs need both.

• The platform is never the intervention.

  • Platforms exist to affect what humans do beyond them: patients taking pills, clinicians titrating, care teams closing the loop.

  • A delivery vector, not the mechanism.

• Audience decay is the SMS-vs-App economic divider.

  • App audiences decay continuously through uninstalls, OS push restrictions, and device churn; the reach floor surfaces only when re-engagement fails.

  • SMS preserves the addressable list across years through phone-number persistence.

  • Healthcare CAC is paid upfront, so audience decay erodes the reach floor before outcomes are measured;

  • Medicaid weakens this but doesn’t invalidate it.

• Program adherence (actual patient behavior) is the leading indicator of clinical outcomes; engagement should be limited to a product-health metric.

  • Track completion at 80% (or the condition-specific equivalent).

  • Engagement stays on the product dashboard.

  • The patient heals and buyer renews on health and economic outcomes.

    • Those are what should drive your operational decisions and dashboards.

• Combined modalities beat single-channel automation.

  • Across the trial record, pairing a delivery channel with a staffed action layer beats single-channel automation.

  • Program architecture beats channel.

• Stand up regulatory compliance before launch.

  • Register every brand and use case with The Campaign Registry; build to the FCC treatment-purpose exemption and the CTIA floor.

  • Retrofitting TCPA, HIPAA, or 10DLC after launch is the most expensive architectural choice an SMS program can make.

Author’s note

For the past six years, I’ve mentored healthcare founders through NYU accelerators and public workshops that I host in New York City. The thesis behind that work has been consistent:

Only new mistakes.

VC-backed healthtech has a pattern of new builders impaling themselves upon old failure modes. Research findings stay trapped in journals. Operational lessons stay trapped inside health plans and care teams who’ve done this work for decades.

These Field Guides exist to help close that gap.

The goal is effective digital health programs: interventions that survive their unit economics, improve the outcomes enterprise buyers renew contracts on, and most importantly reduce patient suffering.

Part of my unfair advantage as a mentor is that I have operators in my network who have built and delivered care at population scale. They’ve already learned many of the expensive lessons and have generously provided their feedback and comments on this piece.

• Evan Huang, developed SMS programs reaching 2% of the U.S. chronic care population, First as co-founder of CareSignal and then as CTO of Lightbeam after the acquisition.

• Ramon Lizardo, MD, MBA, built the nation’s largest ED diversion company, covering 9.5M lives across 150 health plans as CEO of Tele911.

• Brittany Sigler, MPH, DrPH, was a product leader across multiple population-scale outreach programs at CVS Caremark and Aetna.

This field guide distills lessons from 30 years of clinical, operational, regulatory, and behavioral literature surrounding healthcare communication systems. It is not clinical, legal, or compliance advice.

This document was developed using Claude Cowork + custom skills as my primary collaborator, with Gemini, ChatGPT, and Perplexity providing redteam support. My use of AI is loud and proud. The arguments and conclusions are mine.

Vadim Gordin.
Only new mistakes.

© 2026 Vadim Gordin. This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/

Contents

The piece is organized in three parts. Each can be read on its own.

Part 1. The System.

• The conceptual frame: why the channel choice is the wrong unit of analysis,

• the six-layer causal chain from message-delivered to therapeutic and economic outcomes that every digital chronic care program runs,

• the economics of reachability between SMS and Apps, and

• how scaffolding can dramatically improve retention and engagement rates.

• Action beyond the platform is the actual intervention.

Part 2. What the Evidence Supports.

• The five clinical truths the literature supports,

• where apps and SMS each win on a per-task basis,

• what is the cost and consequence of platform failures, and

• how LLMs fit inside the operational stack.

Part 3. Operating the System.

• How to measure and instrument an effective program and

• common failure modes in both commercial and research programs.

• Will me compliance architecture survive scrutiny or create liability?

Appendix

• The appendix is intended to be a reference tool rather than something to read end-to-end

• All 75+ cited clinical trials by topic and specialty.

  • If you’re building a renal care or ob/gyn platform for Medicaid and want to read up on how previous SMS interventions have performed, the appendix is your on-ramp.

• The 3 major regulatory regimes governing SMS in the US organized into an actionable reference for builders.

Read more

For 5 years, I’ve run a mailing list of NYC digital health events and organized curated workshops on topics that regularly impale founders and builders in digital health. My consistent theme has been

Only new mistakes.

Healthcare is hard enough as it is. If a founder in our community repeats a mistake that I, one of our more experienced operators, or the research literature could have predicted that’s a failure on my part as a mentor and convener. The various workshops and round-tables I’ve hosted over the years have functioned to bridge that gap.

The workshops have been fantastic, but had two failings that were intrinsic to their form factor.

1. The workshops were ephemeral.

   1. Aside from a slide deck, there was never an actionable artifact that you could take from “Selling to Health Plans” workshop on a Thursday night on Union Square into a GTM strategy session on Monday morning.

   2. By the same token, there was nothing substantive that you could forward to your friend who could have gotten value but was home changing diapers.

2. The workshops were geographically constrained.

   1. I firmly believe that current NYC is the best place in history to build healthcare innovation.

   2. At the same time, I have persistently gotten messages from builders around the world asking for recordings or captures of our conversations.

   3. These are folks who I genuinely want to support, but zoom conferences and recording our panels would have broken what made the events special.

The answer to both of those tensions is a new subscription series that I will be publishing starting June 1st called Healing Healthtech Field Guides. Each Field Guide will focus on a specific operational topic that I’ve seen make or break digital health startups. Each Field Guide will distill decades of research literature, anecdotes from builders and investors who have operated at population scale, regulatory framings, and clinical guidelines into actionable workflows, references, and best practices you can start implementing on Monday morning.

Over the next two weeks until the first Field Guide goes live, I will be sharing content and takeaways that I believe are deeply impactful to builders and operators.

My ask of you until then is that when you see these posts on LinkedIn and X, please engage with them.

Like, repost, comment, agree, disagree.

Tell me when my opinions conflict with your lived experience as an operator, provider, patient or caregiver. Insult my taste in suede vintage-inspired sneakers. It’s all good. The goal is to have a meaningful and active conversation.

The NYC Events Roundup will continue after NY Tech Week and remain free. You’ll have the ability to choose whether you get the Field Guides, the Events Roundup, or neither here on Substack. In terms of events, I’m pivoting to 20-30 roundtable roundtables and dinners. The 100+ person events at midtown law firms were both exhilarating and exhausting to run. I will do a few of those a year, but fortunately, we now have a thriving community of new NYC healthtech community organizers who have taken up that mantle and whose events I have been very happy to promote since day 1.

Lastly, please consider becoming a paid subscriber.

It’s been a privilege to support your work for all these years and I hope to continue to do so for a long time to come.

Warmly,

Vadim
Only new mistakes.