We didn't buy the Doximity AI rally.
We watched it the way I watched the Terra peg begin to crack in May 2022. The stock gapped up on an earnings print, the financial press reached for the phrase 'AI-driven physician productivity,' and the retail order flow did what retail order flow always does: it translated a narrative into a price. The narrative was familiar. A profitable healthcare software company that has been around for more than a decade suddenly becomes an AI company. The price action looked like a token launch, not an earnings re-rating. Volume appeared. Momentum chasers arrived. The company announced AI tools, and the stock moved as if those tools had already been deployed at scale.
I did not buy the rally. I also did not short it. A battle trader does not need to take both sides of every emotional move. What I needed was the underlying order flow. What I saw was not a product. What I saw was a liquidity event: the market monetizing a trust network under a new label. That label is AI. The network is Doximity. The doctors are the liquidity. And the pharma companies are the taker fees.
This article is not a price target. It is a structural audit. I want to explain why the Doximity medical AI hype is not a software revolution, why it is a toll booth operator adding faster turnstiles, and why the same fragmented-liquidity disease that infected Layer2s in crypto is now showing up in the medical AI landscape. By the time you finish this, you will understand why I treat Doximity's AI announcement as infrastructure, not innovation.
We didn't read the headline and call that research. The research starts with the network.
Context: What Doximity Actually Is
Doximity is a network for physicians. It is often called LinkedIn for doctors, but that comparison undersells the architecture. Doximity is a closed professional identity layer, a communication system, a medical news feed, a fax network, a telehealth switchboard, a marketing platform, and a workflow aggregator. It has more than two million verified medical professional members and claims coverage of the overwhelming majority of U.S. physicians across every specialty and all 50 states. Physicians use it to read medical news, collaborate on cases, message colleagues, review their own public profile, complete continuing medical education credits, communicate via fax API, and participate in telehealth visits.
The commercial engine is not the doctor. The commercial engine is the relationship between the doctor and the institutions that want something from the doctor. Pharmaceutical companies want attention, trust, and prescribing behavior. Hospitals and health systems want recruitment, referrals, and patient volume. Doximity sits between these parties. On one side of the balance sheet is a verification graph. On the other side is a marketing toll booth.
Doximity's revenue is dominated by pharmaceutical marketing solutions. Pharma companies spend significant sums to reach physicians with product information, clinical trial recruitment materials, sponsored medical education, and branded content. Doximity sells that access with a layer of professional credibility that a social media platform cannot reproduce. If you are a pharma company, you do not want your message placed next to random crypto spam. You want your message inside a clinical news feed used by a verified oncologist with a National Provider Identifier number. Doximity provides that environment. In crypto terms, Doximity is the settlement layer for a very specific form of attention: physician attention at the moment of clinical workflow.
This distinction matters because the AI narrative is being sold as if Doximity is a software company that suddenly discovered machine learning. The more accurate framing is that Doximity is a closed network that is adding an execution layer to its existing settlement layer. AI is not changing what Doximity sells. AI is changing how it produces the thing it sells.
Doximity has launched multiple AI features. There is Doximity GPT, a HIPAA-compliant generative AI tool for drafting clinical communication. There is ambient AI documentation that listens to a physician-patient conversation and produces a note. There are AI tools that parse faxes, automate prior authorization, and structure unstructured medical documents. These are not clinical diagnostic models. They are workflow automation products. They reduce administrative friction. They make the physician faster at doing things inside the Doximity environment. And that is exactly why they are valuable, and exactly why the market should not confuse them with a fundamental change in Doximity's competitive position.
The real product has always been the graph. The graph is the verified physician identity graph. The AI is just another query interface into that graph. The moment you understand that, the AI hype becomes a lot less magical.
Core: Reading the Order Flow
When I audit a crypto protocol, I do not start with the token price. I start with the state management. Where does value enter the system? Where does it exit? Who can create state transitions? Who pays gas? Who is the sequencer? Who can bribe the sequencer? You cannot understand Doximity's AI story until you apply the same framework to physician attention.
Where does value enter the system? Pharma marketing budgets enter. Health system hiring budgets enter. Telehealth billing enters. Where does value exit? It exits as physician time. The physician has a limited number of working hours. Every minute spent in Doximity is a minute not spent in the EMR, not spent with the patient, not spent reading a journal, not spent on a competing pharmaceutical marketing channel. Doximity monetizes that time by being the platform that connects the physician to the institution before the physician reaches a static alternative.
Who can create state transitions? Doximity can. Doximity decides what appears in the physician's news feed. Doximity decides which AI summaries surface, which studies are promoted, which continuing medical education courses are recommended, which clinical trial recruitment messages are displayed. The physician creates the final state transition by acting on that information: prescribing a drug, referring a patient, signing a contract, completing a form. But Doximity shapes the probability distribution of that final transition.
Who pays gas? The pharma company pays. In the same way a trader pays gas fees to move value on-chain, a pharma company pays Doximity to move its message into a physician's attention. AI reduces the cost of producing that message. It does not eliminate the toll.
Who is the sequencer? Doximity is the sequencer because it orders information for physicians. This is the most dangerous power in healthcare media. A centralized sequencer can front-run attention. It can show a sponsored AI-generated answer before an unsponsored clinical answer. It can choose which trial recruitment message appears first. It can prioritize messages from the highest bidder. In crypto, we would call this maximal extractable value. In medical AI, it is called marketing optimization.
This is the insight that the retail narrative misses. The market is not buying an AI product. The market is buying a more efficient extractor of physician attention. Doximity's AI features do not need to be the best clinical AI. They only need to be good enough to keep the physician inside the Doximity environment while Doximity captures a larger share of the physician's limited attention. The AI is not the alpha. The attention is the alpha.
The Verified Physician Graph Is a Closed Mainnet
The key asset is the verified graph. Every physician on Doximity is credentialed. Doximity uses public data sources, NPI records, professional directories, and a verification process to ensure that a given profile belongs to a real physician. This is a major moat. In a world of AI-generated fake identities, verified human professional identity becomes more valuable, not less. In crypto terms, Doximity is a proof-of-humanity registry, but it is not permissionless. It is a closed mainnet. You cannot join without proving that you are a licensed clinician. You cannot earn verification by staking tokens. You cannot fork the graph. The graph is the collateral.
This is also where the bull case gets tricky. A closed mainnet is valuable only as long as the sequencer remains credible. The moment physicians believe that the platform is simply an advertising network, they lose trust. The moment they lose trust, they stop logging in. Doximity's AI features create a conflict between credibility and monetization. If the AI summary is clearly sponsored, the physician will ignore it. If it is subtly sponsored, the physician may be misled. Both outcomes are bad. One kills engagement. The other kills the brand and invites regulatory action. The design of the AI interface is not a technical detail. It is a risk management problem.
During my 2020 DeFi audits, I learned that the most dangerous vulnerabilities are not always reentrancy attacks. The most dangerous vulnerabilities are incentive mismatches. A yield aggregator can be functionally correct and still drain users if the owner can switch the reward asset at the worst moment. Doximity's AI is the same. The smart contract here is the recommendation engine. The reward asset is physician attention. The owner is the pharma sponsor. If the recommendation engine is programmed to maximize pharmaceutical marketing revenue, physician trust is the exit liquidity.
AI Is the Gas Fee, Not the Block
Think about the AI products again. Doximity GPT drafts referral letters. Ambient AI listens and writes notes. Prior authorization automation moves documents between providers and insurers. Fax parsing extracts data from one of the oldest technologies in medicine. Every one of these products is a gas optimization. It reduces the amount of physician effort required to complete a workflow. It does not create a new category. It does not replace the physician. It makes the existing workflow cheaper and faster.
That is valuable. In crypto, gas optimizations matter. A rollup that cuts transaction fees by 80 percent is valuable. But a rollup is not a new chain. It is an extension of the same chain. Doximity's AI products are extensions of the same physician graph. They deepen the integration between the physician and the platform. They increase switching costs. Once a physician trains an ambient AI tool on their voice, their documentation style, their referral patterns, and their prior authorization templates, leaving Doximity becomes painful. The AI is a stickiness engine.
The market, however, is pricing it as a land-and-expand narrative. The market assumes that Doximity can expand from marketing revenue into clinical decision support, population health, precision medicine, and every other AI-adjacent healthcare vertical. That assumption is not supported by the technical architecture. Doximity's AI is not a clinical reasoning engine. It does not have FDA-cleared diagnostic capabilities. It is a workflow tool. Workflow tools produce retention. They do not necessarily produce expansion.
The Layer2 Trap: Too Many Scribes, Same Doctors
Now let me address the structural problem I mentioned at the beginning. In crypto, there is an obsession with Layer2 scaling. Dozens of rollups claim to solve Ethereum's throughput problem. But the user base remains the same. The total liquidity is often fragmented. Bridges add risk. Token standards multiply. Composability becomes a research project. The same user is now using seven networks instead of one. This is not scaling. It is slicing.
Medical AI is doing the same thing to physicians. There are dozens of AI scribe startups. Abridge, Ambience, Nuance DAX, Suki, Nabla, Heidi, and countless others all pursue the same user: a physician who hates charting. They train models on the same clinical documentation patterns. They sell to the same health systems. They deploy into the same EMR environments. The user base is not expanding. Physician time is not expanding. There are only so many hours in a day. The AI scribe market is fragmenting physician attention into more products, not creating new physician capacity.
In this world, the most valuable asset is not the best model. It is the best distribution layer. Doximity is a distribution layer. It already has the physician identity graph. It already has the trust. It already has the relationship with pharma. It can route the output of many AI scribes to the right doctor at the right moment. It does not need to build the best speech recognition model. It needs to be the aggregator that every scribe startup eventually has to use to reach physicians. That is the Layer2 aggregator of the medical AI universe.
But wait. The VC narrative says the problem is liquidity fragmentation. The VC solution is one more product to fix fragmentation. This is the manufactured problem. The real issue is that most AI scribe startups do not have a distribution moat. They raise money, build a demo, sign a few pilot health systems, and burn through cash trying to acquire physician workflows one hospital at a time. Doximity has the distribution. The fragmentation is an opportunity for Doximity, not a threat to Doximity. The market is right to see that. The market is wrong to believe that this makes Doximity an AI-native company. It makes Doximity a toll bridge on an increasingly crowded highway.
An Audit of the AI Claims
Now let me put on the auditor's hat. I spent years auditing smart contracts before I trusted them. I do not trust press releases. I want to see the code. I want to see the model card. I want to see the evaluation set. I want to see the update log. Doximity's AI claims do not include that level of transparency. This is not necessarily a reason to sell, but it is a reason to discount the bullish narrative.
First, what exactly is Doximity GPT? The name implies a proprietary model. It may be an orchestration layer around a third-party model. If so, the moat is not the model. The moat is the API endpoint, the compliance wrapper, the physician graph, and the workflow integration. That moat is real, but it is not the AI moat that the market is pricing. It is a compliance and distribution moat.
Second, where is the evidence that the AI improves clinical workflow outcomes at scale? Doximity has reported growth in physician engagement and AI use, but I have not seen rigorous, externally validated studies on diagnostic accuracy, information retrieval precision, or safety incidents. In crypto, we learned the hard way that unaudited smart contracts are land mines. In healthcare, unverified AI models are worse. A hallucinated referral letter might get caught. A hallucinated dosing recommendation can cause harm. The absence of incident data is not proof of safety. It is proof of insufficient adversarial testing.
Third, the liability structure is unclear. If a physician uses Doximity GPT to draft a note and the note contains an error that leads to a misdiagnosis, who is responsible? The physician? Doximity? The underlying model provider? If Doximity is simply a workflow tool, the physician bears the liability. That will make adoption slower than the stock price suggests. Doctors, unlike retail traders, do not rush into unproven infrastructure.
Fourth, there is the commercial bias problem. Doximity's business model is funded by pharma. AI features that summarize clinical information will inevitably be under pressure to favor sponsors. If Doximity has an AI tool that provides a quick answer to a clinical question, do you think the answer will prominently mention all competing drugs? The economic incentive says no. The regulatory risk says yes. The tension between these two forces will define the next several quarters. In DeFi, we call this oracle risk. In medical AI, it is worse because the stakes are clinical.
The Institutional View: A Stacked Assumption Problem
An institutional allocator looking at Doximity has a different problem. He cannot simply buy the AI narrative because his mandate requires discounting future earnings. At a high multiple, the market is embedding a massive assumption: that AI will not increase rivalry between platforms, that Doximity can keep its take rate, that health systems will not build their own tools, that regulatory costs stay flat. Each assumption individually is plausible. Together, they are not. The market does not price the joint probability; it prices the most attractive scenario.
The same thing happened with early DeFi tokens. Investors bought the narrative that a protocol would capture all value in an ecosystem. Then they discovered that liquidity providers demanded yield, competitors copied the code, and governance votes turned into extractive events. Doximity is not vulnerable to code forks, because the graph is closed. But it is vulnerable to integration forks. If Epic or Microsoft builds a verified physician network with AI-native workflows, Doximity's closed graph becomes less special. The value proposition shifts from 'we have all the doctors' to 'we have a slightly better doctor directory.' That is a terrible place to be in a bull market.
Doctors are not just users. They are liquidity providers. Every minute a physician spends in Doximity is a liquidity contribution. Doximity's job is to aggregate that liquidity and sell it to pharma. The AI is a mechanism for increasing the velocity of that liquidity. But liquidity providers need to be compensated. In DeFi, LPs earn fees. Physicians earn nothing except the work product. Their compensation is time saved. Doximity monetizes that saved time. This creates an eventual tension. If Doximity captures too much of the productivity gain, physicians will feel extractive pressure. They may not leave, but they will not increase engagement. The growth story depends on physicians feeling that AI makes their lives better, not that AI turns their clinical judgment into a sellable signal.
There is also an AI saturation risk. Every physician is already bombarded with AI scribe pitches. The same hospital gets twenty vendor demos per week. The same overworked physician is being asked to adopt five different tools. AI saturation is real. This is exactly the Layer2 fragmentation problem. Doximity can cut through by being the single platform. But if the market overprices that position, a disappointment in user growth will hit the multiple hard.
As a battle trader, I cannot ignore the options market. High implied volatility after an earnings gap suggests that the market is not sure about the AI story. If institutions were confident in the AI thesis, they would sell downside protection. Instead, the options market is charging a premium. This is a signal of uncertainty, not confidence. Retail traders see a trend. I see an uncertainty premium.
The Contrarian: The Short Thesis Is Also a Meme
Now let me take the other side. The mainstream bear case on Doximity is that AI will destroy the company. The logic is simple: if a doctor can open ChatGPT and get instant clinical summaries, why does the doctor need Doximity? If an ambient AI scribe can generate notes without the platform, why does Doximity matter? This bear case misses the most important lesson from the last decade of technology: raw model capability does not beat distribution.
OpenAI has capability. It does not have a physician graph. It does not have two million verified doctors. It does not have the workflows embedded in existing practice patterns. A doctor might use a generic AI chatbot for a quick question, but that doctor will not use it for prior authorization, for secure messaging with a colleague, for CME tracking, for fax interoperability, or for clinical trial recruitment. Doximity owns the workflow stack. The AI scribe startup owns one vertical slice. The generic chatbot owns no workflow. The value of a model is a function of how easily the user can act on the output. Doximity makes action easy. That is the context moat.
The contrarian twist is that the AI narrative is strengthening Doximity's moat, not weakening it. Every new AI scribe produces more data. Every AI scribe wants to reach physicians. Every AI scribe needs distribution. Doximity can become the neutral router between AI models and physicians. In crypto, this is the aggregator play. The same way 1inch routes trades across liquidity pools, Doximity can route physician workflow events across AI vendors. The more AI products enter the market, the more valuable the router becomes. This is the opposite of disruption.
But there is a real bear case, and it is not the one retail investors are discussing. The real bear case is not ChatGPT. It is the EMR incumbents. Epic and Microsoft have deep integration into the live clinical workflow. They see the patient chart. They operate inside the hospital. They can deploy AI ambient documentation directly in the system where the physician already works. If Epic builds a physician network and an AI documentation layer that includes secure messaging and a marketplace for pharma content, Doximity's position as the neutral workflow layer is threatened. The doctor may not need to leave the EMR to get AI assistance. Doximity then becomes an optional add-on instead of an essential front door.
The second real bear case is regulatory. If the FDA or CMS decides that AI-generated clinical communication should be treated as medical decision support, Doximity's compliance burden increases. If the FTC audits pharma's sponsored AI content and finds that AI summaries blur the line between evidence and marketing, the toll booth model is at risk. Regulatory risk is systematically underpriced in the current AI rally, because bull markets do not reward questions. They reward narratives.
The Blockchain Angle: Why This Is Not a Different Animal
You may be asking why a crypto analyst is writing about a healthcare company. The answer is that the same mental model applies. Doximity's AI is not fundamentally different from an on-chain oracle. It provides an answer to a clinical question, but the answer is only as good as the source. In DeFi, we learned to attack oracles. In healthcare, the attacks will come from commercial bias. The AI output is an oracle. The pharma sponsor is the price source. The physician is the protocol. The patient is the end user. Any failure in the oracle propagates to the end user. That is why I treat Doximity's AI as a risk management challenge, not as a growth hack.
Some people will ask whether Doximity should tokenize its physician graph. There are startups proposing decentralized physician networks, patient-owned data marketplaces, and clinical trial recruitment DAOs. The same pattern repeats. They want to take a closed, verified graph and make it permissionless. They forget that permissionless identity is an oxymoron. You cannot prove you are a physician without a trusted issuer. The issuer is either the government, the medical board, the hospital, or a centralized credentialing aggregator like Doximity. Tokenizing the graph does not add credibility. It adds anonymity. Anonymity is not a feature for medical decision-making. It is a bug.
Doximity is a reminder that the market rewards verification, not decentralization. The blockchain industry has spent years trying to solve identity with zero-knowledge proofs. The problem is not the proof. The problem is the root of trust. Doximity has a root of trust. It is not decentralized. It is not open. But it is liquid. It is practical. It is exactly the kind of infrastructure that makes the AI narrative possible. The market is pricing Doximity's AI, but it is really pricing the scarcity of verified clinical identity.
This may disappoint crypto purists, but it should not. The most useful blockchain market structure is not the fully autonomous DAO. It is the hybrid network where a centralized sequencer provides order and decentralized participants provide liquidity. Doximity is a hybrid network. Physicians are the liquidity providers. Pharma companies are the takers. Doximity is the sequencer. AI is the matching engine. No token exists because the stock is the token. The stock's float is the liquidity pool. The earnings call is the governance forum. The SEC is the smart contract auditor.
The same architecture that makes a protocol fragile also applies here. A protocol is fragile when collateral is not observable. Doximity's collateral is physician attention. You cannot see it on-chain. You cannot count it in a smart contract. You have to trust the company's disclosure. In the current AI rally, that trust is being extended far beyond what the evidence supports. This is not fatal. But it is a structural risk.
What My Own P&L History Says About This Setup
I have made this mistake before. In late 2017, I allocated forty thousand dollars into Waves based on technical pedigree. I thought that a blockchain project with strong engineering fundamentals would outperform the hype. The launch was chaotic. Transaction fees spiked. My position dropped thirty percent before the crowdsale closed. I learned that infrastructure strain is the silent killer of new protocols. Doximity is not a new protocol, but its AI rollout faces the same strain. The market is assuming that the AI infrastructure can handle a massive influx of physician activity. But the healthcare industry has old infrastructure. Fax machines. EMRs. Legacy practice management systems. Insecure APIs. The AI layer is being built on top of a system that was not designed for AI. This is a source of hidden operational risk. It is not priced into the stock. It only appears during a crisis.
In 2020, I audited yield aggregators before public adoption and found a reentrancy vulnerability in one. It rewarded me with fifty ETH. The lesson was not that audits make systems safe. The lesson was that adversarial review is the only way to find the real failure points. Doximity's AI cannot be fully audited by outsiders. The evaluation data is not public. The model cards are not public. The adverse event logs are not public. This asymmetry is not a problem in a bull market, but it becomes a problem when the narrative shifts. Every unverifiable claim will be questioned. Every question will create selling pressure. The market is long unverified AI infrastructure.
In 2021, I treated Bored Ape Yacht Club as a liquidity play. I calculated that the floor price premium against secondary volume was unsustainable. I sold fifteen percent of my position at the peak. The correction came. Doximity's AI narrative has a similar liquidity structure. The stock price jumped on AI enthusiasm. The secondary volume of the stock increased. But the fundamental liquidity, the physician attention that generates pharma revenue, did not jump in the same proportion. A price can lead liquidity for a while. Eventually, price follows liquidity. If the physician engagement data does not confirm the AI story, the stock will correct to the old liquidity baseline.
In 2022, I shorted the Terra peg three days before the collapse. The trade was not about predicting the exact day. It was about recognizing that algorithmic stablecoins without sufficient collateral are mathematical time bombs. Doximity is not a time bomb. It has real revenue and real profit. But the market is treating its AI narrative as if the future revenue is current collateral. It is not. The future revenue has to be earned through adoption, regulatory clearance, and physician trust. Until those things are proven, the AI premium is an unverified claim.
In 2025, I built a platform that tokenizes the strategies of verified human traders and lets AI agents execute them. I learned that the hardest part is not building the agent. It is keeping the agent honest when it has access to capital. Doximity is facing the same challenge. The AI agent is the model. The capital is physician trust. If the model is allowed to accept sponsored answers, trust compounds into monetary extraction. The platform needs a governance layer. Doximity's governance layer is its brand and its compliance function. The market is trusting that brand. The question is whether the brand can survive the pressure of quarterly revenue targets.
The OpenSea story is also relevant. OpenSea had the dominant NFT marketplace, but in chasing volume it stopped enforcing creator royalties. That move destroyed the creator economy and alienated the community. Doximity must not make the same mistake. If it monetizes AI by prioritizing sponsored clinical content in a way that violates physician trust, it will be the equivalent of a royalty surrender: short-term revenue, long-term brand death. There is no sustainable business model on-chain for creators who do not control distribution. Doctors are creators in a strange sense: they create clinical decisions. Doximity controls the distribution. It must protect the clinical decision as the product, not the ad impression.
The Takeaway: The Only Structural Levels That Matter
I do not give price targets. I give structural levels. When I analyze a protocol, I identify the assumptions that must remain true. Here are the levels for Doximity.
First, watch physician engagement velocity. The AI story is empty if physicians do not use it. The only metric that matters is the change in weekly active physician engagement and the change in marketing revenue per physician. If AI features are creating stickiness, those numbers will trend up together. If engagement goes up but revenue per physician goes down, Doximity is converting attention into a cost center. If revenue goes up but engagement goes down, the commercial pressure is destroying trust. Both paths end badly.
Second, watch for an AI marketplace. If Doximity opens a public API or an AI marketplace where third-party models can compete for physician workflow tasks, that is a signal that Doximity sees itself as the settlement layer, not the model owner. That would be bullish for the network. If Doximity continues to keep everything proprietary, that is a signal that the company is afraid of commoditization. Fear creates fragile architecture.
Third, watch the regulatory calendar. The AI narrative is being priced as if compliance is solved. It is not. Every major AI scribe announcement is followed by questions about HIPAA, FDA clearance, and state medical board rules. One negative regulatory event could erase the multiple expansion that the market just gave Doximity. This is the same problem as unverified collateral in Terra. The market assumes the collateral is there until the arbitrage stops working.
Fourth, watch what pharma does. If pharma marketing budgets continue to shift toward programmatic targeting of physicians, Doximity is a prime beneficiary. If pharma starts building its own AI-native sales agents that can reach physicians through EMR systems without Doximity, the network premium collapses. The decision will be made by procurement teams, not by technologists. Doximity needs to be deeply embedded in the procurement cycle, not just the engineering cycle.
This is the difference between a protocol and a product. A protocol has a clear value capture point. Doximity's value capture point is the physician's decision. The AI increases the value of that decision because it makes the physician faster. But it also increases the surface area for manipulation. The market is not pricing that manipulation risk. Markets never price existential risk during the expansion phase. They price it during the unwind phase.
We didn't buy the Doximity AI rally. We didn't short it. We watched it with the cold discipline of someone who has seen too many protocols with strong technical demonstrations and weak settlement assumptions. Doximity is a real company. Its physician graph is a real asset. Its AI features are real workflow improvements. But the AI hype is a liquidity event, not a fundamental re-rating. It is a toll booth operator adding faster turnstiles. Speed does not change the direction of the road. Verify the road before you pay the toll.
