Geoffrey Gewurz

Investor for 40+ years. Focused on markets, trading, long-term compounders, precision medicine, and diagnostics. Caris shareholder. Personal views only.

[Geoffrey Gewurz @GG57TLV]

Important work from Aditya Shreenivas, MD MS, Razelle Kurzrock, MD, and collaborators, using Caris data to profile 206,750 tumors across the HER2/ERBB2 landscape.

That dataset size is not trivial. In oncology, most studies are still limited by narrow cohorts, single tumor types, or fragmented biomarker data. A real-world analysis across more than 200,000 solid tumors allows patterns to emerge that smaller datasets simply cannot resolve.

The key point is simple: HER2 biology is more complicated than ���amplified or not amplified.”

ERBB2 alterations, HER2 IHC expression, RNA expression, tumor type, and prognosis can diverge in meaningful ways. One particularly interesting finding is that ERBB2 mutations may be associated with increased HER2 IHC expression even without amplification.

That matters as HER2-directed therapies and ADCs expand beyond traditional HER2-positive breast cancer. The broader HER2/ERBB2 landscape needs to be mapped cancer by cancer, alteration by alteration, and modality by modality.

Great example of how academic leadership and industrial-scale oncology data can come together to make precision medicine more precise.

@Dr_R_Kurzrock

[Aditya Shreenivas MD MS @AdityaSunavs]

Was worth the wait, sharing our real-world analysis of HER2 across solid tumors! 206750 tumors profiled by @carisls for ERBB2 alterations, revealing complex relationships between ERBB2 alterations and HER2 IHC expression. Forever grateful to @Dr_R_Kurzrock for her mentorship.

[Caris Life Sciences @carisls]

Precision medicine is transforming the way cancer is treated and changing the way oncologists care for their patients. With Caris testing, Dr. Christopher Vaughn has gained deeper insight into the origins of each patient’s cancer, turning that knowledge into more informed, personalized treatment decisions. Learn more: https://t.co/iTJJdNKW2t

[Geoffrey Gewurz @GG57TLV]

1/ Trying to pick a bottom in a single name is two bets pretending to be one. You bet the sector turns. You bet you picked the right horse. The second bet is the one that ruins you.

2/ Biotech last year bottomed in July. Own the group, you caught the move. Buy the wrong single name in June, you sat in dead money while the index doubled. Right thesis, wrong ticker, no clean exit.

3/ Consumer internet is in that same moment now. Several names near their lows on a mix of real and imagined fear. The clean expression is a basket. Buy the turn, let the group sort out the winner.

4/ Inside that basket I favor Zillow. Pinterest's upside rests almost entirely on international monetization. Most of the users, almost none of the revenue, twenty cents abroad against near seven dollars at home. Hard and slow to close.

5/ And Pinterest only earns a premium multiple if the market sees it as a scarce visual asset someone strategic buys. Without that takeout frame it is a social network, and social is under multiple pressure. I do not want a thesis that needs a buyer to show up.

6/ Match throws off more cash than Zillow and gets paid for it. What is underpriced is the risk it carries. Demand destruction. If a younger generation stops dating through apps, the network runs in reverse fast. That is a cliff, not a slope.

7/ Zillow's risk is friction, not demand. People will search for homes forever, in any rate environment. The only question is whether that demand keeps flowing through Zillow or gets intermediated by Google and AI search. Real, but slow and contestable.

8/ And fightable. Lock listings ahead of the MLS, deepen Home Loans and closing, lean on traffic that already arrives direct. A slow risk you can fight is a different animal than a fast risk you can only watch.

9/ The enterprise value makes the point. Zillow trades near half of Match's EV while growing faster, holding net cash instead of net debt, addressing a far larger pool. The market has the riskier demand base priced at the premium.

10/ The prize is asymmetric. Zillow's earnings sit in a frozen housing market. Volumes come back, earnings step up off a low base, and if the capture held you get operating leverage and a re-rating at once. Two levers. One hundred to two hundred dollars is not a fantasy on that path.

11/ The behavior is forever. The capture is Zillow's to lose. I am betting it does not.

[Geoffrey Gewurz @GG57TLV]

I should probably revise something I once said too categorically.

I once argued that GRAIL would never get FDA approval for Galleri because the test did not meet the key clinical endpoint in the NHS-Galleri trial. That may be too simplistic.

FDA approval does not necessarily require proving full clinical utility, mortality benefit, or that the test changes population outcomes. In many cases, the FDA can approve a diagnostic test if the company shows analytical validity, clinical validity, and an acceptable benefit-risk profile. In plain English: the test does what it says it does, the claims are supported, and the harms from false positives/false negatives and downstream workups are judged acceptable.

So yes, GRAIL could still potentially get FDA approval.

But that is not the same thing as saying the product works commercially.

FDA approval does not automatically mean CMS reimbursement. It does not automatically mean private-payer coverage. It does not automatically mean doctors believe the test has enough utility to use it broadly. And it does not automatically mean the company can make money if its cost structure is not aligned with the revenue it can actually collect.

That is the real issue.

A test can be clinically valid — meaning the claim it makes is basically correct — and still fail commercially if payers do not see enough utility, if physicians do not adopt it, if guidelines do not support it, or if the economics do not work.

This is especially important in MCED. A positive result in a healthy person can trigger imaging, referrals, biopsies, anxiety, and cost. So payers are not only asking, “Is the test valid?” They are asking, “Does this improve outcomes enough to justify the system-wide cost?”

That is why FDA approval, if it comes, would be meaningful — but not decisive.

The real hierarchy in diagnostics is something like:

CLIA lets you operate.
FDA can validate the claim.
CMS and private payers decide whether you get paid.
Guidelines and physician behavior decide whether you become standard of care.
Unit economics decide whether the company actually works.

So the right conclusion is not “GRAIL can never get FDA approval.”

The right conclusion is: even if GRAIL gets FDA approval, that does not prove Galleri becomes a reimbursed, broadly adopted, profitable population-screening product.

Regulatory validity is not the same thing as commercial inevitability.

[Geoffrey Gewurz @GG57TLV]

Regarding $GRAL. This is a company that before the blow up was doing 40 million a quarter and trading at 30x revenue for a test company! Not a data rich operating system, profitable like $CAI. The entire story was based on FDA approval which will not come. The bar for FDA is adding years of life for patients and detecting at earlier stages. I am shouting this everywhere, Caris Life Sciences is the only company in the universe that will meet that standard when Achieve 1 readout will come in the next month. For those holding positions in $GRAL in a best case scenario they are worth 5x revenue which is 17 dollars. More likely that they will not survive. I don’t trade, don’t know or care what shorts, longs or algos will do. This is the fundamental analysis and I stand behind it.

[Geoffrey Gewurz @GG57TLV]

Worth reading. Antonio just started on Caris and already caught the things most people miss. One engine feeding many products. The flip to cash. The pricing edge. That’s a strong first read.

One thing he priced wrong, and it’s the one that matters. He measured the sequencing. How many genes, how many patients. That’s the floor, not the asset.

The asset is the outcomes data sitting on top. Seven hundred thousand patients with years of follow-up, heading past a million. That is what lets Caris build an early detection test. A six-hundred-gene panel cannot get there. And early detection is the bigger market by a wide margin.

So the race he scored was the small one. The big one is not even open to the panel model.

[Antonio Linares @alc2022]

$CAI and $TEM are the same thing, except for two caveats:

1. $TEM operates at a larger scale.
2 $CAI prioritises depth over breadth and is already printing cash.

Don't be fooled by current market caps.

This is a trillion dollar race to become the dominant Biology Ontology.

[Geoffrey Gewurz @GG57TLV]

Reduction and Repetition (Part 2 of 2)

Last time I argued that Palantir and Astera Labs prove the same thing in opposite styles: a company doesn't need a particular persona, but the market needs to know what to repeat. War room or engineering lab, either works, as long as the story reduces.

So where do the four big names in precision oncology sit? Guardant, Natera, Tempus, Caris.

One distinction first, because it does most of the work: clarity is not the same as loudness. A company can reduce its story beautifully and still be reducing the wrong sentence, one that runs ahead of what the business has proven. Loud and repeatable isn't the same as accurate.

Tempus sits closest to the Palantir end. Founder-led, AI-forward, data-heavy, big-platform language. The market repeats it without effort: AI plus multimodal medical data makes medicine smarter. That's a clean reduction. The question isn't whether it's repeatable, it plainly is, but whether it's running at the right distance from the business or a step ahead of it.

Natera is the opposite temperament, closer to Astera. No prophet, no drama, an execution machine. Cell-free DNA platform across women's health, oncology, organ health, reimbursement, scale. Repeated commercially, relentlessly. It works because the sentence and the business run at the same speed: the story isn't ahead of the P&L, it's describing it.

Guardant already had a great reduction, and that's its problem. "Liquid biopsy" made the company instantly legible. But Guardant is now two stories with two buyers and two clocks, therapy selection on one side, screening on the other, wearing a single phrase. The original sentence no longer tells you which movie you're watching, and a phrase that once clarified can start to blur.

Caris is the hardest to place. Not Palantir, not Astera, not Natera. Founder-CEO, scientist-president, major clinical voices, pharma relationships, AI, tissue and liquid, WES and WTS, MI Profile, Assure, Detect, MI Clarity, ChromoSeq, POA. On the asset side, the breadth may be the whole advantage. On the storytelling side, it's a liability, because the market cannot repeat a list.

Tempus has "AI medicine." Natera has "cell-free DNA platform." Guardant had "liquid biopsy." Caris doesn't have the phrase yet, four audiences each seeing one face, none holding the whole. And the breadth makes that the expensive problem, not the small one: the market prices what it can repeat, so a company that asks it to repeat a list gets priced as a pile of parts, including the integrated whole that's supposed to be worth the most.

The exact words don't matter. The discipline does: one identity, repeatable in a breath, said until the market says it back.

Complexity without reduction is noise. Complexity with repetition becomes a category.

Caris investor.

[Geoffrey Gewurz @GG57TLV]

Reduction and Repetition (Part 1 of 2)

I've been studying the storytelling architecture of recent IPO winners, how a company shows itself to the market, separate from the financials or the product spec.

Two examples could not be more different.

Palantir is narrated by one voice across fifty channels. Usually Alex Karp, sometimes Shyam Sankar, but whoever speaks, the message never changes, and it isn't really about software. Karp rants about the West, institutions, adversaries, national security, why the world is more dangerous than people admit. Palantir feels like a war room.

Astera Labs is the opposite. Three cofounders in chairs, presented as equals, calmly explaining connectivity bottlenecks and rack-scale architecture. No prophet. No drama. Astera feels like an engineering lab.

Both have been rewarded enormously, each up roughly tenfold or more from their lows. And here's what people miss: the persona fits the product.

What Astera sells is boring, the plumbing between chips, and a calm team of engineers is exactly right for it. What Palantir sells is mission-critical, the difference between a missile hitting in one second or not. Karp's intensity matches the stakes. Each narrator is connected to what's narrated.

Maybe none of this is the cause. Both stocks may just be riding the same AI wave, the narrative a passenger on a move the market would have made anyway. I can't disprove that. But the AI tide doesn't explain the difference in how these two are understood, or why each story fits its business so precisely.

This isn't a post about hype either. Hype is when the story runs ahead of the business, and that breaks. Both delivered. If you don't deliver, no story saves you.

The real question: when a company is delivering, how does it realize the full potential of its stock currency? That's where storytelling matters, as translation. One reduces emotionally, one technically. Same discipline underneath: make the market understand what's changing, why you matter, and repeat it back.

The market doesn't need every company to have an Alex Karp, or to sound like engineers in a lab. It needs to know what movie it's watching.

War room? Engineering lab? Clinical science? Data infrastructure? A platform hiding inside a product?

Part 2: how the four big names in precision oncology (Guardant, Natera, Tempus, and Caris) score on exactly this.

[Geoffrey Gewurz @GG57TLV]

A quick thought on Caris Life Sciences vs Personalis.

I like Personalis.

It is doing important work in ultrasensitive MRD, and the company may become very relevant in recurrence monitoring.

But I do not think the market should view Personalis and Caris as similarly positioned companies.

Caris has the whole stack.

It has comprehensive genomic profiling.

It has more than a decade of accumulated oncology data.

It has tissue, blood, DNA, RNA, AI, pharma relationships, physician workflow, and a growing clinical platform.

It has launched ChromoSeq.

It has launched MI Clarity.

It has Detect coming in MCED.

It has MRD coming, and I expect that product to be very competitive.

And over time, the most important layer may not be any single test. It may be the decision layer: how the system helps physicians decide what to do next across profiling, recurrence risk, MRD, early detection, and treatment selection.

That is the difference.

Personalis may have a very good product.

Caris is trying to become oncology infrastructure.

Financially, the contrast is also clear.

Caris is guiding to roughly one billion dollars of revenue this year, has turned free cash flow positive, and has already authorized a one hundred million dollar buyback despite being a very young public company.

Personalis is much smaller, guiding to roughly eighty million dollars of revenue, and is still in the investment phase.

That does not make Personalis bad.

It just means the strategic position is very different.

Personalis is building an important MRD company.

Caris already has the broader operating system: CGP, data, AI, pharma, Clarity, ChromoSeq, MCED, MRD, and eventually the clinical decision layer.

In diagnostics, the dataset and the workflow position matter as much as the assay.

That is where Caris has the structural advantage.

[Geoffrey Gewurz @GG57TLV]

https://t.co/UyjQq1tswh

[Geoffrey Gewurz @GG57TLV]

Post 3: The commercial engine

The next question is whether Caris is actually building the commercial engine for Detect.

Julie Whelan is now Head of Marketing at Caris, with major consumer-brand experience from Harley-Davidson. She is not there only for Detect. Caris is building a broader brand and commercial engine across the company.

They are assembling a regional team, adding direct salespeople, using Everlywell as a channel, and likely targeting concierge medicine. Over time, international self-pay markets can probably be approached through agents and partners.

Then there is the factor nobody can model.

What happens if Elon Musk does the test and talks about it?

What happens if Joe Rogan has someone on to discuss MCED seriously?

What happens if Andrew Huberman frames it as part of a premium prevention stack?

What happens if a billionaire founder like Larry Ellison, who already discussed AI-driven cancer detection on the White House lawn, publicly says this is the direction medicine is going?

That is not in the stock because it cannot be in the stock. It has not happened yet.

So this is the seesaw.

On one side: price, brand, reimbursement, historical self-pay skepticism, and the difficulty of building a new medical behavior.

On the other side: category education, repeated marketing, multiple companies creating the norm, Caris’ deeper architecture, commercial buildout, wealthy self-pay demand, grief, fear, prevention, and awareness.

Nobody knows which side wins yet.

But MCED is not really being valued as upside today. It is almost being treated like a cost, a weight, or a source of uncertainty.

That may be backwards.

Profiling, MI Clarity, ChromoSeq, the pharma/data platform, and the existing oncology data assets may already justify much more than the current market cap. MCED is not needed to make the stock interesting.

But if Detect works commercially, even partially, it changes the discussion.

Then MCED is not a cost.

It is a profit center.

I am a shareholder.

[Geoffrey Gewurz @GG57TLV]

Post 2: The commercial quandary

The real question around Caris Detect is commercial.

Can a premium, expensive, self-pay cancer detection test become a real market before broad reimbursement?

That is the quandary.

Earlier cancer detection is becoming a collective effort. Guardant Shield, Exact/Cologuard, Freenome, Natera, Grail, and others are training the market in different ways. Grail has done the heavy lifting in MCED. Caris is coming with a deeper WGS/WTS-based approach, which today is premium, expensive, and harder to model.

Premium self-pay medicine is hard. Most medical products do not scale without reimbursement. Analysts are right to be skeptical until there is real volume.

But here is the heart of the quandary.

There may be 20 million people in the U.S. and another 50 million globally who can afford this test if it becomes part of the premium prevention conversation.

Most of them are over 50. Many have lost a friend, a parent, a spouse, a sibling, or a business partner to cancer. Almost all of them know someone whose life was changed by cancer. Some are fearful of disease. Some are health-obsessed. Some are probably hypochondriac. That may not sound elegant, but it is real human behavior.

So if they are offered a premium cancer detection test, and they know it exists, what percentage actually does it?

I have no idea.

Is it 0.1%?

Is it 1%?

Is it 5%?

That is the whole commercial question.

It is either a rocket, or it is a dud. The flower either blooms, or it dies on the vine. Nobody knows yet.

For many of these people, price is not the real obstacle. They spend that amount on a weekend, a vacation, one hotel night, a watch, a dinner, or a wellness program.

The real obstacle is awareness, marketing, and norm creation.

People need to know this exists. They need to hear about it from physicians, concierge doctors, friends, podcasts, public figures, and multiple companies attacking the problem from different angles.

Over time, the question has to move from “what is this?” to “should I be doing this?”

That is how a self-pay medical behavior forms.

More in Post 3: whether Caris is actually building the commercial engine to make that behavior possible.

[Geoffrey Gewurz @GG57TLV]

Post 1

I asked ChatGPT and Claude to model what Caris Detect could do in the self-pay MCED market.

They both basically gave me the same answer: slow launch, maybe 10,000 tests in the first six months, maybe 20,000 to 30,000 the next year, and then a gradual ramp from there.

And honestly, I understand why.

The models start with Grail, and Grail deserves real credit for creating the MCED category. We are after several years of Grail educating the market. Without Grail, most investors and consumers would not even know what multi-cancer early detection means. So this is not day one anymore.

Then the models do what models do. They adjust down.

Caris is not yet a consumer brand. Detect is more expensive. Broad reimbursement is not here yet. The self-pay medical market has very few examples of real scale. LASIK / laser vision correction is probably one of the better historical examples, but even that is very different. LASIK was visible and immediate: you could not see well, you paid, and then you saw better. Cancer detection is invisible, emotional, scientific, and trust-dependent.

So I do not think analysts are crazy to discount MCED today.

The conservative spreadsheet is doing the logical thing. It starts with Grail, adjusts down for price, adjusts down for brand, adjusts down for uncertainty, and gives Caris Detect very little value.

That is one side of the seesaw, and it may be right.

But the other side of the seesaw is where this gets interesting, because the spreadsheet is looking mostly at adoption history. It may not be looking hard enough at product differentiation, wealth, grief, fear, belief, and amplification.

That is Post 2.

[Geoffrey Gewurz @GG57TLV]

Caris opened up roughly 6% today, faded to down roughly 4%, and then recovered slightly into the close.

With my market experience, I thought there was a real chance the buyback would be taken the wrong way.

We have recently seen buybacks rewarded in beaten-down names. Pinterest bought back approximately $2B around the first quarter. Zillow bought back roughly $600M to $700M and authorized an even larger program.

But those companies are more mature, generate far more cash, and are not launching four or five major clinical products over the next year.

Caris is different.

Caris is launching liquid profiling, Detect/MCED, MI Clarity, ChromoSeq, MRD, AI pathology, and pharma discovery. So a $100M buyback naturally raises the question: should that cash be going into R&D, trials, reimbursement, and commercialization instead?

I understand the concern.

I also saw the R&D comparison on X: Guardant spends roughly 3x Caris’ R&D, and Natera roughly 6x, including women’s health.

That looks concerning at first glance.

But it misses the operating leverage question.

Caris spent more than a decade building the platform: WES, WTS, matched normal, tissue and blood integration, AI pathology, clinical data, pharma relationships, and a longitudinal oncology database.

If each new product needs its own separate R&D engine, the market is right to worry.

If these products are extensions of the same molecular engine, lower R&D intensity is not weakness. It is the thesis.

Anyone who really wants to understand the company should watch Dr. David Spetzler’s recent ASCO interview. That interview explains why Caris is not just running tests. It is building a comprehensive molecular engine.

That said, the volatility is real. Ten percent intraday swings have been almost normal since Caris went public. A $100M buyback may signal confidence, but it will not fix a broken chart.

Execution will.

Clinical adoption. Reimbursement. Recurring volume. And eventually, something like 100,000 Detect/MCED tests.

The buyback is not the thesis.

Platform leverage is.

I am a shareholder.

[Caris Life Sciences @carisls]

Caris Life Sciences announced that its Board of Directors has authorized a share repurchase program for up to $100 million of the Company’s outstanding common stock.

Learn more: https://t.co/sgcrkrA8eX

[Geoffrey Gewurz @GG57TLV]

R&D dollars matter, but architecture matters more. Guardant and Natera are spending heavily to win specific categories. Caris is trying to leverage one comprehensive molecular engine across profiling, liquid, MCED, Clarity, ChromoSeq, MRD, and pharma. If that architecture works, lower R&D intensity is operating leverage, not weakness. A $100M buyback authorization does not mean Caris is choosing buybacks over science. It means management likely believes the stock is materially undervalued while still investing in trials, commercialization, and platform expansion. That is not starving R&D. That is capital allocation.

[Geoffrey Gewurz @GG57TLV]

This speaks for itself.

ASCO26: Agentic AI supercharging genome sequencing for early cancer dete... https://t.co/ostdBiCfgp דרך @YouTube

[Geoffrey Gewurz @GG57TLV]

Another great one. Only Caris has a real test.

https://t.co/LHlPYwHXqs

[Geoffrey Gewurz @GG57TLV]

Grail, interesting post.

[Geoffrey Gewurz @GG57TLV]

@LogicalThesis Ok

[Geoffrey Gewurz @GG57TLV]

@LogicalThesis @shortbus_ace Your thesis is right on. This is the knowledge layer. The leaders will be large, very large.