AI in Oncology — KOL Digest · Aug 18, 2026

This edition

JAMA: the case for AI alone, vibe coding, and a hallucinated peer-reviewed pub

In this edition: JAMA's new Perspective argues AI alone may exceed physicians on cognitive medical tasks · co-authors Zeke Emanuel, MD and Vinod Khosla make the case · Eric Topol, MD flags the argument · Yakup Ergün, MD largely agrees — and says what physicians still own · physician-founder Joseph Younis, MD on the paper's conflicts of interest · Santhosh Ambika, MD on workforce oversupply · Jeff Ryckman, MD on Abridge saving an hour a day — and on an EviCore denial built on AI-hallucinated literature · Jean-Charles Soria, MD on AI in pharma R&D · The ASCO Post's AI week in review.

JAMA 'AI alone' PerspectivePhysician reactionConflicts of interestAmbient scribesAI-hallucinated citationsPharma R&D

JAMAPublication@JAMA_current on X

International peer-reviewed general medical journal, JAMA Network.

Aug 17, 2026

First page of the JAMA Perspective 'Will Autonomous AI Exceed AI-Aided Physicians as the Best Medical Care?' by Ezekiel J. Emanuel, Abe Baker-Butler, Neal Khosla and Vinod Khosla, published online August 17, 2026, arguing AI-alone care is likely to be better than physician-only or physician-AI hybrid care at five cognitive medical tasks.

View post on X ↗

Zeke Emanuel@ZekeEmanuel on X

Oncologist and bioethicist. Vice Provost for Global Initiatives and professor, University of Pennsylvania. Co-author of the JAMA Perspective discussed in this edition.

Aug 17, 2026

View post on X ↗

Vinod KhoslaInvestor@vkhosla on X

Venture capitalist, founder of Khosla Ventures. Co-author of the JAMA Perspective discussed in this edition. Not a clinician.

Aug 17, 2026

Game mostly over for human doctors vs. AI? @ZekeEmanuel @AbeBakerButler @nealkhosla and I just published in @JAMA_current: When AI alone performance is consistently superior to human-alone performance, AI alone surpasses human-AI hybrids. Paradoxically, hybrid care in which humans are in the loop to correct AI errors is likely to worsen rather than improve AI performance. Medicine is rapidly approaching the transition point at which AI alone will exceed physicians and physician-AI hybrids in providing the best care.

@DrOz @1klomp time for @CMSGov @CMSinnovates to take note: Will Autonomous AI Exceed AI-Aided Physicians as the Best Medical Care?

jamanetwork.com/journals/jama/… ↗

View post on X ↗

Free Access · KOL Pulse Intelligence

Track oncology’s top KOLs in your specialty

Create a free account, pick the tumor types you cover, and go beyond the tweet — the intelligence pharma teams use to map influence and prepare for engagement:

✓Enhanced KOL profiles
✓Pharma influence & potential advisory-board patterns
✓2025 Open Payments financial analysis
✓Social-media sentiment & trend signals

Get free access — pick your specialty

Choose the tumor types you follow. No cost, unsubscribe anytime.

Neal KhoslaCompany executive@nealkhosla on X

CEO and co-founder of Curai Health, an AI-driven virtual care company. Co-author of the JAMA Perspective discussed in this edition. Not a clinician.

Aug 17, 2026

View post on X ↗

Eric Topol, MD@EricTopol on X

Cardiologist and physician-scientist. Founder and Director of the Scripps Research Translational Institute; author of Ground Truths.

Aug 17, 2026

Full page of the JAMA Perspective 'Will Autonomous AI Exceed AI-Aided Physicians as the Best Medical Care?' with passages highlighted, including sections 'When Does AI Alone Exceed Physicians?' and 'When Do Humans Alone Equal or Exceed AI Alone?', citing diagnostic-accuracy and guideline-concordance studies.

View post on X ↗

Yakup Ergün, MD@dr_yakupergun on X

Medical oncologist, Bower Hospital.

Aug 17, 2026

Page 1 of 4 of the same JAMA Perspective 'Will Autonomous AI Exceed AI-Aided Physicians as the Best Medical Care?', downloaded from jamanetwork.com, laying out the authors' case that AI-alone medical care is likely to be better at five fundamental cognitive medical tasks.

View post on X ↗

Joseph Younis, MDPhysician-founder — not an oncologist@YounisJoseph on X

Physician-founder working on AI in research, medicine and medical training. Not an oncologist.

Aug 18, 2026

Let me make a few important qualifications:
1. First and foremost, the COIs: One of the authors is the CEO of Curai Health, which unsurprisingly is a AI physician startup. Another author is the investor in Curai Health.

2. "Published" is actually an overestimation of the qualitative threshold of this "review": This is, in fact and contrary to the paper's claim, NOT a "review" or "of all published articles on AI in medicine since Jan 1 2024." There is no review methodology, databases searched, search queries/strings, inclusion/exclusion criteria, PRISMA structure, risk-of-bias assessments, preregistered protocol, or really anything scientific.

3. Cherry-picked epistemology: When a paper showed physicians or physicians + AI > AI, they rationalize it as "old LLM model" or "sub-optimal prompting." Vice versa is, of course, evidence AI > physician. Its a unidirectional epistemic criteria.

4. Endpoints are proxies rather than real clinical outcomes: Correct diagnosis/appropriate test/guide-line concordant treatment/simulated cost/top-1 diagnosis vs. mortality rate/re-hospitalization rate/adverse drug events/morbidity/healthcare utilization rate

5. The review is misleading: they cite an RCT with their claim that "AI is better at chronic management for conditions including diabetes" but in that RCT before the AI was activated, each patient's diabetes clinician selected the protocol... So in other words, it was a human-based AI management

6. The review is again misleading, and possibly confused: They cite a meta-analysis with their claim that "When AI is better than humans, adding a human can actually make AI performance worse." This paper covered 106 experiments, of which only about 20 were medicine related. And here's the kicker... a human always signed off on the AI-generated medical plans.

tl;dr: Dad funds son's health startup and write an opinion piece. Embarrassingly low scientific comprehension by the authors but they're the equivalent of 'non-technical' in this domain, as they like to say in Silicon Valley, so we'll give them a pass. But, @JAMA_current do better? I get it's a perspective but like... at least make sure the studies cited say what the perspective cites them for?

@CuraiHQ if need some consulting before waging war against physicians, I am open for a curbside peer-review next time

View post on X ↗

Santhosh Ambika, MD@RenoHemonc on X

Hematologist-oncologist, Reno, Nevada.

Aug 17, 2026

View post on X ↗

Jeff Ryckman, MD@jryckman3 on X

Radiation oncologist and former medical physicist; builder of RadOncCalc.

Aug 18, 2026

has allowed me to save about an hour a day doing notes, using more accessible language than my notes prior to abridge. However, even if I am seeing a female patient, it will often attribute my voice to the patient or their family members and incorporate it in the HPI or AP section.

Has the @AbridgeHQ team ever thought about storing a voice clip of the clinician +/- additional HCWs who may be in the room within the local healthcare system to be able to tell HCW vs patient and family member voices, say, by using software such as FluidAudio + sortformer, NeMoSpeaker-iOS or speech-swift to be able to tell HCW vs. “other” more reliably?

I was able to vibe code software that uses FluidAudio + Sortformer just over this past weekend using Claude Code and Grok on two independent projects, so I’m thinking this may be low hanging fruit if felt to be of value by the @AbridgeHQ team! Thanks for building such a great tool 🙏

View post on X ↗

Aug 13, 2026

“Reliance on AI-Hallucinated Literature

EviCore bases its non-coverage decision on a recent meta-analysis (PMID: 40888940). Independent review of this publication reveals pervasive LLM hallucinations and fabricated data points. Notably, the lead author operates a commercial platform (clever-academy.com ↗) selling access to automated AI tools designed specifically to rapidly create these reviews. Formal misconduct investigation and retraction requests have been submitted to the journal Editor-in-Chief and publisher (Springer Nature). Unverified, AI-churned literature fails the sound scientific evidence threshold mandated by CMS.”

@evicorehc 👀

View post on X ↗

Jason Beckta, MD, PhD@drbeckta on X

Radiation oncologist.

Aug 12, 2026

Great idea, Sean. Don't mind if I do:

"This filing requests immediate CMS intervention to suspend EviCore's November 2026 policy update classifying Low-Dose Radiation Therapy (LDRT) for Osteoarthritis (OA) as "not medically necessary". EviCore has classified LDRT for OA as medically necessary since at least 2019. Under 42 CFR Section 422.101 (CMS-4201-F), internal utilization management criteria must be grounded in widely accepted clinical practice guidelines and sound, peer-reviewed scientific evidence. EviCore's reversal violates federal standards by relying on AI-hallucinated literature, distorting federal research, contradicting major clinical guidelines, and acting arbitrarily following national media exposure.

Reliance on AI-Hallucinated Literature

EviCore bases its non-coverage decision on a recent meta-analysis (PMID: 40888940). Independent review of this publication reveals pervasive LLM hallucinations and fabricated data points. Notably, the lead author operates a commercial platform (clever-academy.com ↗) selling access to automated AI tools designed specifically to rapidly create these reviews. Formal misconduct investigation and retraction requests have been submitted to the journal Editor-in-Chief and publisher (Springer Nature). Unverified, AI-churned literature fails the sound scientific evidence threshold mandated by CMS.

Distortion of Federal Evidence (2024 VA ESP Report)

EviCore cites the 2024 VA Evidence Synthesis Program report to justify non-coverage. This directly distorts the report's explicit executive summary conclusions:

"Despite the gaps in the evidence, we found no indication that RT should not be used after conventional therapy fails for [osteoarthritis]."

The report formally assesses that "there is equipoise about the clinical utility of RT in patients failing conventional therapies."

A finding of "clinical equipoise" combined with an explicit statement that there is no indication against second-line use supports patient access when conservative measures fail, not an administrative ban.

Direct Contradiction of Recognized Clinical Guidelines

CMS-4201-F mandates alignment with recognized multispecialty guidelines:

2026 American Radium Society (ARS) Guidelines (Radiation Oncology, Rheumatology, Orthopedics): Classifies LDRT as "Usually Appropriate" for refractory knee OA and "May be Appropriate" for hip, hand, foot, and shoulder OA (60%-90% response rates, positive RCT evidence).

2022 DEGRO Guidelines: Recommends LDRT (Category C) for knee/hip OA with zero secondary malignancies documented in literature.

Paradoxically (and incorrectly), EviCore cites the DEGRO guidelines to support its new non-coverage decision. Though these guidelines have been published for the last four years, EviCore has not cited them until now.

Arbitrary and Capricious Timing

PDF metadata shows EviCore's "not medically necessary" guideline was created on June 17th, 2026, at 11:13 AM, approximately 22 hours after NBC News/TODAY published a national profile highlighting LDRT's benefits on June 16th, 2026, at 1:08 PM. Reversing a 7-year coverage standard within 24 hours of major national media exposure reflects a reactive effort to curb patient volume rather than genuine, evidence-based clinical evaluation.

REQUESTED ADMINISTRATIVE RELIEF

To ensure our patients with refractory osteoarthritis do not lose access to effective care, I urge CMS to immediately pause EviCore's November 2026 non-coverage policy and maintain existing coverage criteria while initiating a compliance review. Before any future coverage modifications are permitted, EviCore should be required to undergo an independent, multispecialty clinical review alongside practicing radiation oncologists, rheumatologists, and orthopedic surgeons. Halting this hasty policy reversal will give CMS time to ensure utilization criteria reflect true clinical consensus rather than flawed administrative decisions."

Screenshot of a CMS.gov Provider Complaints Form confirmation page reading 'Submission successful' with a partially blurred confirmation number - the complaint about a Medicare Advantage plan's use of EviCore referenced in the post.

View post on X ↗

Jean-Charles Soria, MD, PhD@jsoriamd on X

Medical oncologist. Professor, Université Paris-Saclay; former Director General of Gustave Roussy.

Aug 14, 2026

AI in Pharma R&D: time to move from better models → better medicines
• 🎯 Optimize what matters: improving Phase II success has far greater potential impact than simply accelerating hit discovery
• 💊 Ligand ≠ drug: AI can generate binders rapidly—but efficacy requires the right biology, exposure, safety & patient
• 🧬 Better data ≠ more data: biological data are highly contextual; what matters is predictive validity for human outcomes
• 📊 Benchmark ≠ impact: model accuracy matters only if it improves a real R&D decision—advance, redesign or stop
• 🔄 Next frontier: start with the clinical question → generate fit-for-purpose human biology → integrate clinical feedback → continuously improve the model
👉 Perhaps the end of the beginning: less “technology push,” more “science pull.”
#AI #DrugDiscovery #DrugDevelopment #PrecisionMedicine
nature.com/articles/s4157… ↗

Journal table of compound properties in AI-driven drug design: rows for on-target activity, physicochemical properties, off-target activity/selectivity, in vitro ADME/PK and in vivo ADME/PK, comparing each property's consideration in ligand design versus what drug discovery actually requires.

View post on X ↗

The ASCO PostPublication@ASCOPost on X

Oncology and hematology news publication, produced in partnership with ASCO. A publication, not an individual clinician.

Aug 15, 2026

Still from an ASCO 'AI in Oncology' news video showing an AI-generated news anchor in glasses and a dark blazer in a virtual studio. — Still from ASCO's AI in Oncology weekly news video, linked in the post.

View post on X ↗