Editor’s Note

In this post, Steve Kirsch responds to critiques of his KCOR method for analyzing COVID vaccine safety using Czech national data. The exchange features Stanford Professor Konstantina Stankovic, and illustrates a growing tension between credentialed experts and independent analysts over transparency and methodology in public health science.

Executive Summary

I just love it when people with academic credentials try to dismiss my arguments and methods with hand-waving critiques unsupported by evidence.

It never works.

Here’s the most recent example from Konstantina M. Stankovic, MD, PhD, FACS—an MIT/Harvard-trained auditory neuroscientist and skull base surgeon, now Professor and Chair of Otolaryngology–Head and Neck Surgery at Stanford. She also holds a courtesy professorship in Neurosurgery and was elected to the National Academy of Medicine in 2024.

She has 29 honors and awards and a very respectable h-index of 48 with over 6,600 citations.

By contrast, I’m just an Electrical Engineer with a couple of degrees from MIT in EECS and no medical credentials whatsoever.

In her critique, she claims I’m wrong. She called me a “deceiver” and a “con man.” I’m not buying it.

ChatGPT politely informs her that she and her graduate students are wrong on every point. KCOR was designed to be a conservative estimator of harm. So when it shows a safety signal, people should pay attention.

Rather than engage, she dismissed the AI rebuttal without citing any evidence. I gave her the link. She could have picked up the conversation and shown where ChatGPT went wrong—but she didn’t.

She could have:

• Called for transparency of COVID data. She didn’t.

• Shown the “correct” analysis of the Czech data. No interest.

• Proposed a better method than KCOR. Silence.

But she did none of those. Instead, she defaulted to saying I was wrong, and refused further discussion.

That’s how science works nowadays, apparently.

None of her colleagues have called for data transparency or provided an alternative analysis. They just attack methods that expose truths they missed.

The Conversation

See the full conversation here.

The conversation below was edited for brevity.

Dr. Stankovic:

First, like so many would-be analysts, you apparently over-relied on and failed to heed the warnings of AI systems. With my graduate assistants, we ran the Czech data with your “system” and found systemic errors:

1. Mortality doesn’t follow a straight line. It varies by comorbidities, lifestyle, and seasonality.

2. Mortality depends on comorbidities, not just age.

3. You oversimplify mortality into two numbers, ignoring other factors.

4. Hazard functions aren’t identical across age-matched groups due to other variables.

5. Your assumption of stable death ratios across time unless stressed is naive.

Epidemiology is not your strength.

My Response:

Time for you and your grad students to go back to school. You didn’t find a single valid point: https://chatgpt.com/share/685610a1-ad08-8009-a94c-63ab8e5510b9

ChatGPT Summary:

Her critique misunderstands your objective and method. She applies individual-level epidemiology to a population-level tool, misreads simplifications as literal claims, and misses how cohort-level frailty is inferred. Most of her objections are off-target or address things you never claimed.

Stankovic’s Students Reply:

Student 5: We know better than to rely on ChatGPT.

Student 3: You don’t understand epidemiology and are using AI to cover for it.

Everyone else went home, but I know what they will conclude, if only because I only pick the best of the best. You would do well to stick with the tried and true principles that we – as students and caretakers of the longstanding evidence-based principles we were taught over the course of our long careers in the field rely upon – and give up these foolish efforts to manipulate data to say what you want it to say, rather than what it actually states. I have no respect for deceivers and con men who attempt to “teach” and distort the AI systems – which is quite easy to accomplish – in order to achieve a false goal. You will always be discovered and found it for what you are. Always.

My Counterpoint:

AI was an objective third party. I didn’t ask it to agree with me—I asked it to judge argument quality.

I’ve shown this to Sander Greenland and Norman Fenton. Neither could find a hole. If we’re playing “appeal to authority,” they outrank your grad students.

If your team is so confident, show the correct analysis. Why haven’t you?

Lives are at stake. Let’s have a civil, on-camera debate or publish competing analyses. Don’t just walk away.

Stankovic replies:

You’re out of your league. Our critiques are standard epidemiology. You can’t respond to five basic corrections. This is over.

My Final Reply:

If you’re so capable, why can’t you or your colleagues show the correct analysis of the Czech data?

ChatGPT rebutted your arguments point by point. Claiming we didn’t respond is false. You simply couldn’t stomach the reply.

And no one in your profession has acknowledged that combining HVE + non-proportional hazards can make a placebo COVID vaccine appear 90% effective. Not a single paper accounts for it. Not one.

Also, where are your calls for data transparency? Why aren’t you demanding release of record-level data or sharing methods that work with KCOR?

You should be demanding answers—not attacking people who are asking the right questions.

Summary

I could have been super nice, but it wouldn’t have changed anything. She didn’t pull punches, so I didn’t either.

This exchange illustrates how credentialed experts often refuse to engage with novel methods that challenge their frameworks. Rather than test KCOR on real data or improve upon it, they cling to what they were taught.

If anyone believes they can produce a better, more objective analysis of the Czech data than KCOR, I’m all ears.

Show us your work.

Share