Gray market peptides: A harm reduction guide

Hi there,

This week I'm publishing something I've been asked about for months but most physicians won't touch.

Gray market peptides.

The "research chemicals" currently flooding Silicon Valley. Tech workers mixing powders in their kitchens. Injecting themselves with insulin syringes ordered from Amazon. Attending "peptide raves" with mix-your-own workshops.

People are doing this anyway. They deserve to know what a test report can tell them, and what it can miss.

This is a harm reduction guide. Not an endorsement.

The TL;DR if you only have 30 seconds:

• Peptides: Why a purity percentage cannot tell you an injectable is safe.
• Health AI: My experience with the January 2026 launches and their data connections.
• SleepFM: A research model ranks future disease risk from overnight recordings; it is not a personal forecast.
• Dietary guidelines: What the 2025–2030 update says about whole foods, dairy, and saturated fat.

Let's get into it.

💉 99% Pure ≠ Safe: A Physician's Guide to Gray Market Peptides

The New York Times covered this market in January 2026. The coverage brought more attention to questions patients had already been asking.

Coverage can miss the costs, unmet needs, and frustration that lead people to consider these products. We need to understand those reasons to have an honest conversation about risk.

The Promise: Why Intelligent People Take the Risk

I understand why this market exists. The traditional medical system has gaps, and patients are falling into them.

• Injury: People want faster repair. BPC-157’s promise comes largely from preclinical work; it has not established a safe, effective injury treatment.

• Weight management: Insurance exclusions and drug costs can make unregulated alternatives seem attractive. A research vial is not equivalent to an approved semaglutide or tirzepatide medicine.

• Persistent fatigue: Patients deserve to be heard. A proposed mechanism for MOTs-C or SS-31 is not clinical proof that a gray-market product treats their symptoms.

The desire for autonomy is valid. But if you step into this world, you must understand the difference between a patient and a "researcher."

What “Research Use Only” does and does not mean

A “Research Use Only” label states that a product is a research reagent, not a medicine for a person. It doesn’t establish the product’s identity, dose, sterility, or safety.

But of course, the biohackers purchasing these RUO peptides do indeed put them in their bodies, at their own risk.

That label does not let a seller bypass drug law or transfer all liability to a buyer. FDA has warned sellers that products marketed for human use can be unapproved drugs despite research-only labels.

The Purity Trap: Why 99% Doesn't Mean Safe

This is where I see smart people make the biggest mistake. You buy a vial, send it to a third-party tester like Janoshik or Finnrick, and get a report saying "99.5% Purity." You feel safe.

A purity percentage describes a result from a particular analytical method. It does not, on its own, verify the labeled dose or even establish identity. It also does not prove sterility or exclude endotoxins. A peptide can look highly pure in one test and still be unsafe to inject.

Gray Market vs. Compounding Pharmacies: The Critical Difference These two are often confused, but they are not the same.

• 503A compounding: Generally uses a prescription for an identified patient and follows the applicable pharmacy rules.

• 503B outsourcing facilities: Register with FDA, follow current good manufacturing practice requirements, and can supply healthcare facilities without a patient-specific prescription. A prescription is still needed to dispense a prescription drug to a patient.

• Gray-market sellers: A chemical storefront and a test certificate do not establish compliance with either system.

Check the exact compound and current rule: FDA has identified safety gaps for several proposed peptide products. Its lists and compounding policies have changed since this issue appeared. A nomination, advisory vote, or removal from a category is not the same as approval for human use. Ask the pharmacy to identify the current legal basis for the exact product and use.

​Read the Full Guide ​

In the full article, I break down:

• Why the route of administration changes the risk.
• Which tests address identity, dose, sterility, and endotoxins.
• Why testing one vial does not prove that every vial in a batch is safe.

Health AI: my January 2026 testing

In January 2026, I tried the new health features announced by OpenAI and Anthropic within days of each other. I wanted to see how well they could use my actual data.

The Strategy: Two Different Visions

ChatGPT Health emphasized bringing personal records and wellness data into one place. Anthropic’s launch also covered healthcare and personal data connections. Neither product is a substitute for a clinician’s diagnosis or treatment.

My data connections needed work

The connection problems made it hard to judge the analysis. The model’s reasoning was little use if it couldn’t get the right records.

My experience with both platforms was defined by friction:

• The "Eternal Loading" Loop: Integrations that were supposedly synced kept timing out when I actually asked a question.
• The Apple Health Lag: ChatGPT took over 24 hours to index my Apple Health data. In the age of real-time AI, waiting a day for your step count to sync feels archaic.
• The Data Gap: The Function Health integration, which was the headline feature, was surprisingly opaque. It gave me the number of abnormal biomarkers but struggled to pull the raw data for analysis.

Before connecting your health data

Read the product’s current connection and privacy settings. Check what data it can access, whether it has the records you expect, and whether you can verify an answer against the original report.

For ChatGPT (OpenAI):

For the launch context, see OpenAI’s January 2026 announcement. Its current instructions may differ from the menus I saw then.

For Claude (Anthropic):

For Anthropic’s launch context, see its January 2026 announcement. Availability and connector menus can change.

At the time, copying a verified report often worked better for me than an unreliable connection. That was my experience at launch; the connections may have improved since then.

SleepFM: predicting risk from overnight recordings

A research model can use overnight sleep recordings to rank future disease risk across a group of people. That is different from telling one person when they will get sick or die.

In a January 2026 paper, researchers trained SleepFM on more than 585,000 hours of clinical sleep recordings from about 65,000 participants.

They asked whether brain, breathing, and heart signals could help predict later diagnoses.

What the model predicted

For selected outcomes, the researchers reported a concordance index, or C-index:

• Dementia: 0.85
• All-cause mortality: 0.84
• Heart failure: 0.80
• Parkinson’s disease: 0.89

A C-index measures how well the model ranks risks between people. It is not “85% accuracy” for an individual diagnosis.

The next question is whether using these predictions improves care. This study did not establish a routine screening test.

The model combines several signals

The combined recordings carried information beyond a single sleep measure.

Brain activity, breathing, and heart signals all change across sleep stages. SleepFM analyzed their patterns together.

The results do not establish one “discord” mechanism as the cause of future disease. They show that patterns in the recordings contain predictive information.

Sleep studies may hold useful information beyond sleep apnea. Researchers still need to test how clinicians should use it and in whom.

For now, this is a promising research use of clinical sleep data, not a reason to buy a sleep test solely to forecast disease.

The 2025–2030 dietary guideline update

The 2025–2030 U.S. Dietary Guidelines put more emphasis on whole foods and allow full-fat dairy. The details matter more than a new graphic.

The big shifts:

• Full-fat dairy can fit the guidance.
• The document emphasizes protein, produce, and fiber-rich whole grains.
• It advises limiting highly processed foods.
• It still says saturated fat should generally stay below 10% of daily calories.

The gut health section is actually good, and the anti-processed food stance is overdue.

My gripe is still the labels. I want them to make ingredients and portions easier to judge. The current label already includes calories, fat, and added sugars; I would like it to say more about processing.

🗣️ The Debate: One Size Does Not Fit All

I posted about this on LinkedIn, and readers raised useful objections about how the guidance fits their own health.

Three specific points from the community that added necessary nuance:

• Individual response: Dr. Joseph Raffaele raised genetics and saturated fat. ApoE4 may change risk, but a genotype does not tell us that butter will cause Alzheimer’s disease in one person. Lipids and the overall diet still matter.

• Dietary patterns: Steve Ardire and Matthew Amsden asked how the guidance fits with research on long-lived populations. Those observations cannot isolate meat or any other single food as the cause of longevity.

• Additives: Toni Maraviglia raised questions about ingredients such as emulsifiers. Different additives need their own evidence; a long ingredient list does not establish an immune effect.

Federal guidelines provide a population-level starting point. Your health, preferences, lab results, and goals help determine how to apply them.

⚡ Longevity Quick Hits

The Science

🦠 CAR-T cells and gut aging: A mouse study used anti-uPAR CAR-T cells to target senescent cells and improve age-related intestinal problems. It doesn’t establish a safe human aging treatment.

👁️ A retinal patch trial: In December 2025, USC announced a phase 2b study for advanced dry macular degeneration. A small earlier study reported some vision improvement in 27% of patients. That is preliminary evidence, not proof that the treatment reverses blindness.

🧬 Immune cells in very old adults: The linked report discusses immune adaptation in exceptional longevity. These observations do not prove that making immune cells more aggressive will extend life.

Industry Pulse: JPM26 Updates

These announcements came from the January 2026 J.P. Morgan Healthcare Conference.

🤖 Lilly and NVIDIA: On January 12, 2026, the companies announced an AI research lab and up to $1 billion of investment over five years. The plans include robotics and drug-discovery models. An announcement is not proof that the full system is already operating.

🐶 Loyal’s dog-drug program: Loyal described progress toward conditional approval of LOY-002. Its January safety milestone did not establish a launch date or a regulatory rule for human longevity drugs.

Events listed in January 2026

Is Longevity Medicine a Bubble in 2026? Live discussion with Dr. Hillary Lin & Dr. Anant Vinjamoori – Join us virtually on LinkedIn Jan 14, 2026 @ Noon EST!

Physician Creator Trends – VSP Innovation Center & Matter Fireside Chat (Virtual) – Jan 28 @ 1PM EST link here. So excited about this topic!

Longevity: Living Longer, Living Better? Join myself and fellow longevity doctor Mishal Reja, MD in NYC for this intimate salon at Extension Health!

👱🏻‍♀️👩🏻‍🦰👩🏻👧🏽👧🏾 Livelong Women’s Health Summit – April 17-18, 2026, SF, CA. Delighted to join 50 other thought leaders in speaking on women's longevity!​

Offers listed in January 2026

This issue included an affiliate offer from NeuroAgeTx. I may earn a commission from qualifying purchases. Check current services and prices before buying.

The four-week MitoImmune trial found changes in selected immune-cell measures in fifty healthy middle-aged adults. It did not show fewer infections or longer life. This issue included a Timeline offer; check current terms.

GlycanAge reports an age estimate from IgG glycosylation. The estimate does not diagnose the cause of inflammation or establish a treatment.

More autonomy needs better information

All of these stories involve autonomy.

People are reading their own sleep data, considering compounds they can’t get through ordinary care, and changing their diets while guidance keeps changing.

Patients can now act on more information than ever. They still need a way to judge it.

My goal for 2026 is to help readers judge the evidence and avoid risks they did not know they were taking.

Stay safe, stay curious.

Hillary Lin, MD

Co-Founder & CEO

​Curator​

​

Follow me for more longevity insights: YouTube | LinkedIn | Instagram | TikTok​

Learn about Curator’s platform.