Summary

What can currently be managed reliably in gut health is “function,” not an idealized roster of microbial species. Healthy adults carry more than 100 billion bacteria and over 1 million microbial genes, yet the microbiomes of any 2 people are often only 10%—20% similar. The health signals researchers can observe are mainly higher diversity, stronger short-chain fatty acid production, and greater resilience to disruption—not a target mix such as “one species at 10%, another at 5%.” 史寒朵’s summary: “Unhealthy guts are all alike, but every healthy gut is healthy in its own way” (“不健康的肠道都是相似的，但是健康的肠道各有各的不同”).
Consumer microbiome tests costing hundreds to thousands of yuan can usually measure accurately, but rarely produce an actionable optimization plan for asymptomatic people. Greater sequencing depth makes it more likely to detect rare species such as Clostridium butyricum below 1%, or even below 0.1%; science still has no benchmark for the “optimal microbiome.” For healthy users, the result is often little more than “one more data point for science.” Clinical settings such as inflammatory bowel disease and C. difficile infection are different: doctor-ordered testing can support diagnosis and affect treatment.
Probiotics are not pure “IQ tax,” but for most healthy people they look more like a short-term service requiring ongoing payment than a one-off ecological reset. A typical capsule contains around 10^9—10^10 live bacteria and may increase Bifidobacteria and related metabolites during use; once discontinued, external strains usually disappear within days. 寒朵’s view is blunt: “It is perfectly possible from the standpoint of the human body, but from the standpoint of your wallet, that is a very personal question” (“从你的钱包上来说，这是一个很个人的问题”).
The higher-return intervention is to keep feeding the existing microbiome, rather than betting on a single “star strain.” In a roughly 6-week intervention, participants added 2 servings a day of high-fiber or fermented foods, each about the size of a fist, to their usual diets. Both groups improved on microbiome and health measures, with the fermented-food group showing a particularly pronounced decline in inflammatory markers. Many participants kept eating more of those foods after the intervention, suggesting that “sticking with it for 6 weeks” may create a habit, not just a temporary change in biomarkers.
Evidence for exercise, antibiotics, and dietary changes varies, and cannot be packaged into a single “microbiome formula.” Interventional studies support moderate exercise as a way to improve short-chain fatty acid production and reduce some inflammatory markers; sleep evidence remains mainly correlational and animal-based. Antibiotics clearly disrupt the microbiome, and taking probiotics alongside them can reduce diarrhea, but has not been shown to accelerate long-term recovery. Most healthy people return close to their original state within several months to half a year.
The claim that “the gut determines depression and anxiety” has run far ahead of the evidence. The current chain of support comes mainly from mouse behavior, serotonin production in vitro, and psychiatric drugs affecting bacterial growth in test tubes. It has not been shown that these changes generate enough molecules in humans, enter circulation, and reach therapeutic concentrations. 寒朵 sees the microbiome as “adding one more link to this complex system,” while using ordinary probiotics to treat mental illness leaves “far too many gaps in the chain” (“这个链条现在的缺口还实在是太多了”).
The microbiome therapies with genuine drug potential are concentrated in reproducible clinical endpoints, not broad wellness marketing. Fecal microbiota transplantation is already a first-line treatment for recurrent C. difficile infection, with 寒朵 recalling cure rates of roughly 80%—90% and most patients requiring only 1 treatment. Voust in 2023, defined microbial consortia, and engineered bacteria are all attempts to turn invasive FMT into standardized oral products. By contrast, the US FDA regulates ordinary probiotics as dietary supplements, requiring safety but not efficacy—a clear dividing line between consumer products and microbial drugs.

Deep dive

1. The Human Body Is More Like a Microbial Community Than an Independent Host

史寒朵 illustrated the scale by pointing to the “more than 100 billion” bacteria in a healthy adult: bacterial cells may be comparable to, or even outnumber, human cells. Humans have roughly 20,000 genes, while gut bacteria collectively contribute more than 1 million. The industry’s exaggerated but vivid line is: “We are actually more like bacteria than humans” (“我们其实更像是细菌，而不是人类”).
Most gut microbes live in the large intestine, feeding on food residue the human body can no longer process. They break down rough fiber and other material, then leave partially processed metabolites for the host to absorb. This is not one-way parasitism, but a continuous cycle of “human supply—microbial processing—human recovery.”
寒朵’s own research focuses on how external chemicals reshape the microbiome. Bacteria compete, but also secrete substances that help specific neighbors grow; chemicals first act on different individuals, then reorganize the community, with long-term consequences potentially flowing through to human health.

2. The Same Food Can Yield Different Calories to Different Microbiomes

Yushan asked why 2 people can eat the same food while 1 gains weight more easily. 寒朵’s mechanism is that the body absorbs the readily digestible portion first, while the microbiome processes the remaining fiber; different microbial communities extract different amounts of calories, produce different molecules, and leave different metabolites for the host.
One major research focus is short-chain fatty acids. 寒朵 cited lactic acid and butyrate: Bifidobacteria can produce large amounts of lactic acid, while the popular Clostridium butyricum efficiently produces butyrate. Yet among the hundreds or thousands of gut bacterial species, only a small number are well understood; a single sequencing result cannot calculate a person’s XYZ output.
寒朵 said a substantial share of short-chain fatty acids can raise basal metabolic rate, improve muscle energy use, and increase calorie expenditure. Elite athletes often have stronger fatty acid metabolism, with their muscles relying more heavily on these molecules for energy; ordinary people may rely more on glucose.
The boundary worth preserving is that the link between short-chain fatty acids and improved metabolism has relatively clear support, while lower blood sugar and greater insulin sensitivity remain substantially observational. They travel with health; that does not prove every step of the causal chain.

3. The Gut Microbiome Blocks Pathogens and Trains the Immune System

C. difficile shows that carrying a bacterium is not the same as having a disease: 1 or 2 out of 3 people may carry it and coexist with it quietly for life. Only when the existing microbiome loses its suppressive power can C. difficile expand, causing inflammation, diarrhea, and more severe symptoms. The healthy microbiome itself functions as an immune barrier.
Large numbers of immune cells sit around the gut epithelium because eating means daily exposure to foreign bacteria and potential pathogens. 寒朵 described the microbiome’s constant interaction with immune cells as a learning process through which the immune system distinguishes “normal, healthy bacteria” from genuinely harmful targets.
Babies born vaginally initially encounter birth-canal microbes, primarily lactic-acid bacteria. Babies born by C-section encounter bacteria from hospital air, skin, and other environmental sources, making their early composition more random. Evidence suggests their long-term probability of immune problems may be slightly higher, but this is not a deterministic conclusion that every C-section child will become ill.
Some hospitals and doctors recommend specific probiotics for C-section babies in hopes of making their early microbiomes more like those of vaginally delivered babies. Because immune diseases themselves are uncommon, 寒朵 stressed that the effects require long-term observation, rather than turning early correlations into a validated lifelong intervention.

4. The Gut–Brain Axis Is Running Ahead of Causal Evidence

寒朵 ranked the evidence clearly: randomized double-blind controlled trials sit at the top, followed by non-blinded human interventions; population correlations easily mix in variables such as exercise. Animal studies can control conditions, but “a mouse is still not a human,” so she placed their credibility roughly alongside correlational research. In-vitro studies are further removed from human biology.
Mental-health evidence currently sits mostly at the lower end of that hierarchy. Certain microbiomes may cause anxiety- or depression-like behavior in mice; many bacteria produce serotonin in test tubes, and some antidepressant and anti-anxiety drugs affect bacterial growth. This can generate hypotheses, but not the claim that the drugs work primarily through the gut.
寒朵’s complete validation chain would be: first confirm that psychiatric drugs alter the human microbiome; then test whether serotonin and other molecules change in the gut; next establish that they enter the bloodstream; and finally determine whether directly administering external serotonin or a similar intervention can produce a comparable reduction in symptoms. Only when every link connects would there be a closed mechanism.
Yushan’s counterpoint was that drugs may also act on the liver, brain, or other organs, with microbiome changes representing just 1 of many effects. 寒朵 suggested comparing drug-sensitive and drug-insensitive microbiomes: human genomes are more than 99.9% similar, while any 2 people’s microbiomes are often only 10%—20% similar, creating a natural basis for comparison.

5. Healthy Microbiomes Have No Standard Formula—Only 3 Observable Functions

寒朵’s joke was: “Unhealthy guts are all alike, but every healthy gut is healthy in its own way.” Research is better at identifying what the microbiome loses in disease and then inferring health; it cannot produce a universal list of species and ratios.
The first signal is diversity: a healthy person typically has more than 100 detectable bacterial species, while disease often comes with fewer. The second is metabolic capacity, especially the ability to digest prebiotics and produce substantial short-chain fatty acids; this capacity declines in many patients.
The third is stability. Eating 1 extra piece of cake at lunch or 1 less bite of vegetables creates a small fluctuation, but a healthy microbiome remains relatively stable over a 2-week to 1-month horizon and can absorb a degree of external disruption. A fragile microbiome may undergo larger changes and struggle to return to its starting point.

6. Consumer Microbiome Tests Can Be Accurate Without Guiding Action

寒朵 breaks “reliable” into 2 questions: can the test accurately identify bacteria and their relative abundance in a stool sample, and can it produce a health decision? The first is mature technology. Given reagent costs and retail prices, she believes seemingly legitimate companies are usually sequencing real samples rather than returning random results.
The answer to the second question is colder: science still does not know what the healthiest microbiome looks like. For someone without disease or symptoms, testing mostly proves that “my microbiome is also different, but healthy,” offering little basis for further optimization.
Clinical testing is different. Inflammatory bowel disease, C. difficile infection, and other conditions already have established microbiome-testing pathways; when recommended by a doctor, the result can support diagnosis and affect treatment, giving the test clear decision value.
Price differences may reflect sequencing depth. 3—5 dominant species can account for 80%—90% of the total population, while some important species are scarce. Clostridium butyricum is often below 1%, or even below 0.1%; deeper sequencing can find these low-abundance members, but seeing more does not automatically tell you how to change the system.

7. Short-Term Fasting Disrupts the Microbiome but Usually Does Not Reset It

For a 2- to 3-day fast, 寒朵 offered a deliberately loose but intuitive summary: “As long as you haven’t starved to death, your bacteria probably haven’t either” (“只要你还没饿死，你的细菌应该也没饿死”). A healthy person’s microbiome may change temporarily, but after several days of resumed eating it usually moves close to its pre-fast state, with no evidence of lasting restructuring.
16+8 intermittent fasting broadly overlaps with the circadian cycle, while the microbiome itself also fluctuates over the day. It is therefore difficult to isolate any additional microbiome effect from this eating pattern. The program did not reject fasting’s other effects; it said only that the microbiome evidence is insufficient.

8. Probiotics Can Work Short-Term but Rarely Engraft Automatically

After taking probiotics for 1 month, Bifidobacteria and other strains may increase in the gut and produce more short-chain fatty acids. The problem is that many external strains disappear within days of stopping, and the community returns to its original state. Advertising that probiotics “enter the body and work” may not be wrong; what it leaves out is that the effect may last only while the product is being taken.
寒朵 used ecology to explain failed engraftment: a tropical rainforest and an arid grassland can both be stable and diverse, but move a giant rainforest tree to the grassland and, without rain, symbiotic birds, and surrounding plants, repeated transplantation will keep it alive for only a few days. “Poor adaptation” is not a sign that the strain is insufficiently advanced; the niche simply does not exist.
To make the tree grow long-term, you need to “turn the grassland into a tropical rainforest”—that is, continuously change diet and lifestyle. Yushan pointed out that the grassland itself may not be unhealthy. 寒朵 agreed: the goal is not to turn everyone into the same microbiome, but to maintain an ecosystem that coexists peacefully with its host, processes fiber, and produces useful metabolites.
Taking probiotics every day is therefore perfectly feasible from the body’s perspective; economically, it depends on the individual. “From the standpoint of your wallet, that is a very personal question”—an accurate summary of probiotics’ recurring-subscription economics for healthy consumers.

9. Prebiotics and Fermented Foods Are Closer to Ecological Management Than Single Strains

Prebiotics are essentially fiber that humans cannot digest but gut bacteria can use. Eating more high-fiber food supplies fuel to the existing microbiome, allowing it to grow and produce short-chain fatty acids. Compared with directly inserting external bacteria, this is further upstream and more likely to persist.
Fermented foods sit between probiotics and prebiotics. Yogurt and similar foods may carry live bacteria; production may also involve prebiotics or high-fiber ingredients, while the food already contains short-chain fatty acids generated by fermentation. The body may therefore receive bacteria, bacterial food, and bacterial metabolites at the same time.
Yogurt also provides protein, calcium, and other conventional nutritional value. 寒朵 would personally rather eat yogurt than take a capsule, but she means normal food portions—not “2 jin of yogurt a day.” Sugar content cannot be hidden behind a “probiotic” label.

10. 6 Weeks and 2 Fist-Sized Servings a Day Can Drive Measurable Change

寒朵’s lab divided participants into 3 groups: maintain the original diet, add fermented foods, or add high-fiber foods, with an intervention lasting roughly 6 weeks. Both intervention groups outperformed the control group on microbiome and multiple health measures, supporting 2 dietary paths rather than a single “superfood.”
Participants first recorded their existing diets, then added 2 servings a day from the assigned category. Each serving was about the size of an adult fist—perhaps a small cup of yogurt or a small bowl of celery. Participants could choose the specific foods and timing, trading experimental control for higher adherence over 6 weeks.
The design also preserved real-world noise: some people failed to stick with it, and reported intake may not have matched actual intake. Researchers could serve everyone identical meals and lock them in a lab for 6 weeks for greater rigor, but then “nobody would come do the experiment.” The results must be understood in light of that trade-off.
Measurable benefits included improved microbiome diversity and short-chain fatty acid production, along with lower inflammatory markers in the body—the decline was particularly pronounced in the fermented-food group. After returning to free eating, many participants still consumed more of the relevant foods than before, suggesting that the 6-week intervention may have created a new habit.

11. The Best Daily Plan Remains Simple: Diverse Plants, Moderate Fermented Foods

寒朵 does not take probiotics daily, viewing them as useful in special situations but low-return for most healthy people. When too busy to eat fruits and vegetables properly, she occasionally takes prebiotics; otherwise she prioritizes fruit, vegetables, whole grains, and beans because “I don’t know what the bacteria in my gut actually like, so I try to eat a varied diet.”
A prebiotic powder or capsule made from 1 or a few plant sources may support only specific bacteria, while a natural diet supplies a more complex mix of fibers. Fermented foods require an additional check on sugar and salt: yogurt is often sweetened, and kimchi is high in salt. Eat them in moderation, while fruits and vegetables can generally be consumed more generously within personal tolerance.
泓君 offered the counterexample that dairy may promote skin inflammation. 寒朵 acknowledged evidence linking high intake to localized skin inflammation, but said it is unclear how much “high” means. The conclusion is neither to consume more nor to eliminate dairy across the board, but to control the dose and watch one’s own response.
A sudden switch from processed food to large amounts of fiber mainly overwhelms the person, not the bacteria: producing short-chain fatty acids also produces gas, leading to flatulence and bloating. Increase intake gradually if necessary; if you can tolerate it without feeling overly full, 寒朵 sees no clear downside to increasing vegetables quickly.

12. Exercise Has Stronger Evidence Than Sleep; Households and Pets Also Exchange Microbes

Exercise has interventional support. When normally inactive people sustain moderate-intensity training, their microbiomes shift toward stronger short-chain fatty acid production, and some inflammatory markers decline. Observations that athletes have healthier microbiomes point in the same direction but cannot independently prove causality.
Sleep evidence is weaker, coming mainly from correlation analyses and mouse studies. Short-term sleep deprivation can temporarily reduce microbiome diversity, which usually recovers after sleep is restored. Whether chronic sleep deprivation causes sustained microbiome changes remains an important question without strong human research.
People who live together have microbiomes that become “more similar,” potentially because they share spaces and microbes, or because their diets and schedules converge. In households with pets, children show greater similarity to the pets than adults do, and children also have a lower probability of immune problems. The leading hypothesis is that richer exposure trains the immune system, but this remains a statistical association.

13. Probiotics After Antibiotics Can Stop Diarrhea but Not Accelerate Long-Term Recovery

Antibiotics kill large numbers of gut bacteria and temporarily alter the environment, making diarrhea a common side effect. Evidence supports taking probiotics during antibiotic treatment to reduce diarrhea and protect the gut, but this does not mean they restore the original ecosystem faster.
Over longer follow-up, additional probiotics do not appear to accelerate recovery. The good news is that most healthy people, whether they take probiotics or not, return close to their pre-treatment microbiome within several months to half a year. If an infection genuinely requires antibiotics, “you still have to take the drug.”
The first product check is whether the bacteria are alive when consumed. 寒朵 generally trusts refrigerated preparations more because bacteria are more likely to remain viable; room-temperature and freeze-dried technologies may also work, but consumers should check the labeled live count, commonly around 10^9—10^10 per capsule. All else equal, “higher is better.”
Strains matter too. Bifidobacteria are traditional probiotics; Clostridium butyricum is popular because it produces butyrate, but is harder to keep viable. If other conditions are equal, 寒朵 would favor Clostridium butyricum. She also likes combination products: Seed Health’s flagship product reportedly contains around 20 probiotic species, so “there’s bound to be 1 that suits my gut better.”

14. The Commercialization Threshold for Probiotics Is Safety, Not Efficacy

The probiotic industry is not new. 寒朵 joked that “America’s largest probiotic company may be Danone or a dairy company.” Bifidobacteria and similar strains have long been used in yogurt and cheese fermentation, with extensive safety histories and little practical scope for exclusive patents, so they generally lack strong intellectual-property barriers.
Manufacturers may conduct safety studies and patent new strains. Regulatory and development costs remain far below Phase I, II, and III drug trials; supplements require no prescription and address a broader pool of potential consumers, allowing patent and validation costs to be spread across a larger market.
The US FDA classifies ordinary probiotics as dietary supplements. They do not have to prove efficacy, only safety, and packaging is restricted from making direct health claims. 寒朵 therefore puts safety first and places more trust in major brands; whether a product “works” is something regulation itself does not answer for consumers.

15. Fecal Microbiota Transplantation Is One of the Few Microbiome Therapies with a Strong Clinical Loop

Recurrent C. difficile infection is the most mature case in microbiome medicine: the patient’s original ecosystem can no longer suppress the pathogen, while a healthy donor’s complete microbiome restores competition and the immune barrier. For patients without other contraindications, FMT is already a recommended first-line treatment.
寒朵 recalls cure rates of roughly 80%—90%, with 1 transplant enough for most patients. After success, sequencing shows greater microbiome diversity, while C. difficile levels fall or become undetectable, creating a far clearer endpoint than broad wellness products offer.
For some cases of inflammatory bowel disease, FMT can be a second-line option after other treatments fail. The boundary remains limited to digestive diseases with relatively clear mechanisms and stronger confidence; it cannot be directly extended to every condition merely associated with the microbiome.

16. High Cure Rates Come from an Aggressive “Clear and Recolonize” Procedure

Donors are typically young, of normal weight, and free of other diseases because their microbiomes may carry metabolic or disease risks. Stool is collected in a specialized container, refrigerated immediately, and delivered to the hospital as quickly as possible. The core principle is to do as little as possible to preserve live bacteria.
Hospitals mix the sample with saline into a uniform paste. Part is tested for pathogens, parasites, and other hazards, while the rest remains refrigerated. Once it passes screening, 1 sample is typically enough for 3—5 patients.
To create an ecological niche for the new microbiome, patients may first receive a combination of powerful antibiotics, followed by a physical enema to clear the original community. The transplant is not simply swallowed: it is delivered through a tube from the mouth to the intestine, or through a colonoscopy-like procedure, to bypass stomach acid and improve engraftment.
The process is invasive, but matches the severity of C. difficile infection itself. The long-term direction is sustained-release capsules and encapsulation technology that can pass safely through the stomach and release in the intestine. 寒朵’s view is that once a product can be standardized at scale, oral delivery is “basically an inevitable trend” (“基本上是一个必然的趋势”).

17. Defined Microbiomes and Engineered Bacteria Are Turning Stool into Standardized Drugs

寒朵 mentioned 2023’s “Voust”: it consists of specific bacteria and cannot directly treat an active C. difficile infection, but can prevent recurrence. Another company she recalled from Australia was developing a fixed multi-strain combination to replace whole stool and was in Phase I at the time.
The next step is not to restore the original microbiome, but to give it new functions. 寒朵 cited a startup that had advanced to Phase II by introducing oxalate-metabolizing bacteria, with the aim of lowering oxalate levels in the body and reducing acute attacks in gout patients. The key question remains whether the bacteria can engraft consistently and generate enough metabolic activity.
Another program for phenylketonuria added a specific metabolic enzyme to bacteria so the microbiome could process a substance that patients cannot properly metabolize themselves. The project appeared to encounter problems after reaching Phase III. 寒朵’s assessment retained the practical constraint: the direction is promising, but many projects stall at different clinical stages because of details, funding, or resources.

18. Weight Loss, Autism, and Depression Still Cannot Move from Correlation to Treatment

Claims about using FMT for weight loss, autism, or psychiatric disease are “highly exaggerated and designed to grab attention.” These fields have not clarified how strong the microbiome–symptom relationship is, or whether the microbiome is a cause or merely a parallel change. Treatment is therefore “not even ready to be discussed,” and the success of FMT for C. difficile cannot endorse other indications.
Follow-up studies have observed obese patients losing weight and moving closer to a healthy range after receiving microbiomes from healthy donors. There are also isolated reports of improvements in depression and anxiety. 寒朵 does not call these targeted studies: donors were already screened for healthy weight, while patient numbers and other confounding factors were insufficient to prove a direct microbiome effect.
On whether depression is fundamentally a brain disease or a gut disease, 寒朵 believes the evidence for the brain is stronger. Neurotransmitters can enter the brain and act there, and psychology has a relatively complete imbalance-based mechanism. The gut microbiome currently “adds one more link to this complex system,” rather than overturning the existing theory.
Treating mental illness with probiotics would require proving that every downstream response in the body was actually caused by the ingested bacteria, and that the chain ultimately produces a therapeutic effect. 寒朵’s conclusion: “There are still far too many gaps in this chain” (“这个链条现在的缺口还实在是太多了”).

Vol.172 Interview with Qunhe Technology’s Huang Xiaohuang: 14 Years as a Technical Entrepreneur Mechanistic Interpretability: Philosophy, Practice & Progress with Goodfire's Daniel & Tom