gshc2020.com

AI and robotics accelerate search for better gut microbiome therapies

Biomedical engineers at Duke University have demonstrated a method for systematically developing novel, complex combinations of probiotics and prebiotics to more effectively maintain gut health and treat various gastrointestinal diseases.

… %!s()

By strategically designing experiments and robotically automating thousands of parallel trials to fill knowledge gaps and make the model more accurate, the approach can reveal complex interactions among many microbial species, nutritional sources and the surrounding environment.

One challenge in designing probiotics and prebiotics that improve gut health is the substantial variability in diet, organisms present in the gut, medications and other host factors. In a proof-of-concept study, researchers used this method to identify specific bacterial combinations and dietary fibers that together consistently promote the health and function of the human gut regardless of variability in the organisms present. The approach could be a boon for a rapidly growing prebiotic and probiotic industry that has long struggled to demonstrate consistent and predictable results for customers.

"If you take a seed and plant it in soil that is depleted of nutrients and minerals, it might not grow very well," said Ophelia Venturelli, associate professor of biomedical engineering at Duke. "The same thing applies to putting microbes into a person's gut microbiome. This work aims to engineer both the environment and the bacteria to improve human health."

The results appear online July 27 in the journal Nature Chemical Biology.

A crowded and variable ecosystem

The human gut is a highly complicated environment that hosts thousands of different species of microorganisms. The number of microbial cells in the human gut is roughly equal to the number of human cells, and they weigh up to nearly half a pound (225 grams).

To further complicate matters, the makeup of species and their relative ratios vary greatly from region to region and even from person to person, depending on a wide range of factors like diet, lifestyle and environmental exposures. Although bacterial groups in the human gut microbiome have much in common, different people harbor different bacterial species, creating substantial variability that can make treating disruptions in their functions difficult.

And disruptions can be serious. Microbes in the gut prevent unwanted pathogens from taking hold while breaking down food like dietary fiber and complex carbohydrates that the stomach can't. This process provides nutrients essential for the health of gastrointestinal cells. Bacteria also produce other molecules that shape immune system development, energy balance, cardiometabolic health and even brain and neurological activities.

Searching trillions of combinations

To address these issues, introducing healthy microbial communities and making sure they have the nutrients they need to thrive has become a major industry, currently valued at about $130 billion worldwide. Demonstrating consistent results, however, has remained a challenge.

"Probiotics alone might not be able to colonize a person's gut long enough to do anything helpful because there's so much living there already," Venturelli said. "There is substantial uncertainty about whether these current methods for restoring gut microbiome health are effective at all, since outcomes are still largely unpredictable."

Venturelli's new approach is meant to bring more predictability to the whole enterprise.

In the new paper, Venturelli and her colleagues demonstrate a method for intelligently and systematically exploring this large design space by closing the loop between experiments and computational models. In benchtop experiments, her team forced various combinations of 15 different species of gut microbes to live and eat together. Their menu consisted of six types of dietary fibers known to affect the production of butyrate, a short-chain fatty acid critical to gut health.

Even with just 21 different variables on the table, the potential combinations of microbes and their diets easily numbered well into the trillions. Faced with such an intractable problem, the researchers turned to machine learning, computer models, automated experiments and active learning—also known as Bayesian optimization.

Bayesian optimization is an approach that designs experiments to simultaneously fill gaps in the computer model and achieve specific goals. Armed with autonomous robotic experimental systems, the team was able to plow through five batches of high-throughput experiments—up to 390 conditions at the same time—to build their knowledge of interactions and outcomes.

"We found a lot of unexpected effects from interactions between the microbes and the fibers," Venturelli said. "These effects are very complicated and couldn't have been predicted before we ran the experiments."

A butyrate-producing trio emerges

The results revealed that the dietary fiber inulin, along with two inulin-hungry microbial species (Bacteroides uniformis and Anaerostipes caccae), interacted with a third bacterial species (Prevotella copri) to reliably produce the desired butyrate. What's more, this combination remained effective regardless of what other species and factors were introduced.

The researchers are now working to see whether this combination can effectively treat a mouse model of inflammatory bowel disease, with promising early results. Beyond this specific combination, the researchers believe their approach could be used to discover many more promising combinations for a wide range of gastrointestinal disorders.

"There are companies combining bacteria and dietary fibers in their gut health products already, but they are not yet trying to identify and match the right synergies to the right problem," Venturelli said. "I believe using this process would help create more tailored—and more effective—solutions to a wide range of gastrointestinal issues."

Publication details

Designing fiber–gut microbiome interactions with active learning, Nature Chemical Biology (2026). DOI: 10.1038/s41589-026-02272-4

Who's behind this story?

Swati Mestri

Swati Mestri holds a bachelor's degree in Electronics Engineering and has worked as a content editor since 2019. She has experience editing research documents across technology, health care, and materials science, and has a particular interest in technology and space. Full profile →

Robert Egan

Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →

Citation: AI and robotics accelerate search for better gut microbiome therapies (2026, July 27) retrieved 27 July 2026 from https://phys.org/news/2026-07-ai-robotics-gut-microbiome-therapies.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only. %!s()

My husband and I planned completely different activities on our trip. No one had to compromise, and it brought us closer.

Travel has a way of exposing the tiny negotiations that make up a relationship. Where should we eat? Beach or mountains? Spa or hike? Early morning or sleep in?

My husband and I have similar travel styles in a lot of ways. We both love wandering through new towns and staying somewhere that feels special rather than generic. But when it comes to how we actually want to spend our days, we're complete opposites.

He's happiest when he's climbing a mountain or squeezing in one more hike before sunset. I can appreciate a beautiful trail, but my idea of a perfect vacation usually includes a good book, a spa appointment, and no place I have to be.

For years, I assumed one of us had to compromise. Then, during a recent trip to Lake Tahoe, we tried something different.

Instead of trying to find activities we'd both enjoy equally, we intentionally planned parts of the trip apart. It ended up being one of the best decisions we made.

We stopped treating every vacation activity like a team sport

We booked a room at Desolation Hotel, an eco-conscious micro-resort tucked into the heart of South Lake Tahoe.

Despite being just steps from restaurants, shops, and the Heavenly Gondola, it felt like a quiet retreat, making it easy to slow down and enjoy the soaking tub on the private balcony.

One afternoon, instead of debating whether to book a massage or tackle another trail, we each chose exactly what we desired. I caught a ride from the hotel and headed to Tahoe Beach Club for a massage while my husband set off for an afternoon outdoors.

For a few hours, we each had the vacation we actually wanted.

When we reunited later that afternoon, something unexpected happened: We had stories to tell each other. Instead of saying, "Remember when we..." all day, we swapped experiences.

He showed me photos of sweeping lake views and told me about spotting wildlife, while I described the tranquil treatment room, the sound of the water outside, and how relaxed I felt afterward.

A little time apart gave us something new to bring back into the relationship.

Spending time apart made our time together feel more intentional

The next morning, I woke up feeling surprisingly energized after my itinerary of solitude.

Instead of lingering in bed, I headed to the hotel's completely aesthetic gym for a workout (and even threw in a sauna sesh) while my husband enjoyed a slower start to the day, sleeping in.

It wasn't a grand gesture, but it reinforced that we didn't have to follow the same schedule to enjoy the same vacation.

Later that morning, we met up and hiked to Eagle Lake, a turquoise alpine oasis tucked into the Desolation Wilderness. The moderate trail climbed past Eagle Falls before opening up to still water framed by granite peaks and towering pines. It was exactly the kind of hike we both could appreciate: enough adventure for my husband, plenty of beautiful scenery for me.

Come evening, we boarded a sunset boat tour of Emerald Bay. Cruising across the impossibly blue water as the sun dipped lower in the sky, we admired the cliffs and forested shorelines.

After carving out space for our individual interests, sharing one of Lake Tahoe's most iconic experiences together felt even more meaningful.

The rest of the trip felt lighter. We wandered downtown, ducking into local shops and taking our time without an agenda. Dinner at the hotel's on-site Maggie's restaurant turned into one of those meals that stretches well past dessert because you're simply enjoying the conversation.

On our final morning, we had brunch at Edgewood Tahoe Resort and wandered down to the resort's waterfront beach. Sitting by the lake and watching paddleboarders glide across the crystal-clear water, I realized neither of us was wishing we'd spent the previous days differently. We were simply enjoying being there together.

In everyday life, couples naturally separate for work, hobbies, workouts, or time with friends. Yet on vacation, there's often an unspoken expectation that every experience should be shared.

That pressure can turn even small decisions into unnecessary compromises. By giving each other permission to spend a few hours apart, we removed that tension entirely.

We'll be planning future trips this way

Our time apart didn't make the vacation feel less romantic. If anything, it made it better.

We both returned feeling fulfilled rather than resentful that we'd sacrificed something we'd been excited about, which made the time we spent together afterward feel more present and connected.

I still want shared experiences when we travel. Some of my favorite memories are the ones we've created side by side. But I've stopped believing that togetherness has to look like doing everything together.

Sometimes the healthiest compromise isn't meeting in the middle at all. It's trusting your partner enough to let them spend an afternoon — or even a morning — exactly where they want to be, while you do the same.

In the end, we didn't come home talking about the hours we spent apart. We came home talking about how surprisingly close we felt because of them.

Read next

Lexi Marshall is a freelance writer and editor covering health, wellness and community. Her reporting has taken her to places like Morocco, Croatia, Belize, Hawaii and Mexico. Outside of work, she is an avid reader and certified yoga teacher based in Colorado.

• Essay
• Travel

Nintendo’s ‘Splatoon Raiders’ Just Set A Metacritic Score Record

Splatoon Raiders

Nintendo

It has not exactly been the most publicized game in the world, but Splatoon Raiders is delighting Switch players at the moment, a break from its usual genre and turning it into a surprisingly effective looter shooter. A Destiny 2 replacement after that game died. That may not actually be a joke.

Why Kimi K3 Signals A Convergence Toward Open-Weight Models

SHANGHAI, CHINA - JULY 20 2026: People visit the booth of Kimi, an LLM developed by the Chinese startup Moonshot, during the World AI Conference in Shanghai, China, Monday, July 20, 2026. (Photo credit should read LONG WEI/ Feature China/Future Publishing via Getty Images)

… %!s()

Future Publishing via Getty Images

This month, Moonshot AI released Kimi K3, an open-weight model with 2.8 trillion parameters. It demonstrates how open-weight systems are competing with closed models on architectural design, enterprise adoption, and operational cost. The data from this release suggests a shift in the competitive balance of the AI industry, grounded in both technical capability and structural distribution economics.

Parallel Orchestration vs. Sequential Execution

The technical distinction in Kimi K3 originates in the agentic framework introduced in Kimi K2.5, termed “Agent Swarm.” The Kimi K2.5 technical paper outlines a core limitation of existing systems that rely on “sequential execution of tool calls”: “As agentic workloads grow in scope and heterogeneity—e.g., building a complex project that involves massive-scale research, design, and development—the sequential paradigm becomes increasingly inefficient.”

To address this, the authors propose Agent Swarm and Parallel-Agent Reinforcement Learning. The framework equips the orchestrator with interfaces for “sub-agent” creation and task delegation. Agent Swarm reduces inference latency by up to 4.5× in wide-search scenarios. Kimi K3 inherits this architecture, executing tasks via up to 300 sub-agents, managing over 4,000 tool calls per task, and running 4.5x faster than single-agent sequential execution. This transforms task completion time from a function of sequential steps to a function of parallel processing capacity.

Native Multimodal Integration

Kimi K3 incorporates the vision architecture of Kimi K2.5, which consists of a three-dimensional native-resolution vision encoder (MoonViT-3D), an MLP projector, and the Kimi K2 MoE language model. This vision component is integrated into the pretraining phase rather than added as a post-hoc adapter.

DeepSeek’s current open-source offerings remain text-centered. The native vision capability in Kimi K3 enables visual-to-code generation and supports real-world software engineering tasks that require interpreting screenshots and visual layouts. For technology startups and developers building applications that require joint text and image editing and designing, this eliminates the need to integrate separate vision models or rely on closed multimodal APIs.

MORE FOR YOU

Enterprise Adoption in the United States

Adoption data indicates that U.S. companies are integrating Kimi models into production workflows. DoorDash has adopted Kimi for internal use, applying it to lower-level tasks while reserving Anthropic’s Fable for higher-level functions. Coinbase has confirmed internal use of Kimi. Cursor, the coding startup acquired by SpaceX, developed its product using a Kimi model as foundation. Data from OpenRouter, an API marketplace, shows that Chinese AI models now account for approximately 60% of token usage by U.S. companies on the platform.

Nvidia CEO Jensen Huang stated in an Axios interview that U.S. companies “absolutely should be allowed to use Chinese models.” Nvidia’s position, given its role as the primary hardware provider for AI workloads, carries weight in the infrastructure decisions of enterprise customers.

Cost and Parameter Scaling

Kimi K3 is the first open-weight model to reach 2.8 trillion parameters. It uses a Mixture-of-Experts architecture with 896 experts, activating 16 per inference to maintain computational efficiency. The model also implements Kimi Delta Attention (KDA), a hybrid linear attention mechanism that provides a 6.3× decoding speedup in million-token contexts.

The cost comparisons against closed models are measurable. On the BrowseComp benchmark—which evaluates real-world research and browsing—Kimi K3 scores 91.2, compared to Claude Fable 5 at 88.0 and GPT-5.6 Sol at 90.4.

In coding benchmarks, the cost differential is large. The rollout cost is $4.65 for Kimi K3 compared to $13.41 for Claude Fable 5. At a full 452-rollout sweep, the total cost is $2,103 versus $6,010. Kimi K3 delivers 2.8× more solved tasks per dollar on this benchmark.

The Structural Necessity of Open Distribution

The competitive standing of large language models is secondary to a broader structural condition governing the diffusion of general-purpose technologies. AI builds an infrastructure-level innovation, analogous to the steam engine for industrial revolution, the microprocessor for personal computers, or the optical fibers for modern telecommunication. The economic and social value of these technologies materialized only when they transitioned from centralized artifacts to distributed assets accessible to the majority of economic actors.

Automobiles did not transform transportation while restricted to a single factory fleet. Televisions did not reshape media consumption as studio-exclusive hardware. Personal computers did not create the software industry when confined to mainframes. In each case, the inflection point occurred when the technology achieved near-total distribution—ownership, modification, and integration into diverse local contexts by a broad population of users, developers, and enterprises.

AI follows the same diffusion curve. Its projected economic impact—estimates range from $17.1 trillion to $25.6 trillion in annual global GDP contribution—depends on its penetration into every sector, device category, and regulatory jurisdiction. This requires deployment at the edge, integration into proprietary data pipelines, fine-tuning for domain-specific vocabularies, and compliance with local data sovereignty laws. These requirements are inherent to the utility function of the technology.

Closed models impose structural friction on this diffusion process. Per-inference pricing creates a variable cost that scales with usage, disincentivizing high-volume or low-margin applications. Opaque weights prevent fine-tuning for specialized verticals such as healthcare diagnostics, manufacturing quality control, or financial risk modeling, where proprietary data cannot be transmitted to external APIs.

Open weights eliminate these frictions. Distribution carries zero marginal cost per copy. Modifications require no external permissions. Deployment can occur on local hardware, air-gapped networks, or specialized chips without per-token accounting. This model aligns with the infrastructural role that AI is expected to occupy: embedded and ubiquitous.

The competitive pressure to achieve maximal adoption compels a convergence toward openness. Closed systems may sustain temporary leads in benchmark performance or specific capability sets. However, the distribution economics favor the paradigm that enables the broadest and deepest integration into the global economy. Over time, this structural advantage erodes the market position of closed models regardless of their technical merit at any given release cycle. The long-term equilibrium of AI infrastructure will favor open-weight architectures, not because of ideological preference, but because the diffusion requirements of the technology itself preclude sustained closure. %!s()