For generations, chemists have written reactions as simple equations: A plus B gives you C. Clean, efficient, and - frankly - not quite the whole story. Sure, we’ve known that byproducts often form, but they’re usually treated as annoyances (unwanted leftovers from a process meant to produce one specific thing).
But what if that mindset is too narrow? What if reactions - even ones we’ve known about for over a century - are actually part of much richer, more flexible networks? And what if, with the right tools, we could nudge those networks to produce entirely different outcomes using the same ingredients?
That’s exactly what researchers at the Institute for Basic Science (IBS) in Ulsan, South Korea, have been exploring.
Led by Professor Bartosz A. Grzybowski, the team at the Center for Robotized and Algorithmic Synthesis (CARS) has developed a low-cost robotic system that does something surprisingly powerful: it maps out how chemical reactions behave under a wide range of conditions. And in doing so, it reveals new products, new pathways, and a new way of thinking about chemical synthesis.
Their findings, published in Nature, challenge the way we’ve traditionally looked at chemical reactions. Instead of static formulas, the CARS team sees dynamic networks, systems that can shift and adapt depending on things like temperature and concentration, not unlike how biological systems regulate themselves through enzymes and feedback loops. Under different conditions, the same reactants don’t just behave a little differently - they can lead to completely different products.
So where did this idea start? With curiosity. The team wanted to know: if we really explored a reaction under lots of different conditions, would we always get the same product? Or might one of those so-called byproducts actually take over?
Turns out, this wasn’t something chemists had fully tested before. Not because they didn’t want to - but because doing it manually would be a monumental task. Just varying three parameters - say, concentrations of A and B, and temperature - across ten values each means running 1000 separate experiments. That’s a lot of time, money, and effort. And that’s for just one reaction.
Enter the robots. The CARS team built an affordable platform that could not only run thousands of reactions but also analyze them quickly without relying on expensive tools like NMR or HPLC. Instead, they came up with a clever optical method that lets them estimate product concentrations using fast spectral readings - basically, by "photographing" the reaction. This gave them an edge: up to 1000 reactions analyzed per day.
With that kind of throughput, they started exploring what they call “reaction hyperspaces” (multi-dimensional maps defined by variables like concentration and temperature). And what they found was surprising. Even well-characterized reactions turned out to have hidden regions, where entirely new products formed, sometimes doubling the number of known products for a given system.
Clearly, the old A + B → C equation doesn’t cut it anymore. These aren’t single reactions - they’re networks, filled with possibilities depending on how you set the stage.
But the team didn’t stop at discovering new products. They also wanted to understand how these products connect - how one reaction pathway flows into another. Borrowing an idea from electrical engineering, they looked at how different “inputs” (like concentrations) led to different “outputs” (products), and used that data to reconstruct the hidden wiring - the underlying network of reactions. With the help of chemical AI and kinetic modeling, they were able to map these networks in detail.
And here’s where it gets really interesting: they could now control those networks. By tweaking conditions, they could steer the reaction toward different major products on demand. Think of it like flipping a switch: same ingredients, same flask, different result. That kind of flexibility is common in biology, but until now, it hadn’t been purposefully engineered into chemistry.
The implications are big. In a world where raw materials are costly and sustainability matters more than ever, being able to generate multiple valuable products from the same starting materials is a huge win. The team even identified compounds with applications in pharmaceuticals, dyes, and organic electronics, all emerging from reactions we thought we already understood.
Zooming out, this work opens the door to a new way of doing chemistry. Instead of exploring one reaction at a time, scientists can now scan huge landscapes of reactivity and discover hidden patterns that were simply out of reach before. The CARS team refers to this as exploring the “DarkNet” of chemical space, not because it’s shady, but because it’s been invisible until now.
In short, the team has created a kind of chemical GPS: a tool that helps scientists navigate vast networks of reactions, uncover unexpected pathways, and even switch between them. And they’ve done it with a mix of robotics, clever analytics, and a healthy dose of curiosity.
Journal Reference:
Jai, Y., et al. (2025) Robot-assisted mapping of chemical reaction hyperspaces and networks. Nature. doi.org/10.1038/s41586-025-09490-1