Thomas J L Mustard Ph.D. · 9 entries
The competitive advantage of scientific software is shifting. When AI can replicate proprietary workflows, the real moat becomes the data underneath them — and the scientists who know how to use it.
ViewTracing the arc from a graduate-school question — can we automate catalyst design? — through a decade of building AutoRW, and into the current age of agentic scientific computing.
ViewWhen even the best individual tools aren't enough. The third part in the series examines why integrated, agentic systems are the next step in the evolution of scientific software.
ViewHow automation and thoughtful UX transformed computational chemistry from a specialist skill into something bench chemists could run — and what the tools that enabled this looked like from the inside.
ViewThe standard go-to-market playbook breaks when the product requires a PhD to evaluate. Four principles for navigating it, drawn from a decade of launching technical platforms for Fortune 50 clients.
ViewAn early examination of where computational chemistry workflows were breaking down — not for lack of compute, but for lack of tools to manage, generate, and analyze data at scale. The gap identified here took a decade to close.
ViewAn inside look at the Agentic AI Co-Researcher built at SandboxAQ — an orchestration layer where intelligent agents conduct research, plan multi-step workflows, execute simulations, and synthesize results to automate the full DMTA cycle at scale.
ViewAn honest update on OINSMILES, an open-source project for the computational chemistry community. Not ready yet — not because the code is broken, but because the science is genuinely hard.
ViewUsing Google AI Studio to convert a decade-old computational chemistry script into a working graphical interface in a single hour — and what that says about how scientists will build tools in the next decade.
View