Agents have burst onto the scene in the last year. Agentic AI refers to AI systems that can pursue a goal, make decisions, take actions, and then adapt based on the results.
Unlike traditional AI models that mostly answer questions or classify information, an agentic system can:
Continue readingFor more than a century, chemists have been trying to squeeze the beautifully messy, quantum-smeared reality of molecules into tidy digital boxes, “formats” such as line notations, connection tables, coordinate files, or even the vaguely hieroglyphic Wiswesser Line Notation. These formats weren’t designed for machine learning; some weren’t even designed for computers. And yet, they’ve become the wedged into the backbones of modern drug discovery, materials design and computational chemistry.
The emergent use of large language models and natural language processing in chemistry posits the immediate question: What does it mean for a molecule to have a “language,” and how should machines speak it?
if molecules are akin to words and sentences, what alphabet and grammatical rules should they follow?
What follows is a tour through the evolving world of chemical languages, why we use them, why our old representations keep breaking our shiny new models, and what might replace them.
Continue readingReinforcement learning (RL) is pretty simple in theory – “take actions, get rewards, increase likelihood of high reward actions”. However, we can quickly runs into subtle problems that don’t show up in standard supervised learning. The aim of this post is to give a gentle, concrete introduction to what RL actually is, why we might want to use it instead of (or alongside) supervised learning, and some of the headaches (figure 1) that come with it: sparse rewards, credit assignment, and reward shaping.
Figure 1: I’d like to help take you from confusion/headache 🙁 (left) to having a least some clarity 🙂 (right) with regard to what reinforcement learning is and where its useful
Rather than starting with Atari or robot arms, we’ll work through a small toy environment: a paddle catching falling balls. It’s simple enough to understand visually, but rich enough to show how different reward designs can lead to completely different behaviours, even when the underlying environment and objective are the same. Along the way, we’ll connect the code to the standard RL formalism (MDPs, returns, policy gradients), so you can see how the equations map onto something you can actually run.
Continue reading
Tiles, tiles, as far as the eye can see. Conquerors on horseback storming into the breach; proud merchant ships cresting ocean waves; pious monks and shepherds tending to their flocks; Christ bearing the cross to Calvary—in intricate tones of blue and white on tin-glazed ceramic tilework. Vedi Napoli e poi muori the Sage of Weimar once wrote—to see Naples and die. But had he been to Lisbon?
The azulejos of the city’s numerous magnificent monasteries are far from the only thing for the weary PhD student to admire. Lisbon has no shortage of imposing bridges and striking towers, historically fraught monuments and charming art galleries. Crumbling old castles and revitalised industrial quarters butt up against the Airbnbs-and-expats district, somewhere between property speculation and the sea. An endearing flock of magellanic penguins paddles away an afternoon in their enclosure at the local aquarium (which is excellent), and an alarming proliferation of custard-based pastries invites one to indulge.
This blog post is supporting my poster at Young Modellers Forum and makes things way easier to see and understand. Underneath each GIF, is the explanation of what you should look for as things denoise throughout the diffusion trajectory. Click the GIFs for higher quality viewing!
Continue readingThe Molecular Graphics and Modelling Society began life as the Molecular Graphics Society. It’s hard to imagine a time without computer graphics, but yes, it existed. The MGS was formed by the pioneers who made molecular graphics commonplace.
In 1994, the MGS organized an Art and Video Show (Goodsell et al., 1995), and I submitted some of my own work. One of the other images — inspired by Magritte‘s “Ceci n’est pas une pipe”, depicts a molecule with a remarkable similarity to a pipe — and to a molecule… It was submitted by Mike Hann (of GSK):
Welcome to a slightly different blog post than usual. Today I am sharing an insight into my life at Keble College, Oxford. I am the Chair of Cheese and Why?, which is a talk series we host in our common room during term. The format is simple: I provide cheese and wine, and a guest speaker provides the “why”—a short, thought-provoking talk to spark discussion for the evening.
To kick off the series, I opened with the question of artificial intelligence replacing human thought. I am sharing my spoken essay below. The aim of a Cheese and Why? talk is to generate questions rather than deliver answers, so I hope you’ll forgive me if what follows doesn’t quite adhere to the rigorous structure of a traditional Oxford humanities essay. For best reading, I recommend a glass of claret and a wedge of Stilton, to recreate the full Oxford common-room experience.
Oxford University has recently announced that its students will receive free access to a professional-level subscription of ChatGPT Education. This decision is more than just a perk, it’s a signal. One of the world’s leading universities is openly acknowledging that generative AI will be central to the academic experience of its students. But what does this mean for learning? For education? For scholarship itself?
To frame this question, it is worth beginning with a macro view: Mary Meeker’s AI Trends Report (2025) argues that AI is accelerating the transformation of knowledge work, pushing tasks once reserved for experts into more automated or semi-automated regimes. In her framing, AI is less a standalone innovation than a “meta-technology” that amplifies other domains.
Continue readingIn the world of computational structural biology you might have heard of diffusion models as the current big thing in generative modelling. Diffusion models are great because primarily they look cool when you visualise the denoising process to generate a protein structure (checkout RFdiffusion Colab notebook), but also because they are state of the art at diverse and designable protein backbone structure generation.
Originally emerging from computer vision, a lot of work has been built up around their application to macromolecules – especially exciting is their harmonious union with geometric deep learning in the case of SE(3) equivariance (see FrameDiff). I don’t know about you but I get particularly excited about geometric deep learning, mostly because it involves objectively dope words like “manifold” and “Riemannian”, better yet “Riemannian manifolds” – woah! (see Bronstein’s geometric deep learning for more fun vocabulary to add to your vernacular- like “geodesic”, Geometric Deep Learning).
But we’re getting side tracked. Diffusion is a square to rectangle case of score-based generative models with the clause that diffusion refers explicitly to the learning of a time-dependent score function that is typically learned via a denoising process. Checkout Jakub Tomczak’s blog for more on diffusion and score-based generative models. Flow matching, although technically different to score-based generative models, also makes use of transformations to gaussian but is generally faster and not constrained to discrete time steps (or even Gaussian priors). So the big question is, how does one flow match?
Continue reading
Researchers from Apple have released SimpleFold, a protein structure prediction model which uses exclusively standard Transformer layers. The results seem to show that SimpleFold is a little less accurate than methods such as AlphaFold2, but much faster and easier to integrate into standard LLM-like workflows. SimpleFold also shows very good scaling performance, in line with other Transformer models like ESM2. So what is powering this seemingly simple development?
Continue reading