Writing

• 1 Oct 2026

  Deep dive into flow models

  Flow models reach the data distribution with an ODE rather than an SDE. Velocity fields, marginal versus conditional, and why the two flow matching losses have the same minimiser.

• 24 Sep 2026

  What are diffusion models, and how exactly do they stand out from Autoregressive Models?

  Autoregressive models generate left to right and pay for it in both accuracy and cost. What diffusion models do instead, and where each one actually wins.

• 5 Sep 2026

  We tried to beat Pfam for BGC retrieval. Pfam won.

  A Set Transformer over frozen ESM-2 embeddings lost to plain Pfam-domain Jaccard on biosynthetic gene cluster retrieval. What a negative result says about baselines and evaluation.

• 8 Jul 2026

  Fine-tuning ESM-2 for protein subcellular localization: LoRA matches full fine-tuning with 2.4% of the parameters

  Rank-16 LoRA on ESM-2 650M reaches the accuracy of a fully fine-tuned 150M model while training 2.4% of the weights — and a frozen 8M linear probe is already most of the way there.