Agent-Rendered Infrastructure: The End of Unnecessary Servers
AI agents that regenerate and deploy artifacts on demand can collapse entire runtime stacks — and the economics are forcing the transition.
Blog
Dictate and Ship
The gap between having an idea and publishing it is now basically zero.
When P-values Hit the Floor: Handling Astronomically Small Statistics in QQ Plots
How fastqq had to confront the limits of floating-point arithmetic when meta-analysis p-values underflow to zero — and the log-space fix that makes extreme signals visible again.
Guest Lecture at Reykjavik University's Executive MBA
I gave a guest lecture in the Data Analytics for Management course at Reykjavik University's Executive MBA programme.
Back to Blogging: AI is Moving Fast
Restarting the blog after a long break, with a new focus on modern AI developments.
Variant in a Taste Receptor Locus Tied to Changes in the Use of Insomnia Medication
A new first-author paper in Biological Psychiatry Global Open Science links a taste receptor variant to switching away from zopiclone — a result that is directly actionable in the clinic.
The Role of BMI in Disease Risk
Our new paper in Nature Communications uses genetics to show that BMI itself — not just factors correlated with it — drives risk for a wide range of diseases.
fastqq R package
How I built and published fastqq, an R package that speeds up quantile-quantile plot generation by 80x using C++.
Speculations on 2021
Speculating on how 2021 might unfold as COVID vaccines roll out globally, looking at herd immunity strategies and distribution.
Simplifying blog
A brief update on moving back to Iceland, starting at deCODE genetics, and simplifying the blog setup.
emplace_back vs push_back
A practical look at the difference between emplace_back and push_back in C++, and when you should actually use each.
Eating Less vs Fad? Diets
Tracking a weight loss journey after a serious knee injury: comparing calorie restriction vs Huel over several months, with data.
Leaving Academia
Reflecting on the decision to leave academia after completing my PhD, and what I'm looking for next.
Part 2: Deep Learning with Closures in R
Part two: completing the MLP in R with backpropagation and stochastic gradient descent, and visualising what the network learns.
Part 1: Deep Learning with Closures in R
Part one of a series on implementing a neural network from scratch in R, using closures and functional programming.
Visualizing Strange Attractors
Rendering animated Clifford strange attractors using C++ and OpenGL, and turning them into looping GIFs.
Got some new books!
A quick post about some new books on R and machine learning I was excited to get into.
First Post!
Introducing the blog: a place to write about things I learn, set up with blogdown and Hugo.