Work

Pinned pictures

Pinned figures

The same figures, pinned to the sheet instead of hung in it. Six come from the case studies with their captions and scopes untouched; three are studies I redrew after pinned references, and they are not results.

Forty-four stacked ink curves of Burgers’ equation, from a smooth sine wave at t = 0 to a sharp shock at x = 0 by t = 1.
Fig. 1Burgers’ equation with ν = 0.01/π, drawn one time slice at a time: the sine wave steepens into a shock at x = 0. Exact solution by the Cole–Hopf transform, computed for this figure. It is the benchmark PINNs are fitted to, not output from the repository’s networks. From Physics-Informed Neural Networks.
Endoscopic frames beside DINOv2 patch-norm and centre-similarity maps through an ESD sequence.
Fig. 2DINOv2 feature diagnostics across an endoscopic submucosal dissection sequence. Selected frames from one ESD sequence. From AIMS: Surgical Phase Detection.
A distance-5 surface code: three X errors light four checks, a matching joins them, and the errors and the correction together form a stabiliser.
Fig. 3A decoder sees only the four lit checks, and its correction differs from the errors yet cancels them. An illustrative error pattern on a d = 5 rotated surface code. The matching is an exhaustive minimum-weight pairing computed for this figure, not Stim or PyMatching output. From Quantum Error Correction.
Eleven ink dots on warm paper joined by a tangled closed path, which untangles step by step into a short loop through every dot.
Fig. 4A tangled tour through eleven dots shortens, one swap at a time, into the shortest loop. The tour is computed by a 2-opt search over these eleven points, not by the paper’s Hopfield network. The final loop is the optimal tour for them. From Hopfield Networks & TSP.
Four PNW dashcam frames with red boxes drawn on labelled potholes, from the paper’s Figure 1.
Fig. 5Sample dashcam frames from the PNW set, with the labelled potholes marked. Figure 1 of Rout et al., arXiv:2401.08588. The boxes are the dataset labels, not model output. From Pothole Detection (arXiv).
Thirteen ink squares flip one at a time while twelve autocorrelation bars shrink, ending at the Barker sequence with every bar within ±1.
Fig. 6Tabu search on the 13-bit LABS problem, one flip a step, from energy 302 down to 6: the Barker sequence. A classical tabu search from a random start, computed for this figure. In the team’s workflow, CUDA-Q samples seed the classical search instead. From iQuHACK 2026 NVIDIA LABS.
Fig. 7Study I — a spiral tunnel, redrawn after a pinned reference. A drawing exercise from geometry. Not a result from any project.
Fig. 8Study II — wave lines, redrawn after a pinned reference. A drawing exercise from geometry. Not a result from any project.
Fig. 9Study III — a circuit tree, redrawn after a pinned reference. A drawing exercise from geometry. Not a result from any project.

Six figures kept with their captions and scopes. Three studies, drawn from geometry.