https://zanderpolk.com/writing/ Notes on machine learning, space systems, and building software. en https://zanderpolk.com/writing/from-two-orbits-to-one-number/ https://zanderpolk.com/writing/from-two-orbits-to-one-number/ Sun, 09 Aug 2026 10:00:00 +0000 A conjunction lasts about a second, which is what licenses collapsing two orbits in three dimensions into a flat integral. The encounter plane, why along-track uncertainty dominates, and why the crossing angle decides how much of it counts. https://zanderpolk.com/writing/a-geocode-is-not-a-location/ https://zanderpolk.com/writing/a-geocode-is-not-a-location/ Sun, 09 Aug 2026 09:00:00 +0000 A geocoder returns coordinates and a match type, and only one of them survives the pipeline. What the four tiers actually are, why parcel centroids carry bias and street interpolation carries variance, and why point-in-polygon is the operation that will not tolerate either. https://zanderpolk.com/writing/your-validation-set-is-next-door/ https://zanderpolk.com/writing/your-validation-set-is-next-door/ Sat, 08 Aug 2026 09:00:00 +0000 Spatial data breaks the assumption cross-validation rests on. A random split puts held-out points a few hundred metres from training points, so the score describes interpolation rather than prediction. Blocking, how big the blocks should be, and the good objection to all of it. https://zanderpolk.com/writing/a-force-that-does-nothing/ https://zanderpolk.com/writing/a-force-that-does-nothing/ Wed, 05 Aug 2026 09:00:00 +0000 I went looking for why my game dropped from 200 fps to 40 near a boat. I found a rigid body that falls asleep with a force still on it, and stops receiving that force entirely. Two wrong theories died on the way, with interactive figures for the frame budget and the sleep timer. https://zanderpolk.com/writing/observation-estimate-inference/ https://zanderpolk.com/writing/observation-estimate-inference/ Sun, 02 Aug 2026 09:00:00 +0000 Satellite conjunction triage and commercial property intelligence are the same problem: remote observations turned into a number someone acts on. Why every value needs its evidence tier attached. https://zanderpolk.com/writing/json-in-polytoria/ https://zanderpolk.com/writing/json-in-polytoria/ Sun, 02 Aug 2026 09:00:00 +0000 GetAsync returns a string, not a table. Start with the two lines that fix that, then read the engine's own JSON library to find out why null is not nil and why every number passes through a decimal. https://zanderpolk.com/writing/how-buoyancy-works/ https://zanderpolk.com/writing/how-buoyancy-works/ Sat, 01 Aug 2026 09:00:00 +0000 The buoyancy system behind Whatever Floats Your Boat, starting from the smallest version that floats. Then why it bounces forever, why it cannot capsize, and what sampling points actually approximates. With six interactive figures you can push, load, destabilise and capsize. https://zanderpolk.com/writing/building-a-fan-game-without-taking-anything/ https://zanderpolk.com/writing/building-a-fan-game-without-taking-anything/ Sat, 01 Aug 2026 09:00:00 +0000 Whatever Floats Your Boat is a fan remake. What that obligates: separating inspiration from assets, crediting per file with the licence named, and not claiming buoyancy as an invention. https://zanderpolk.com/writing/instanced-rendering-and-its-bugs/ https://zanderpolk.com/writing/instanced-rendering-and-its-bugs/ Fri, 31 Jul 2026 09:00:00 +0000 Draw calls dominate cost, so engines batch. The buffer that makes batching fast is a copy of scene state, which turns correctness into an invalidation problem — and invalidation fails silently. https://zanderpolk.com/writing/shipping-on-a-platform-and-fixing-it/ https://zanderpolk.com/writing/shipping-on-a-platform-and-fixing-it/ Thu, 30 Jul 2026 09:00:00 +0000 Two open pull requests to the Polytoria engine: a transform-inheritance bug you can only find by building something real, and Luau IntelliSense for module requires. What contributing upstream actually costs. https://zanderpolk.com/writing/a-transform-bug-in-a-scene-graph/ https://zanderpolk.com/writing/a-transform-bug-in-a-scene-graph/ Wed, 29 Jul 2026 09:00:00 +0000 Parts inside a folder stopped following their model. The cause was an organisational node missing from the spatial hierarchy, hidden by three independently correct optimisations. https://zanderpolk.com/writing/what-actually-crosses-the-wire/ https://zanderpolk.com/writing/what-actually-crosses-the-wire/ Tue, 28 Jul 2026 09:00:00 +0000 Server authority, a 50ms batch interval, reliable versus unreliable channels, and a per-call interpolation flag. How Polytoria replicates state, and the trade-offs every multiplayer engine makes. https://zanderpolk.com/writing/what-a-script-is-allowed-to-reach/ https://zanderpolk.com/writing/what-a-script-is-allowed-to-reach/ Mon, 27 Jul 2026 09:00:00 +0000 Letting users make HTTP requests from your server hands strangers a machine inside your network. Reading Polytoria's URL gate, and why resolve-then-connect is hard for everyone. https://zanderpolk.com/writing/you-cannot-test-a-timeout-by-waiting/ https://zanderpolk.com/writing/you-cannot-test-a-timeout-by-waiting/ Sun, 26 Jul 2026 09:00:00 +0000 Structured concurrency in Luau: cooperative cancellation, cleanup scopes, and a virtual-time scheduler that turns flaky async integration tests into deterministic unit tests. https://zanderpolk.com/writing/how-a-csharp-object-becomes-a-luau-value/ https://zanderpolk.com/writing/how-a-csharp-object-becomes-a-luau-value/ Sat, 25 Jul 2026 09:00:00 +0000 Luau knows eight types and none of them is Part. The marshalling layer that closes the gap uses userdata, metatable hooks and a weak identity cache — and explains why you should cache property lookups. https://zanderpolk.com/writing/where-early-prediction-fails/ https://zanderpolk.com/writing/where-early-prediction-fails/ Fri, 24 Jul 2026 09:00:00 +0000 BEACON reaches 0.806 PR-AUC one day out and 0.233 at the earliest available observation. Why the collapse is partly measurement, mostly missing information, and what still survives. https://zanderpolk.com/writing/rare-events-are-a-measurement-problem/ https://zanderpolk.com/writing/rare-events-are-a-measurement-problem/ Thu, 23 Jul 2026 09:00:00 +0000 With twelve positives in a test split, the gap between two models is smaller than the noise from reshuffling the split. Why the binding constraint is the split, the baseline, the metric, and the number of repeats. https://zanderpolk.com/writing/accuracy-is-the-wrong-question-for-rare-events/ https://zanderpolk.com/writing/accuracy-is-the-wrong-question-for-rare-events/ Mon, 20 Jul 2026 09:00:00 +0000 A model that flags nothing is 99.4% accurate on satellite conjunction data and completely useless. What I measured instead while building BEACON: PR-AUC, calibration, and ensemble uncertainty as a human-review signal. https://zanderpolk.com/writing/laplace-approximation-and-where-it-breaks/ https://zanderpolk.com/writing/laplace-approximation-and-where-it-breaks/ Fri, 17 Jul 2026 09:00:00 +0000 A posterior over logistic regression weights for the price of one matrix inverse. The derivation, why confidently classified points contribute nothing to parameter uncertainty, and the five ways it fails. https://zanderpolk.com/writing/a-research-interface-should-show-its-seams/ https://zanderpolk.com/writing/a-research-interface-should-show-its-seams/ Wed, 15 Jul 2026 09:00:00 +0000 BEACON's 3D viewer renders uncertainty it cannot physically justify, so it labels the mapping a proxy in the interface, displays which geometry fallback is active, and watermarks its own screenshots. https://zanderpolk.com/writing/bayesian-inspired-is-not-bayesian/ https://zanderpolk.com/writing/bayesian-inspired-is-not-bayesian/ Sat, 11 Jul 2026 09:00:00 +0000 BEACON contains a real posterior and a bootstrap ensemble, and calls them different things. Which sentences each one entitles you to write, and why the cheaper method was enough. https://zanderpolk.com/writing/your-split-is-leaking/ https://zanderpolk.com/writing/your-split-is-leaking/ Tue, 07 Jul 2026 09:00:00 +0000 Splitting grouped data by row instead of by group inflates every metric you report. How event-level splitting works, why label metadata is the same mistake in disguise, and why one split is never a result. https://zanderpolk.com/writing/determinism-is-not-free/ https://zanderpolk.com/writing/determinism-is-not-free/ Tue, 16 Jun 2026 09:00:00 +0000 Send inputs instead of state and thousands of units fit down a modem. The catch is that identical code on two machines does not produce identical floating-point results unless you make it. https://zanderpolk.com/writing/why-large-worlds-shake/ https://zanderpolk.com/writing/why-large-worlds-shake/ Tue, 26 May 2026 09:00:00 +0000 At ten thousand kilometres from the origin, consecutive float positions are a metre apart. Sub-metre motion is not slow, it is unrepresentable — and almost every fix is a better choice of origin. https://zanderpolk.com/writing/measuring-in-the-wrong-coordinate-system/ https://zanderpolk.com/writing/measuring-in-the-wrong-coordinate-system/ Tue, 05 May 2026 09:00:00 +0000 Compute an area in Web Mercator and you get square metres that are not square metres. Why the projection trade-off is a theorem, and why buffering in degrees quietly produces the wrong answer. https://zanderpolk.com/writing/what-a-gnss-fix-actually-costs/ https://zanderpolk.com/writing/what-a-gnss-fix-actually-costs/ Tue, 14 Apr 2026 09:00:00 +0000 A position to a few metres, from satellites 20,000 km away, resting on nanosecond timing. The correction stack between the measurement and the dot on the map, and where each layer fails. https://zanderpolk.com/writing/when-knowing-less-looks-safer/ https://zanderpolk.com/writing/when-knowing-less-looks-safer/ Tue, 24 Mar 2026 09:00:00 +0000 Widening the covariance on a satellite conjunction can lower its computed collision probability. Probability dilution, the non-monotonic curve behind it, and why a summary statistic can invert exactly when it matters. https://zanderpolk.com/writing/kessler-syndrome-is-misquoted/ https://zanderpolk.com/writing/kessler-syndrome-is-misquoted/ Tue, 03 Mar 2026 09:00:00 +0000 The 1978 paper describes a rate equation over decades, not a chain reaction over days. What gets lost in the retelling is the timescale, the altitude dependence, and what it implies for triage. https://zanderpolk.com/writing/reproducing-a-house-price-paper/ https://zanderpolk.com/writing/reproducing-a-house-price-paper/ Tue, 10 Feb 2026 09:00:00 +0000 Reproducing Maloku & Maloku (2024) on the same dataset moved linear regression’s test R² by 0.31. The cause was one preprocessing step, and cross-validation then reversed my own conclusion.