science-math

Thirteen shows up more often in software work than people expect: 13-week quarters, 13-bit IDs in embedded systems, and even the 13th item in a batch size that crosses a page boundary. When I mentor junior developers, I ask them to estimate a few values without a calculator, and the 13 table is usually the […]

Last year I helped a team automate nutrient dosing for a greenhouse. The software looked fine, but the crop still showed zinc deficiency. The root cause wasn’t a bug in the code; it was a formula mismatch. The config assumed anhydrous ZnSO4, while the supplier shipped the heptahydrate. That molar mass gap skewed every calculation,

Last year I was helping a team debug a battery-powered sensor node that kept browning out whenever the radio transmitted. The scope showed a clean 3.3 V rail, yet the MCU still reset. The culprit was a floating ground reference between the sensor board and the programmer. The team was thinking in terms of a

Last quarter I was reviewing a motion‑planning bug in a warehouse robot. The path looked correct on screen, yet the robot drifted sideways whenever it accelerated. The magnitude of the velocity vector was right, but the sideways motion kept growing. After half an hour of log inspection I found the cause: the vector’s x and

Last year I was reviewing a telemetry pipeline that mapped device speed to power draw. The code assumed a single output for each input speed, so we cached by speed and treated the chart as a function. When I plotted the raw points, the curve looked smooth, but I noticed a thick band. I sketched

Last month I was reviewing a billing pipeline where we split cloud spend across teams. A tiny rounding drift turned a clean 1/3 share into 0.3333…, and after a few passes the totals were off by a few dollars. The fix was not a new algorithm; it was a return to exact fractions. Fractions preserve

Last month I was helping a team coordinate two recurring jobs in a data pipeline. One job rotates log files every 12 minutes, the other aggregates metrics every 18 minutes. We needed a single checkpoint where both tasks finish together so a report could be published with consistent data. Instead of simulating a long timeline,

Last winter I was helping a hardware team debug a battery pack that kept tripping its thermal cutoffs during fast charging. The electrical models looked fine, yet the aluminum case was as warm as a coffee mug. We had sensor logs, fan curves, and control firmware, but we lacked a quick estimate for how much

Last week I was sizing a 3D cache for a voxel renderer. The requirement looked simple: store exactly 1000 blocks in a cube. The side length of that cube determines array sizes, index mapping, and even how I shard data across threads. When the side length is wrong by one, you get holes, overlaps, or

The first time I built a radar-style UI for a robotics dashboard, the sensor stream arrived as angle and distance, not x and y. I could have forced everything into Cartesian coordinates, but the visualization felt like I was translating for no reason. The polar coordinate system matched the hardware, the math, and the way

Last week I was reviewing a laser-cut bracket for a small robotics rig, and the CAD export tagged one panel as an isosceles triangle. The design looked symmetric, yet the perimeter in the bill of materials was off by a few millimeters. That tiny mismatch would have triggered a re-cut and a missed delivery slot.

Last spring I helped a team schedule a day of 12 lightning talks. We could choose any subset, but the order changed the energy of the room, so the question became: how many possible lineups are there for 5 talk slots? That’s not a combination problem; it’s an arrangement problem. The nPr formula gives a