Syed's Blog

114 posts · blog home · terminal version · RSS

Designing a Terminal Website (Yes, Like This One)

· design

Why I built my personal site as an interactive terminal, what worked, what didn't, and how to keep a gimmick from becoming a barrier.

UX of Error Messages: 'Something Went Wrong' Is Not a Message

· design

Good error messages say what happened, why, and what to do next, in human words. Before-and-after examples for forms, payments and the dreaded generic failure.

Fonts Have Feelings: A Typography Crash Course

· design

Serif, sans-serif, monospace and the eternal Comic Sans debate. How to choose and pair fonts without a design degree, with a type scale you can copy.

Dark Mode: A Love Letter with Contrast Ratios

· design

Dark mode isn't "invert the colours." Pure black backgrounds, pure white text and saturated accents all cause problems. How to design a dark theme that's actually comfortable.

Whitespace Is Not Empty Space

· design

Clients often want to "fill the gaps." The gaps are working. How spacing creates grouping, hierarchy and calm, with a simple spacing system you can use today.

Why Your Button Is Blue (and Why It Matters)

· design

Blue buttons aren't a lack of imagination. They're decades of learned behaviour, accessibility maths and a little psychology. When to follow the convention and when to break it.

Storytelling with Data: The Chart Is Not the Story

· data-science

A slide full of charts is a data dump. A story has a question, a tension and a recommendation. How to present analysis so people remember it and act on it.

SQL vs Pandas: The Friendly Rivalry

· data-science

Same questions, two languages. When to let the database do the work, when to pull data into Python, and how to stop downloading 40 million rows to compute one average.

Visualisation Sins I Have Committed

· data-science

Truncated axes, 3D pie charts, rainbow colour maps and a legend in a different time zone from the chart. A confession, with fixes.

Simpson's Paradox: When the Data Plays Tricks on You

· data-science

A treatment can look worse overall and better in every single group at the same time. It sounds impossible. It happens constantly.

A/B Testing Without Lying to Your Boss

· data-science

Peeking, stopping early and testing twenty metrics until one looks good are the three classic ways to ship a "winner" that isn't. How to run a test you can defend.

Dashboards Nobody Opens: A Tragedy in Three Acts

· data-science

Months of work, forty charts, eleven filters and three visitors, two of whom were you. How to build dashboards people actually use.