About this site

A personalised lens on the 1,541-session SXSW London 2026 programme, tuned for Alex at Gocertify. The intent is to cut a six-day programme down to the sessions most likely to be worth your time as a product designer with a strong technical streak.

How the match score works

Each session gets a 0-100 score combining two signals: SXSW's own category tags and a keyword pass over the session title, subtitle and description. The heaviest positive weights go to design - For Designers, Design, the UI/UX designer track, and keyword hits on UI, UX, product design, design systems and interaction design - reflecting a product-design lens. Tech and AI are weighted strongly too (For Techies, Technology & AI, plus AI, LLMs, agentic systems, ML, engineering and identity/verification, which is Gocertify's home turf), and a session that sits at the intersection of AI and design gets an extra bonus. Pure-entertainment tracks - music sets, film screenings, morning Zumba - carry small negative weights.

75-100 — Top pick. Don't miss it.
55-74 — Strong. Probably worth showing up for.
35-54 — Maybe. Useful if you've got a gap.
0-34 — Skip unless something else pulls you in.

What's here

A day-by-day timeline grouped by hour; full-text search across titles, descriptions, venues, and categories; category and venue filters; sort by time or match score; and a saved-only filter so you can revisit your shortlist from any view. The My Picks page detects scheduling conflicts and exports your agenda as a calendar file.

Data & hosting

The schedule is loaded from a slimmed, pre-scored JSON snapshot of the official programme, statically generated at build time. The site is plain Next.js (App Router) and is built to deploy to Vercel with no configuration: connect the repo, set the project root to sxsw-london, and it ships.

Caveats

Keyword scoring is fast but blunt - a description that mentions AI in passing scores the same as one where AI is the topic. Sessions with missing or thin descriptions will look worse than they are. If you spot mis-scoring, the weights in rescore.py are easy to tune - edit them and re-run the script.