36 Days Through Their Eyes

Bangladesh's July 2024 uprising is already at risk of being flattened into two things people struggle to connect with: contested political headlines, or dry casualty statistics. For the general public — especially younger people and the diaspora who lived through it via fragmented, blocked-internet social media rather than coherent reporting — it's hard to grasp what the 36 days actually felt like moment to moment, or how differently they landed depending on who you were: a student, a journalist, a rickshaw-puller trying to earn a day's wage through a blockade. This matters right now, not just historically. The International Crimes Tribunal's verdict is still contested, casualty figures still vary by source, and Bangladesh only just held its first post-uprising election in February 2026. In this environment, unclear or unsourced retellings actively risk becoming misinformation, while pure evidence archives — valuable as they are — don't build the emotional understanding of why ordinary people risked everything. There's a gap between historical accuracy and human empathy, and almost nothing sits in both places at once.

36 Days Through Their Eyes is an interactive narrative web app that lets players live through the uprising as one of three ordinary people — a student, a journalist, and a rickshaw-puller — while keeping the actual history completely fixed and unchangeable underneath their choices. The backbone is a granular, day-by-day timeline of 26 real events from June 5 to August 5, 2024, each cross-checked against at least two independent public sources (OHCHR, Reuters, The Daily Star, Al Jazeera, and others) and cited inline. Each of the three characters has their own parallel set of chapters mapped onto that same timeline — so a player choosing the student sees Day 16 (Abu Sayed's killing) through a dormitory window with classmates, while a player choosing the rickshaw-puller experiences the exact same day as a sudden, dangerous drop in fares and safety. Every chapter ends with a real choice, but every path — regardless of what's chosen — converges back on the same documented historical outcome. A "Historical Reality" panel on every scene shows what actually happened, so the fiction never gets to quietly replace the record. Beyond the core narrative, the app has four supporting features built specifically to close the empathy/accuracy gap: Memory Wall — profiles of real, named victims (not composites), each with sourced accounts, distinguishing fictional characters from real people who died. Sources & Methodology — every citation used anywhere in the app, aggregated and searchable in one place, so the "we didn't invent anything" claim is checkable, not just asserted. Human Cost Strip — a quiet, persistent casualty counter that only steps forward at dates where a real cumulative figure was actually reported at the time, rather than faking daily precision. Then & Now epilogue and completion certificate — after finishing a character's path, players get a cited "what's happened since" coda (the Feb 2026 election, the tribunal verdict), and once all three perspectives are completed, a shareable certificate marking that they've now seen the uprising from every angle — worker, witness, and student. Technically, it's a Next.js app with a static JSON data layer driving both the timeline and the per-character chapter engine — no backend, no database, fast and reliable for live judging.

Next.js (App Router) for the frontend and lightweight API routes. React with Tailwind CSS for styling, using CSS custom properties for the design system (colors, typography tokens). Framer Motion for page transitions, scroll-triggered animations, and the timeline's viewport-tracking logic. lucide-react for iconography throughout. Google Fonts — Cormorant Garamond, Libre Baskerville, and Inter — for display, heading, and body typography respectively. Data layer is static JSON (no database) driving the timeline, per-character chapter content, character roster, memory wall, and epilogue content. Deployed on Vercel.

Anthropic's Claude was used throughout development: for structuring and drafting the historical timeline data (with all factual claims independently cross-checked against public sources — OHCHR, Reuters, The Daily Star, Al Jazeera, Wikipedia — and cited inline in the app, not taken from the model's own knowledge); for generating and debugging React/Next.js component code across the timeline, character selection, chapter, ending, and memory wall pages; and for drafting supporting written content including this submission's problem statement and solution description. All historical content was reviewed against cited sources rather than accepted as-is, and all generated code was tested and adjusted in our own development environment.