Rebuilding the World's Most Researched Brain Training App — From Flash to Modern Web and Mobile
The Problem
Cogmed is the world's most evidence-based working memory training program — developed at the Karolinska Institute, backed by 120+ peer-reviewed publications, and used by 200,000+ patients across 20+ countries for ADHD, brain injury, and stroke recovery. Under Pearson's ownership, the core training app ran on Adobe Flash and AIR — a technology Adobe deprecated in 2017 and discontinued entirely in 2020. The entire product — interactive cognitive exercises, adaptive difficulty algorithms, cross-platform delivery, and clinical data logging — needed to be rebuilt on modern web and mobile technology without disrupting the active global user base. Trembit came on as the engineering team to carry out this migration.
Why This Migration Was So Hard
This was not a standard Flash-to-HTML5 conversion. Cogmed's training exercises are clinically validated, adaptive cognitive tasks where difficulty adjusts in real time based on patient performance — and clinical decisions depend on the data they produce. The migration had to preserve:
- Clinical equivalence — the rebuilt exercises had to produce the same adaptive behavior and scoring as the Flash originals, because 120+ research studies validated those specific algorithms
- Cross-platform delivery — the Flash/AIR app ran on desktop, Android, and iOS from a single codebase, and the new version needed to work across web browsers, tablets, and smartphones
- Medical device certification — Cogmed is an approved medical device manufacturer, so any change to the training software required maintaining compliance throughout the rebuild
- Active user base — 200,000+ users and clinical providers in 20+ countries needed continuity during the transition
What We Did
Platform Architecture — Flash/AIR to JavaScript/React
- Rebuilt the training application from ActionScript/Flash to JavaScript and React — a fully web-based HTML5 platform that runs in Chrome, Firefox, Edge, and Safari with no installation required
- Migrated the adaptive difficulty algorithms — the core clinical engine that adjusts exercise complexity in real time based on user performance — ensuring behavioral equivalence with the validated Flash version
- Architected for cross-platform delivery from a single modern codebase and designed the migration path to maintain backward compatibility with existing user data, coaching workflows, and clinical reporting
Rebuilding Interactive Cognitive Tasks
- Rebuilt each interactive training exercise in HTML5/JavaScript — replicating the precise timing, visual sequencing, and audio cues that the clinical research was validated against
- Implemented the real-time adaptive engine that continuously adjusts difficulty based on individual performance, applicable across ages (children through adults) and cognitive levels
- Built session logging that feeds the Coaching Center so clinical providers can monitor progress, with a consistent experience across desktop monitors, tablet touchscreens, and smartphone displays
Mobile Applications & Provider Integration
- Developed native iOS and Android applications for the brain-training exercises, giving patients flexibility to train on the device that works best for them
- Integrated with the Coaching Center (provider portal) for patient profiles, training programs, progress dashboards, and clinical reports, and connected to the Learning Center for coach training and certification
- Engineered the infrastructure for global scale — concurrent users across 20+ countries with varying network conditions and device capabilities
Key Results
In Their Words
What really differentiates them is their ability to deeply understand business needs.
Their proactive team gets things done as if it were their own project.
What We Learned
Migrating clinically validated software is not a rewrite — it is a replication
When 120+ research studies depend on specific adaptive behaviors, timing, and scoring, the new implementation must produce identical outcomes. Standard Flash-to-HTML5 conversion tools do not preserve the clinical logic. Every exercise required manual validation against the original to ensure the research remained valid.
Flash-to-HTML5 for interactive applications requires rebuilding, not converting
Flash's rendering model, animation timeline, and event loop are fundamentally different from the browser's DOM and JavaScript event model. Game-like exercises with precise timing, visual sequences, and adaptive difficulty had to be re-engineered from the ground up — automated conversion tools produce broken results for this class of application.
Medical device migration demands parallel validation at every step
You cannot rebuild a certified medical product and validate at the end. Each exercise, each adaptive algorithm, and each data pipeline needed side-by-side comparison with the Flash original throughout development — not as a final QA step, but as an integral part of the build process.
Modernizing Legacy Software?
Running a healthcare or ed-tech product on aging technology that needs modernization? We migrated Cogmed's entire brain-training platform — from Flash/AIR to modern JavaScript/React across web, iOS, and Android — while preserving clinical validation and medical-device certification. Book a 30-minute session. No pitch deck. Just engineering clarity.