Technologies & Stack
How Muhywyi delivers
Every tool we use has a specific job. This page explains the infrastructure behind our journalistic masterclasses - from delivery platforms to the editing environments our instructors actually work in.
The tools behind
each session
Muhywyi has been building its delivery infrastructure since 2018. The choices made early - favouring stability over novelty - have held up. We run our live masterclasses on a low-latency streaming layer that keeps audio sync tight even on modest broadband connections, because that matters more than visual polish when an instructor is demonstrating sentence-level editing decisions.
Session recordings are stored in a lossless format before being re-encoded at three quality tiers. Participants on slower connections get a watchable experience; those on fast connections get the full resolution of the instructor's screen. Captions are generated automatically and then corrected by hand before any recording goes live in the archive.
-
Adaptive Video Delivery
Three-tier encoding ensures every participant gets a smooth playback experience regardless of connection speed.
-
Hand-corrected Captions
Auto-generated transcripts are reviewed by a human editor before publication. No raw AI output reaches participants.
-
Shared Editing Environment
Instructors and participants work inside the same document layer during live sessions, making feedback immediate and visible.
-
Low-latency Live Stream
Audio and video sync stays within 200ms even at the edges of our distribution network, keeping live Q&A natural.
-
Modular Course Builder
Instructors assemble sessions from reusable modules - video clips, annotated documents, and discussion prompts - without touching code.
Platform decisions that came from actual use
The stack at Muhywyi was not assembled from a wishlist. Each component was added because something else failed in practice. Early sessions used a general-purpose video conferencing tool and the limitations showed up fast - screen-share compression made annotated text unreadable, and the lack of a persistent document layer meant participants had no reference between sessions.
The current setup separates concerns deliberately: streaming handles real-time presence, a document layer handles collaborative annotation, and a separate archive system handles everything that needs to persist. These three layers talk to each other but can be updated independently. When a better captioning tool became available in late 2022, it replaced only the captioning component without touching anything else.
Security follows the same principle. Participant data, session recordings, and instructor materials sit in separate storage environments. Access is scoped - an instructor can see their own sessions and nothing else. A participant can access only courses they are enrolled in. There are no shared admin credentials anywhere in the system.
Platform activity log
Caption review completed for Session 12 - Investigative Structure
New module published: Editing for Clarity - draft to final in four passes
Live session archive encoded at three quality tiers - now accessible to all enrolled participants
Shared editing environment updated - annotation layer now supports inline audio comments
Access scope audit completed - no cross-participant data exposure detected
Streaming layer latency test - average sync delta 180ms across all regions