Who needs Fable? Sunday build
TL;DR
GPT Realtime 2 enables instant transcription and translation: The app transcribes speech in real-time and can translate to languages like Spanish, Vietnamese, and Turkish with near-zero latency, running smoothly even on older hardware via simulator.
Wave engineering orchestrates sub-agents for complex tasks: Fernando's custom skill fans out multiple readonly investigation workers in parallel, each with smaller scoped contexts, then aggregates results to avoid poisoning the main context window.
User feedback drove unexpected design pivots: A 95-year-old user with hearing aids wanted larger text to follow conversations live, revealing a use case for people with memory challenges who need to scroll back through transcriptions.
Cursor makes iOS development accessible: Fernando notes that building the iOS app with Cursor and AI assistance was easier than expected, with models writing solid Swift code.
The technique replicates Fable behavior on existing models: By treating Opus outputs as claims verified by parallel sub-agents, Fernando achieves autonomous multi-step workflows that others struggle to accomplish without the discontinued Fable model.
The Breakdown
Ray Fernando builds a real-time transcription app using GPT Realtime 2 that translates speech across 70+ languages instantly, then demonstrates his "wave engineering" technique that orchestrates multiple AI agents in parallel to replicate Fable-like behavior without needing access to the actual model. The app works on both iOS and Mac, and he's testing it with elderly users who have hearing aids and memory challenges.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.