The Prompt That Started It All
"Build a Gradio app that visualizes AI learning curves across domains with Plotly. Let users select domain, see animated learning progression, and compare pre-ChatGPT vs post-ChatGPT learning rates."
That was the prompt. What followed was a 90-minute vibe coding session that produced a fully interactive visualization tracing AI capability across six domains.
Why Learning Curves Matter
The AI industry loves benchmarks. MMLU scores. HumanEval pass rates. GSM8K accuracy. But a benchmark is a snapshot — it tells you where a model is, not how it got there or how fast it's improving.
Learning curves tell the real story. They reveal:
Acceleration patterns — which domains are improving fastest and why
Regime changes — the visible inflection at ChatGPT's launch
Nascent vs. mature domains — agentic systems exploded from near-zero; text generation improved from an already high baseline
The Two-Sigmoid Model
Each domain's trajectory is modeled as the sum of two logistic functions — one for the pre-ChatGPT era of gradual research progress, and one for the post-ChatGPT acceleration. The result is a smooth S-curve with a visible inflection point.
This is deliberately simple. A single sigmoid would miss the regime change. Three or more would overfit. Two captures the essential story: something fundamentally changed in late 2022.
The Build Process
The entire visualization was built conversationally with Claude Code:
First pass — Claude generated the core Gradio app with Plotly charts, domain data, and sigmoid math
Iteration — Added milestone markers, domain comparison overlays, and the rate acceleration bar chart
Documentation — Added a full Methodology tab with parameter tables and benchmark sources
Integration — Ported the Python Plotly code to a React component for inline rendering in the Astro site
The meta-narrative matters: AI was used to build a tool that measures AI effectiveness. The process itself demonstrates the thesis.
The Acceleration Isn't Uniform
The most interesting finding: post-ChatGPT acceleration varies dramatically by domain.
Agentic Systems saw the steepest curve because the field essentially didn't exist before 2023. Reasoning & Math saw a qualitative shift — from pattern-matching to inference-time compute scaling. Text & Writing improved from the highest baseline but with the least dramatic acceleration.
The floor was raised everywhere, but the ceiling moved at different speeds.
Try It Yourself
The interactive visualization is live at AI Learning Curves. Drag the slider, compare domains, and explore the methodology.