Logo
Overview
The Genesis of Experiments AI: Living Freely in a Goal-Obsessed World

The Genesis of Experiments AI: Living Freely in a Goal-Obsessed World

How reading Anne-Laure Le Cunff inspired me to build a platform for scientific habit formation and AI-powered self-improvement.

June 15, 2025  •  4 min read

experiments-ai productivity flutter firebase gen-ai

The world today is hyper-focused on goals, metrics, arbitrary KPI-driven scores, and productivity tracking. You’re constantly bombarded with apps that gamify your life—awarding you badges, streaks, and punishing you if you fail to meet rigid daily targets.

However, earlier in 2025, I read the book Tiny Experiments: How to Live Freely in a Goal-Obsessed World by Anne-Laure Le Cunff. The book entirely shifted my perspective from rigid goal-setting to curiosity-driven experimentation. I realized that optimizing my life shouldn’t be about stressing over failing goals, but rather conducting mindful behavioral “experiments” to scientifically observe what works.

With this philosophy, I built Experiments AI, a life experimentation platform designed to conduct personal A/B tests through data-driven self-improvement.

Note

The Philosophy: What if, instead of feeling guilty for breaking a streak, we treated habit changes like scientific experiments? The focus changes from “I failed” to “This methodology produced these specific results. Let’s try another variable.” Let’s build a social media for scientists, where the laboratory is your life!

Redefining Metrics: The Science of Happiness

Instead of gamifying every aspect of life with meaningless points, Experiments AI uses a core metric called the Personal Happiness Index (PHI - ϕ). The index operates on a straightforward 0-100 scale.

To establish a baseline, new users complete an interactive Life Sphere Wheel assessment. This UI features a segmented circle covering 7 life categories (Mind & Body, Career & Growth, Financial Well-being, Social Connections, Love & Romance, Joy & Hobbies, Personal Space).

Users drag emoji-based sliders (🥶 to 🔥) to express their current satisfaction. From this, the PHI baseline is formulated via a weighted algorithm.

Social Learning & Community

The app acts as a social learning network for self-improvement. When users create an experiment, they have two standard templates:

  1. Replacement: “I am going to experiment with doing <activity> for <time period>, comparing it with <previous-activity>.”
  2. New Activity: “I am going to experiment with doing <activity> for <time period>.”

Instead of fighting the journey alone, the experiments are shared publicly.

  • Peer Reviews: Other users can leave feedback, mark comments as “helpful,” and build a community reputation (“Helpful Streaks”).
  • Experiment Forking: See someone trying cold plunges and succeeding? You can “clone and modify” their experiment settings straight to your own profile!
  • Live Ticker Animation: Follow the ongoing experiments of others and see real-time tracking updates.
Note

The Engineering Challenge: Creating a highly animated, cross-platform social feed that handles multimedia input (voice, text, images) offline-first while syncing seamlessly in the background to Firestore was one of the hardest technical puzzles to solve.

AI Insights: Integrating Gemini AI

To deepen the experience and avoid making users manually process large amounts of qualitative data, I deeply integrated Google’s Gemini AI.

If you’re unsure what to experiment with, the Hypothesis Helper chat assistant gently interrogates your current struggles and formulates a scientifically structured experiment hypothesis.

Throughout the experiment, users submit Daily Field Notes. This feature allows users to record their thoughts via voice, image, or text and tag an emotion on a visual selector. Here is where the magic happens: The Field Note Analysis system utilizes LLMs to analyze these qualitative daily field notes in the background over time dynamically predicting and recalculating the user’s PHI impact.

lib/features/shared/models/experiment_model.dart
// Exploring the Data Model
class ExperimentModel {
  final String experimentType; // 'replacement' | 'new'
  final String? baselineActivity;
  final String targetActivity;
  final Duration duration;
  final double phiBaseline;
  final double phiCurrent;
  final bool isSharedPublicly;
  final List<FieldNote> fieldNotes; // Houses Multimedia + Emotion Data
  final List<PeerReview> peerReviews;
}

Instead of flooding social feeds with disjointed voice notes, a background queue triggers an AI Summarization pipeline. It takes the day’s field notes, digests the complex experiment updates, and provides highly readable, synthesized feeds to your followers.

Tech Stack & offline-first Design

Experiments AI is built entirely using Flutter, heavily customized with Material 3 UI aesthetics. This ensured the app would deploy uniformly and beautifully across Android, iOS, web, Linux, macOS, and Windows.

Offline-first Robustness: A major technical constraint I placed on the app was that core functionality (PHI tracking, field notes, experiment management) must work offline. I implemented Hive—a lightweight, blazing fast, key-value database written in pure Dart—for local storage.

If the app isn’t connected to the internet, all field notes, emoji ratings, and multimedia files are stowed securely in Hive. Once a connection establishes, the app syncs incrementally with Firebase Firestore.

This robust architecture perfectly embodies the theory of Tiny Experiments: it gives anyone the structure to try new paths toward happiness without the pressure of an absolute end goal—wherever they are in the world.