What We Learned Building 30 AI Apps in 30 Days
In August, we undertook a focused experiment: to build 30 distinct AI applications in 30 days. This marathon of coding tested our endurance and a core philosophy called Generative Work: how we build thoughtfully with AI to achieve novel and superior outcomes while nurturing human development.
The challenge was designed to explore the full range of this approach. We created practical, assistive tools and, in some cases, pushed each of the guiding principles into its most extreme territory, which is where the deepest learning occurred. This journey was guided by three principles that served as our north star.
Principle 1: AI Can Either Foster or Hinder Your Development. Be Deliberate with Your Intent.
Before using a tool, you must clarify your purpose. AI amplifies the strategies you bring to it. We built applications all along the spectrum of learning and doing, trying to get that mix right, moment by moment.
To Scaffold Learning: The framework reminds us that “Most Work Blends Learning and Doing.” We explored this balance with tools like Story Steps. Built with a children's author, it helps kids write their first stories by breaking the daunting task into manageable parts, nurturing their development as storytellers.
To Capture Human Knowledge: In tools like our Ideation Canvas and Design Thinking modules, we built AI agents that facilitate human collaboration. They listen to conversations and structure them into diagrams, freeing up human partners to focus on creative work.
Pushing to the Extreme of Pure Thought: Misusing AI, as the principle warns, "limits your potential." To avoid the trap of "automating when you should learn," we built SocraGPTes. It refactors the Socratic method to turn the AI into a partner that questions your assumptions. Its entire goal is to improve the quality of your own thought, rather than simply “get things done.”
Principle 2: Working With AI is Like Managing a Team. Learn the Art of Guidance.
You are not just a doer now. You are a director. To get quality results, you must guide AI by effectively clarifying and communicating the work to be performed. We found this skill ranges from clear direction to deep empathy.
As a Director of a Team: To create the Generative Work Suite, we had to "Become Multi-disciplinary." The AI acts as a skilled team member, but it requires human leadership. The user must provide the "creative vision, product thinking, [and] business strategy" through natural conversation to guide the AI in producing diagrams and documents.
Become Multi-disciplinary Building the AI Bachelorette required more than simple instructions. To create believable, emergent social drama, we had to give each agent a deep persona informed by clinical psychology, providing them with the context needed to guide their behavior.
Be an AI-Anthropologist: Guiding AI is as Much Art as Science. AI doesn't always see the world like we do--empathy and the humanities matter.” Our attempt to animate Holbein's painting The Ambassadors was the ultimate test. Direct commands failed because the AI is an alien collaborator that perceived two distinct paintings. Success only came when we used intuition to understand its perspective and provided the missing context: a view of the painting within its museum room.
Principle 3: AI Will Revolutionize Work. To Keep Up, You Must Innovate.
The core of this principle is simple: "Don't just optimize the old. Create the new." We aimed to build things that were previously out of reach, setting our sights higher to create truly novel capabilities.
Unstructured-to-Structured Transformation: We discovered a nearly magical ability in LLMs to take chaotic, unstructured human input and instantly pour it into a formal data model. We leveraged this in Clarity Coach, which processes a stream-of-consciousness "brain dump" into an organized GTD-style dashboard. Similarly, Structured Notes turns rambling memos into perfectly formatted notes on the fly, structured according to user-defined templates. This creates entirely new workflows for capturing and organizing information, bridging the gap between fleeting thoughts and actionable data. Finally, Weave formats and organizes unstructured information about your contacts and even makes thoughtful suggestions, turning “John likes chocolate” into “Gift ideas: Chocolate” under John’s contact entry.
Agentic AI as a Force Multiplier: We took seriously the call to "Aim for outcomes neither humans nor machines could achieve alone." We learned that by orchestrating multiple AI agents, we can achieve results far superior to what a single base model can produce. In AI Colab, we can research team dynamics by having AI agents collaborate and critique one another's work. In Backpocket Boardroom, a single user can stress-test a business plan against a virtual panel of executive AI personas, completing in minutes a process that would have taken weeks.
The Final Takeaway
To cap off the challenge and share these learnings with the broader community, we had the privilege of presenting our work at the Generative Series event on AI, art, and culture in Washington, DC. It was an incredible opportunity to move the conversation from our screens into the real world, discussing how these principles apply not just to code, but to creative and cultural pursuits.
Sharing these lessons was powerful, but the real value lies in the process itself. The acceleration and depth of learning possible when you immerse yourself in a challenge like this is an experiment we would recommend to anyone serious about mastering this new wave of technology. It's a transformative journey, and if you're interested in learning how you could structure your own 30-day challenge, please don't hesitate to reach out. We would love to share what we learned.
All content related to this account reflects solely my personal views, not the views of any employer, past or present.
The 30 Apps in 30 Days
SocratGPTes: A conversational AI embodying Socrates to help users examine their own beliefs through rigorous dialectic.
Data Model Design: An AI co-pilot that translates spoken design conversations into formal Entity Relationship diagrams and data dictionaries in real-time.
AI Colab: A simulation platform for analyzing AI teamwork dynamics by running experiments with customized AI agent personas.
Story Steps: A guided storytelling tool for children that scaffolds the creative writing process into manageable, AI-assisted steps.
ScrollyTelling Studio: An AI-powered presentation builder for creating interactive, scroll-driven narratives without needing to code.
MicCheck: A developer utility for simultaneously testing and comparing the performance of multiple real-time speech-to-text APIs.
Back Pocket Boardroom: A strategic simulation tool that stress-tests ideas by generating feedback from a panel of customizable AI executive personas.
À la Recherche du Langage Perdu: An interactive art installation that juxtaposes a generic AI answer with a literary quote from Proust to provoke reflection.
AI Bachelorette: A "social physics engine" that simulates emergent social dynamics using autonomous AI agents with deep psychological profiles.
Pocket Hide & Seek: A location-based mobile web game that digitizes the classic childhood game using GPS and periodic audible "pings."
Treasure Hunt: A location-based multiplayer game where an "Architect" designs a hunt and "Explorers" use GPS to find virtual treasures.
Traffic Simulator App: A full-stack application for designing, running, and analyzing complex traffic simulations on real or custom road networks.
Design Studio: An AI-powered design studio for creating interactive, scroll-based presentations with AI-generated slides, diagrams, and themes.
Tailored News Analysis: A personalized strategic intelligence advisor that analyzes news through the lens of user-defined professional personas.
SerendiCity: An AI-powered travel tool that generates a personalized itinerary of fictional, inspirational events to capture a city's unique "vibe."
Chronicle Weaver: A dynamic, choice-driven storytelling engine where an AI generates a unique, branching narrative guided by a 15-segment story arc.
Market Sim: A behavioral economics platform that simulates stock market dynamics using AI agents with configurable trading personalities.
Sparky's Story Lab: A collaborative, real-time storytelling platform for young writers featuring an AI assistant for brainstorming and analysis.
Structured Notes: An AI-powered application that automatically organizes memos into user-defined structured templates.
Generative Work Lab: An integrated suite of AI-powered productivity tools for strategic thinking, planning, and project execution.
Weave: A contacts app that structures and organizes information about your contacts in a user friendly way.
Ideation (Enhanced): A voice-driven brainstorming tool within the Lab that translates live conversations into evolving diagrams.
Habit Builder: An experimental tool to design small, safe behavioral tests to overcome hidden assumptions blocking personal growth.
Learning Hub: A conceptual space for using AI to structure and accelerate personal learning plans and skill development.
Adaptive Change: A guided coaching module in the Lab to uncover and address deep-seated resistance to personal and professional goals.
Deep Research: An intelligent agent that synthesizes information from multiple sources to provide deep, contextualized research briefings.
Design Thinking: A structured, multi-phase innovation module in the Lab that guides users from problem definition to a planned MVP.
ClarityCoach: A tool within the Lab to process unstructured "brain dumps" into organized, actionable categories using GTD methodology.
North Star: A strategic planning module in the Lab for defining long-term vision, mission, and quarterly OKRs.
Clarity Flow: An integrated project management system in the Lab that serves as the destination for all tasks generated from strategic modules.
Open Design ("Build a Cool App"): A meta-project representing the open-ended exploration of building novel applications with a human-AI collaborative process.
