Start in Step

You're Not Lazy. You're Just Overwhelmed. Stop beating yourself up for procrastinating. Get a 24/7 AI coach that understands your barriers and breaks them down into 5-minute action steps.

Comment

🚀 Start In Step — Project Story

🌱 Inspiration

Start In Step was born from reading Devon Price’s essay “Laziness Does Not Exist.”
It reframed procrastination not as laziness, but as an emotional response to burnout, overwhelm, fear, or shame. That resonated deeply with me. It inspired me to create something that meets people in that moment of stuckness with kindness, not judgment.

The core idea: people don’t need pressure, they need permission to take just one small, doable step.

💬 What it does

Start In Step is an AI-powered mentor that helps people move from overwhelm into motion by:

  • Listening with empathy (no shame, no productivity hype)
  • Identifying emotional blockers
  • Guiding users through a 3-phase conversation:
    Empathy → Clarity → Momentum
  • Generating 5-minute micro-steps to get started
  • Helping build streaks, routines, and confidence over time
  • With the use of modern tech, this coach is available 24/7

It feels more like a conversation with a thoughtful coach than using a chatbot or productivity tool.

🛠️ How we built it

We used:

  • OpenAI GPT-4 for contextual conversation and coaching
  • Bolt.new for rapid no-code frontend development
  • Supabase for user/session data and micro-plan history
  • Mixpanel & GA4 for analytics and behavior tracking
  • A custom TypeScript prompt engine to manage AI state, conversation stages, and logic transitions
  • Carefully structured conversation stages to simulate natural coaching flow

The app supports onboarding without an account, persistent sessions, and context-aware micro-plan suggestions.

😅 Challenges we ran into

  • Too much empathy, too little progress:
    Early versions were validating but didn’t help users act. We restructured the prompts to guide more assertively through emotional friction toward tiny steps.

  • Getting the tone right:
    Striking the balance between warm and actionable was hard. It took dozens of prompt iterations.

  • Trigger timing for action steps:
    We needed to avoid giving suggestions too early (felt robotic) or too late (felt frustrating). Solved with a state-based ladder model.

🏆 Accomplishments that we're proud of

  • Developed a real-time, AI-powered coach that adapts to each user’s emotional state
  • Created a stage-aware prompt system with structured outputs (e.g. STAGE_UPDATE, TRIGGER_MICROPLAN_MODAL)
  • Achieved a working MVP in under 2 weeks using vibe coding principles
  • Built a micro-step engine that encourages forward motion without shame
  • Got early validation from users saying: “This feels like what I needed, not another todo list.”

📚 What we learned

  • Empathy isn’t enough, users need clarity and momentum too.
  • Prompt design is product design when working with LLMs.
  • Micro-wins are powerful motivators when action feels impossible.
  • The best coaching doesn’t solve everything, it helps users feel safe enough to take the next step.

🔜 What's next for Start In Step

  • Launch soft beta for early users and feedback loops
  • Add calendar integrations and gentle reminders
  • Monetize with a freemium + starter subscription model
  • Expand language support (Dutch first, then EU)
  • Partner with educational and wellness orgs to bring Start In Step to students and professionals

We’re just getting started, but even one small step changes everything. And I have been maximally using AI, not only to build the product, but also to structure this as a project, finetune text and all debugging along the way

Built With

Share this project:

Updates