AI Friends That Actually Call You (And They Remember Everything)
with Medi, Commitify
AI Friends That Actually Call You (And They Remember Everything)
Show Notes
Medi built Commitify in a 100-hour sprint with one goal: go viral or make money before the clock ran out. The idea was deceptively simple - an AI that calls your phone, checks in on your goals, remembers what you said last time, and holds you accountable across a cast of distinct personas: Life Coach, Zen Master, Slay Bestie, Hype Beast, Drill Sergeant, and CEO.
It landed on Product Hunt, hit 3rd place, and found an audience nobody had predicted. The people who loved it most were not productivity nerds. They were users who wanted a consistent voice on the other end of the phone - one that remembered their history, matched their personality, and showed up whether they asked for it or not. That is proactive AI accountability, and it is a category that barely exists yet.
The 100-Hour Build Constraint
Medi did not set out to build a startup. He set out to run an experiment with a hard boundary: 100 hours, two goals, pick one or both. Go viral. Make money. The constraint eliminated feature creep and forced a single-question test - does this resonate with real humans fast enough to matter? Product Hunt gave him a forcing function for public launch. Third place on day one was validation, not just vanity.
The lesson is not "build fast." It is "constrain ruthlessly and measure the right output." Most founders spend six months building toward a launch date. The 100-hour model inverts that: launch is the experiment, not the reward. You learn whether the core mechanic works before you have invested enough to be biased about the answer.
Persona-as-Distribution
The six Commitify personas are not just UX choices - they are a distribution engine. Each one speaks to a different identity. The Slay Bestie hits the affirmation community. The Drill Sergeant appeals to ex-athletes and discipline-focused users. The Zen Master lands with mindfulness audiences. When someone shares a clip of their call, they are sharing a persona - and that persona travels with its own built-in audience.
This is the "shareable character" approach to virality. You are not marketing a feature. You are marketing a relationship archetype that people already recognize and want. The product becomes its own ad when the persona is distinct enough to be quotable, memorable, and slightly surprising.
UI-Less Voice-First Design
Most AI products are built around a screen. You open an app, type a prompt, wait for a response, close the app. The friction is high enough that usage decays rapidly after the first week. Commitify removes the interface entirely - the product is a phone call you did not initiate. It finds you.
This is a meaningful design choice with real tradeoffs. There is no dashboard for the user to fiddle with between calls. There is no notification they can ignore. The call either happens or it does not. That limits certain engagement loops but creates a fundamentally different relationship - one that feels less like software and more like a standing appointment. The voice layer, powered by ElevenLabs, carries the emotional weight that text-only AI cannot.
Long-Term Conversational Memory
The product's stickiness depends on one technical and one emotional commitment: the AI remembers what you said. Not just your name and goals - it remembers the conversation from last week, what you said you were going to do, and whether you followed through. Over time, the model builds a picture of each user's patterns, defaults, and growth trajectory.
This is not novel as a technical problem - context windows and memory stores exist. What is novel is the use case: a daily or weekly call where value compounds over time. Most users who try Commitify once get a fun novelty. Users who stay for a month get something that resembles genuine accountability. The retention logic depends entirely on whether the memory layer holds.
Finding the ICP Before Optimizing
Medi does not know exactly who his user is yet - and that is a deliberate choice. After Product Hunt, segments are emerging: ADHD users who respond to external structure, Gen Z users who prefer voice to text, and the affirmation community drawn to the Slay Bestie persona. Rather than picking a lane and marketing hard into it, Medi is watching natural clustering before optimizing the funnel.
The underlying principle: find 10 people who love the product as-is, then ask why. Not "how can we make it better" but "what specifically about this made you come back?" The answers shape positioning, copy, and the persona roadmap - not the other way around.
Q&A
What exactly is Commitify and how does it work?
Commitify is an AI accountability coach that calls your phone on a schedule you set. You pick a persona - Life Coach, Zen Master, Slay Bestie, Hype Beast, Drill Sergeant, or CEO - and the agent calls you, checks in on your stated goals, and maintains a memory of your previous conversations. There is no app to open. The call is the product.
How did the product get built in 100 hours?
Medi set a hard constraint: 100 hours to hit at least one of two targets - go viral or generate revenue. The product launched on Product Hunt during that window and finished 3rd. The constraint prevented over-engineering and forced prioritization to the single mechanic that would prove or disprove the core idea: will people pick up a phone call from an AI and find it valuable?
Why six personas and how were they chosen?
Each persona maps to a recognizable archetype - rooted in Jungian psychology and modern internet culture. The Drill Sergeant speaks to discipline-focused users. The Zen Master speaks to the mindfulness space. The Slay Bestie hits affirmation culture. The CEO appeals to founders. Each persona functions as both product differentiation and a distribution channel - carrying its own built-in audience and shareable identity.
What makes the UI-less approach better or worse than an app?
Better: no friction, no notification to ignore, no app decay. The call finds you. It creates a relationship dynamic closer to a standing appointment than a software feature. Worse: no dashboard, no self-serve configuration beyond initial setup, no user-initiated sessions. The product works best for users who want the AI to hold the initiative - which is not everyone.
What is the pricing model and why weekly?
Commitify runs on a weekly subscription with tiers from 3.75 EUR for 3 calls per week to 15 EUR for 21 calls per week. Three free calls are included before the paywall. Weekly pricing maps naturally to the core use case - a recurring accountability cadence - and makes the value proposition legible: you are paying for a consistent coaching relationship, not a one-time tool.
Who is the target user and how is Medi figuring that out?
The ICP is still emerging. Early signals point toward ADHD users who respond to external structure, affirmation community members drawn to the Slay Bestie, and Gen Z users who prefer voice interaction over text. Medi's approach is to watch natural clustering in the existing user base before committing to a positioning. The goal is to find 10 users who love the product as-is and understand specifically why before optimizing the funnel.
Is the persona IP protectable and what is the real moat?
It is a legitimate question. The personas as designed - specific names, voice styles, and behavioral patterns - represent creative IP that is difficult to patent but potentially protectable as trade dress or copyright. The larger moat is memory: a Commitify relationship that has six months of conversation history has switching costs no competitor can replicate by copying the persona names. The relationship is the moat, not the mechanic.