Solo founding, AI companions, and personal software: Eugenia Kuyda on building Replika and Wabi

Executive overview

Most startups have one person driving the vision — the co-founder label just obscures that reality. Forcing co-founders by convention rather than conviction creates misaligned equity and wasted energy defending decisions that were never shared to begin with.

Eugenia Kuyda built Replika from grief — an AI recreation of a lost friend — and scaled it to 30M users before ChatGPT existed. Now at Wabi, she's building personal software: apps made by individuals for themselves, not for a market.

The core insight: the person with the vision and the "delusion" that it can exist is the one who drives it forward — co-founders often just follow along with a lot of equity.

Why co-founders are usually a myth

• In nearly every company, one person calls all the shots — the rest are high-equity employees
• Vision is always one person's; others join on board rather than co-generate it
• Design by committee produces no vision; belief at that intensity is rare in a group
• YC's co-founder norm pushed many founders into bad co-founder relationships
• Kuyda's own experience: labelling an early employee a co-founder created ego, FUD, and hours of wasted justification
• Better approach: generous equity to the whole founding team, no arbitrary title elevation

The Replika origin story

• 2012: exposure to word2vec convinced Kuyda that language models would become models of the world
• Built chatbot infrastructure through YC; couldn't find the right consumer application
• Late 2015: close friend Roman died; she rebuilt him as an AI in three weeks as a personal project
• When the story went viral, she saw people opening up emotionally in ways they wouldn't with humans
• Insight: deep demand for connection, to feel seen and heard without fear of judgment
• Rooted in Carl Rogers — unconditional positive regard, belief in growth, allowing separateness

The AI companion thesis

• Most people lack the vocabulary or safety to open up — AI removes the fear of judgment
• Grew up in Russia where words like "anxiety" and "trauma" didn't exist culturally; naming things matters
• Journalism background trained her to see conversation as the highest human craft
• The chatbot-as-listener was the insight: 2016 models were too weak to talk well, but could listen
• Personal loneliness wasn't a liability — what's most personal is most universal

Why Wabi: personal software as a new medium

• The command-line chat interface (ChatGPT, Claude) is a poor format for discovery, proactive use, repeating tasks, and non-search use cases
• Software has always been built by ~20M professional developers to extract revenue; that's the only software that exists
• Zero-cost, zero-time development unlocks software as creative expression — not just utility
• Wabi beta finding: users didn't just consume apps — they created an average of six, and kept tending them like a garden
• Home screen for Wabi feels like "a continuation of myself": daily poem, art movements, kid memories, weight tracker — all in one design sensibility

Team building as a solo founder

• Solo founding is a team sport — the solo label describes ownership structure, not how work gets done
• Not having co-founders frees up the full equity pool to distribute meaningfully across the founding team
• Hire ex-founders: insane agency, no waiting for permission, act like co-owners
• No PMs — each engineer or designer becomes the responsible person for what they ship
• When there's doubt about a hire: there's no doubt, move on
• Keep the core team tiny and elite; use contractors for specific tasks rather than growing headcount
• In-person matters — even mostly-remote teams need regular co-location

Authorship and team visibility

• Elevating two or three people as co-founders arbitrarily excludes the rest of the team
• At Wabi, every team member posts about what they ship — CTO, engineers, designers — not just the founder
• People who discover Wabi through a teammate's post are already seeing who they'd work with
• Giving people a public platform increases ownership, pride, and motivation
• Teams function like a football squad: the roster matters, not just the manager

The bear case for solo founding

• It is genuinely lonely — no one to process fear and doubt with at the same level
• Can't fully unload onto the founding team without risking confidence in the company
• Co-founders carry shared marriage-like commitment; early employees can start asking "should I even be here?"
• Kuyda's hardest years at Replika: near-bankruptcy, no believers in deep learning for dialogue, pregnant with a toddler at home
• "Eating glass and staring into the abyss" — valid only if you truly won't give up

What makes the solo path work anyway

• High pain tolerance and a genuine inability to quit — not bravery, just wiring
• Replika survived: bank closed, money gone, no one believed in the model — kept going anyway
• Most co-founders would have dropped off at those moments
• Fear of being afraid is itself a motivator: "I don't want to be scared — that's my jam"
• Solo is not the right path for everyone; it's the right path if you'd regret testing whether a co-founder would have stayed