Waffle

What do they actually do

Waffle is building an AI language tutor delivered as a web app (and likely a simple mobile app) where learners practice through a chat-style conversation. Users pick a target language and level, then run short sessions focused on conversation, grammar micro-lessons, vocabulary drills, and basic speaking practice via the device microphone.

The product provides real-time corrections with brief explanations, queues vocabulary for review, and shows simple progress summaries (recent lessons, streaks, completed units). It likely supports a handful of major languages, offers a free tier for basic practice, and a paid tier for longer sessions and guided curricula. Early versions prioritize smooth conversational UX and repeatable lesson templates over full, human-crafted curricula.

Who are their target customer(s)

• Self-directed working professional needing language skills for work or travel: Has limited time and budget; wants targeted, just-in-time practice and clear corrections, not generic tips or rigid schedules with human tutors.
• University or high-school language student: Needs measurable progress and exam-style practice; currently gets inconsistent corrections and no personalized plan that addresses repeated mistakes.
• Casual learner seeking speaking confidence for trips or social settings: Feels anxious practicing with strangers; most apps don’t provide realistic conversation or actionable pronunciation feedback.
• Language teacher or small language school: Spends time grading and giving individual feedback; lacks affordable tools that automate reliable speaking corrections and provide student analytics.
• Absolute beginner: Gets overwhelmed by grammar-heavy instruction; existing tools move too fast or don’t give immediate, simple corrections for basic pronunciation and sentence formation.

How would they acquire their first 10, 50, and 100 customers

• First 10: Direct founder outreach to known professionals, students, and teachers; provide free access for frequent use, run hands-on onboarding, and collect candid feedback and testimonials.
• First 50: Post in language-learning forums, campus clubs, and meetup groups; run short pilots with a few teachers/classes in exchange for discounts and referrals.
• First 100: Package successful pilots into a teacher-onboarding kit and referral rewards; run small, targeted paid experiments to professionals and students, measuring conversion and cost per paying user to double down on efficient channels.

What is the rough total addressable market

Top-down context:

We frame TAM as the global set of learners and institutions willing to pay for an AI language tutor multiplied by an annual price; under a base-case adoption and pricing mix, this yields about $8B per year.

Bottom-up calculation:

Base case: 80M paying users × $100/year ARPU = ~$8.0B TAM; initial SAM at 20–35% of TAM ≈ $1.6–$2.8B; early SOM targets equate to ~$0.4M–$80M/year depending on penetration and ARPU.

Assumptions:

• ARPU in the $60–$200/year range depending on mix of subscription and upsells.
• Initial support limited to major languages and higher-adoption geographies.
• Institutional/teacher seats included as seat-equivalents within paying user counts.

Who are some of their notable competitors

• Duolingo: Mass-market, gamified lessons with light conversation and speaking exercises; convenient and low cost but corrections skew generic and drill-based rather than deeply personalized practice.
• ELSA Speak: Mobile app focused on automated pronunciation feedback; strong on phonetic accuracy but narrower than a full conversational tutor or curriculum.
• italki: Marketplace for one-to-one human tutors; great for nuanced conversation but lacks always-available, low-cost automated practice.
• Speechling: Speaking drills with coach corrections (human review plus automation); sits between pure AI apps and full human lessons for targeted speaking feedback.
• HelloTalk: Peer-to-peer language exchange with native speakers; offers authentic conversation but limited structured tutoring and consistent corrective feedback.