Killed By Claude Report
← HomeCurrent verdict
Assessment
Anthropic keeps shipping the exact primitives that eat generic language-learning experiences: better reasoning, long memory, computer use, mobile access, voice-style conversational use cases, and explicit education/learning modes. A user can increasingly ask Claude to be a personalized language tutor, generate drills, explain grammar, remember weak spots, and build adaptive study materials. That pressures Duolingo's core app experience. What still saves Duolingo is productization: daily habit loops, gamified retention, structured curriculum, learner data, speech/listening exercises at scale, school distribution, and its English test. Claude is a powerful substitute for lessons; it is not automatically a superior consumer learning product.
Biggest historical hit
2026-02-17 — Introducing Claude Sonnet 4.6. The announcement says Sonnet 4.6 is improved across knowledge work and design, with 1M context, memory-related tooling, and broad app access; paired with Duolingo's own claim of 'Combining the best of AI and language science' and personalized lessons, this is the clearest direct encroachment on AI-personalized tutoring.
What still protects them
Duolingo's moat is not raw model capability. It's behavioral design. They have a polished mobile app, habit formation, streaks/challenges, bite-sized curriculum, language-science packaging, a giant installed base, school workflows, kid literacy products, and a branded assessment product. Claude can impersonate a tutor; Duolingo has a tuned consumer system that gets millions to actually come back tomorrow.
Signals
Why this is in the blast radius
Duolingo explicitly markets 'personalized learning' and 'the best of AI and language science.' Sonnet 4.6 is positioned as a broad upgrade for reasoning, long-context understanding, knowledge work, and design, making Claude much more capable as a personalized tutor that can explain grammar, generate exercises, adapt difficulty, and sustain long learning sessions.
This is the most obvious educational overlap. Anthropic explicitly shipped a 'Learning' style meant to guide users through concepts rather than just answering. Duolingo is a learning product. If Claude can act as a patient, adaptive teacher, it attacks the core value proposition of guided instruction.
Language learning improves when the system remembers mistakes, vocabulary gaps, pacing, and prior conversations. Duolingo's edge includes tailoring lessons to level and pace. Claude memory narrows that gap by enabling persistent, personalized tutoring without requiring a dedicated language app.
This expands Claude from chat tutor to study-material generator: worksheets, vocab lists, grammar drills, flashcards exported into documents, and custom lesson packs. That doesn't replace Duolingo's app loop, but it does make Claude a viable alternative for self-directed language study.
This is about health-data integrations, not education or language learning. It shows Anthropic broadening consumer utility, but it does not specifically attack Duolingo's language-learning app or testing products.