The AI code completion tool with 500,000 developer users that shut down in November 2022 — killed by GitHub Copilot, the same technology it pioneered, built by the company whose code repositories fed its training data
Quiet closure with no public announcement · Fatal mistake: GitHub Copilot launched with OpenAI Codex trained on GitHub's entire public code repository — orders of magnitude better training data and model quality than Kite could access with $17M
Evaluating only Kite’s profile at its peak — without knowing the outcome — the model ranked Competition as the #1 likely cause. That’s exactly how it died.
Key Events Timeline
FOUNDING
Kite founded
PIVOT
Strategic pivot under pressure
PRODUCT LAUNCH
Kite reaches 500,000 active developer users across VS Code, Atom, Sublime Text integrations; among the most widely used AI code tools before GitHub Copilot era.
SHUTDOWN
Silent Shutdown: Kite ceases operations
SHUTDOWN
Kite closes; Adam Smith publishes candid post-mortem noting the company never found a business model despite 500,000 users; GitHub Copilot's qualitative superiority made competition impossible on $17M in funding.
Full Analysis
Free · no account needed
Documented cause
Kite was founded in 2014 by Adam Smith in San Francisco with an idea that proved to be perfectly correct and terminally early: machine learning models could predict what a developer was about to type, reducing repetitive coding work and helping engineers stay in flow. Kite integrated as a plugin into VS Code, Atom, Sublime Text, and other editors, analysing local code context and suggesting completions. Over several years, Kite accumulated approximately 500,000 active developer users — a meaningful distribution in the developer tools market. The company raised approximately $17 million from investors including Foundry Group. In June 2021, GitHub launched GitHub Copilot in technical preview — powered by OpenAI's Codex model trained on GitHub's own repository of public code. Copilot was qualitatively better than Kite in almost every measurable dimension: it suggested entire functions rather than line completions, it had been trained on orders of magnitude more code, and it was backed by Microsoft's Azure infrastructure and OpenAI's research team. Kite attempted to compete by improving its models and expanding features, but the resource gap between a $17M company and a Microsoft-owned product with OpenAI's model research was unbridgeable. In November 2022, Adam Smith published a candid post-mortem announcing Kite's closure. He noted that Kite had never found a business model: even 500,000 users had proven insufficient to generate enough revenue to sustain the company because developers were reluctant to pay for code completion tools. GitHub Copilot, initially priced at $10/month, validated that developers would pay — but only at the point where the product was already orders of magnitude better than Kite. The Kite story is a precise illustration of being right about the thesis, building the category, and then being unable to survive the category's winner.
Lesson
“Being first to prove a category does not protect you when the category's winner is built by the organization that controls the training data, the distribution channel, and the developer workflow that the product lives inside. GitHub owned the code corpus, the developer identity (git login), and the editor integration surface — three structural advantages that made GitHub Copilot better, more trusted, and more integrated than any external competitor could be. Kite was right; it just couldn't survive being right.”
Failure anatomy
Collapse type
Silent Shutdown
🐌 LOW
Hype cycle
ai code completion wave
Moat type
Editor Plugin Distribution + Local Code Context
Fatal mistake
GitHub Copilot launched with OpenAI Codex trained on GitHub's entire public code repository — orders of magnitude better training data and model quality than Kite could access with $17M