ModelYard · Museum of Unfinished AI Projects
An archive of AI projects that were stopped, shelved or transformed: what they set out to do, what is documented about the turn, and what remains.
Browse the archive Suggest a project
10 archive entries
A free web tool that guessed whether a passage of English text was written by an AI — built for teachers and editors, withdrawn after seven months because it could not be trusted.
OpenAI released a public classifier in January 2023 to help educators and editors judge whether text was AI-written, describing it as a stopgap while better provenance methods were researched. In July 2023 the company m…
Copilot Voice — originally shown as "Hey, GitHub!" — was a voice-driven coding experiment from GitHub Next, framed as an accessibility tool for developers who have difficulty typing. It ran as a technical preview and en…
AI Test Kitchen launched in August 2022 as a small set of LaMDA demos the public could try. It later hosted MusicLM and dropped the earlier demos. In August 2023 Google delisted the Android and iOS apps and confirmed th…
The Ai Pin was Humane's attempt to replace the phone screen with a voice-first wearable: a lapel device that projected answers onto the wearer's palm and ran on Humane's CosmOS. It began reaching customers in April 2024…
Kite was a startup that used machine learning to help developers write code, running from 2014 until it stopped working on the product in 2022. Its founder's farewell post is an unusually explicit post-mortem: the compa…
AI Shell was Microsoft's attempt to bring LLM agents into the terminal: an interactive shell plus a PowerShell module that let you ask for a command in natural language, keep an agent pane beside your prompt, and hand r…
provenance (1) detection (1) education (1) api-and-tools (2) developer-tools (4) accessibility (1) voice (1) copilot (1) demo-apps (1) lamda (1) google-labs (1) product-surface (1) ai-hardware (1) wearables (1) voice-interface (1) consumer-device (1) startup-shutdown (2) code-completion (1) post-mortem (1) too-early (1) business-model (1) terminal (1) powershell (1) open-source (1) voice-assistant (1) assistant-retirement (1) consumer-ai (1) microsoft (1) replaced-by-successor (1) ai-companion (1) consumer-app (1) personal-memory (1) emotional-design (1) platform (1) plugins (1) ecosystem (1) superseded (1) image-generation (1) model-lifecycle (1)
Every archive entry separates what a source states from what the curator infers. Status is recorded only when a source supports it; when a reason is not public, the entry says so.