A developer whose Git branching diagram became widely used across the software industry has accused Microsoft of publishing an unattributed, AI-generated imitation of his work on its Learn portal.
Vincent Driessen described the disputed image in a post marking more than 15 years since he published his article, *A successful Git branching model*. He said he designed the original diagram in Apple Keynote in 2010, carefully arranging its colours, curves and branch lanes to show how lines of development relate over time. He also made the source file available so that other people could adapt it.
According to Driessen, the visual subsequently appeared in books, presentations, team documentation and videos. He said that broad reuse did not trouble him because sharing the idea was the point. His objection to the Microsoft Learn image was not simply that it resembled his design, but that it allegedly reproduced its overall structure while removing attribution and degrading its meaning. He distinguished this case from legitimate adaptation: in his telling, the later asset neither acknowledged its source nor improved on it, despite appearing on an official developer-learning service.
Driessen said people alerted him to the image through Bluesky and Hacker News. He pointed to malformed wording, including the phrase “continvoucly morged,” along with missing or misdirected arrows. In his account, those defects indicated that an image generator had been used to rework the diagram. He argued that the result retained enough of the original composition to remain recognisable while losing the precision that made it useful.
The post does not establish who created the Microsoft image, which tools were involved or what review process preceded publication. It also does not include a response from Microsoft. Driessen’s description should therefore be read as the creator’s documented allegation, not an independently confirmed account of Microsoft’s internal process or intent.
His broader concern was about accountability in automated content production. Well-known source material and conspicuous errors made this example easy for online observers to identify, he wrote, but future derivatives could be harder to trace as generation systems improve. The episode, in his view, illustrates how automation can obscure provenance while reducing the quality of a carefully designed teaching aid.
Driessen said the immediate remedy he wanted was modest: a link and credit to the original article. He also called for an explanation of how the Microsoft Learn page and its image were produced, what the page was intended to achieve and why the asset apparently passed without adequate proofreading. Those questions remained unanswered in the supplied account.
