The recent modifications to Tumblr's tagging system focus on delivering more relevant suggestions drawn from a user's personal history and sitewide trends. Meanwhile, an experiment that displayed non-AI related tags prior to typing has been completely retracted due to user dissatisfaction. Staff members are now actively soliciting feedback to determine the future direction of pre-typing tag recommendations. Ideas under review include surfacing frequently used tags, tags from reblogged posts, followed tags, or related tags.
Key points
In-line tagging now emphasizes a mix of personal blog history and popular tags across Tum…: In-line tagging now emphasizes a mix of personal blog history and popular tags across Tumblr.
Community complaints successfully caused the rollback of an automated pre-typing tag reco…: Community complaints successfully caused the rollback of an automated pre-typing tag recommendation experiment.
Tumblr is currently seeking creator input on what pre-typing tag options should be implem…: Tumblr is currently seeking creator input on what pre-typing tag options should be implemented next.
Sources checked by our analysis pipeline: 1
I’ve been working on the suggestions you see when adding tags to a post, both when you open the tag box (before you type anything), and also after you start typing.
I recently launched a change for when you’re already typing, which should show more relevant tags used on your blog, and more relevant popular tags on Tumblr. Let me know how that feels!
For tag suggestions before you start typing, I had implemented suggesting tags related to your last added tag (which are not generated by AI, btw), but I’ve reverted the change based on feedback I saw on Tumblr posts and Support tickets. Thank you for sharing your experience.
Now, I’d love to hear, when creating or reblogging a post…
The project's rollout schedule faced disruptions that prevented it from hitting its targeted distribution window. Stakeholders and users tracking the major database software milestone saw the release window pushed back unexpectedly. The software ultimately reached general deployment only after an unforeseen postponement affected the standard cycle.
Unveiled at the Amazon Accelerate conference, a new no-code plugin links core Seller Central intelligence, advertising metrics, and third-party software into a single analytics environment. The system draws on conversion trends, catalog records, and profit margins to generate proactive operational recommendations for merchants. Built-in privacy controls ensure commercial data remains isolated from artificial intelligence training routines while letting users set granular access tiers.
Following discussions at APNIC 62 in Mumbai, leadership has expanded the review team to five members to revise foundational community guidelines. The initiative seeks to modernize the framework to ensure operational procedures remain inclusive, clear, and effective. Stakeholders will have multiple opportunities to debate draft recommendations before final adoption at APRICOT 2027.