A recommendation is useless if the user has already clicked away. ⏱️ That's why Databricks and @Redisinc have partnered to deliver in-session personalization by combining Databricks Real-Time Mode (RTM) for continuous processing with Redis for sub-millisecond serving, without managing a separate streaming engine. Explore the joint architecture, performance benchmarks, and code snippets:

Account recommendation of the day - please follow @jwt0625 for always striving to push outside five sigma and questioning the status quo! I may not agree with him on everything, but it is worth hearing out his frequent hot takes based on technical knowledge.

Our recommendations offer a broad spectrum of storytelling. A documentary about lions is just as entertaining as a famous cartoon, and a fictional tale documents drug addiction in graphic detail

In generative recommendation, everyone has assumed they know what makes a "good" semantic ID design. So what happens when you actually put that assumption to a large, controlled reproducibility test? 🔍 Generative recommenders turn items into discrete token sequences called semantic IDs, then generate the next item autoregressively. Designs like RQ-VAE, OPQ, and RQ-Kmeans have proliferated, but prior comparisons used different datasets and backbones, so nobody could say which design actually wins. So the authors reran 12 methods side by side under the exact same data splits and evaluation protocol, and the assumptions started falling apart. No single SID design dominates across datasets. Codebook utilization balance barely correlates with recommendation quality. Longer codes don't always help, and bigger backbones can even hurt performance. And when you measure how well an SID preserves an item's local semantic neighborhood, the winning method flips depending on whether you use Jaccard or RBO. The lesson from What Makes a Good Semantic ID for Generative Recommendation? A Reproducibility Study is that good SID design isn't about chasing one metric, it's about balancing several at once. URL: #RecSys# #Reproducibility#

.@benthompson gives his recommendation on how Meta should market Muse: "The way to frame this is to say: 'We're not helping you be more productive, we're helping you get rid of annoying crap in your life.'" "It needs to be framed more as a pain relief as opposed to this aspirational, 'I'm gonna be so productive and get things done.'" "People don't care about being productive. Employers care about their employees being productive. But once you're off the job, you just want to go home and watch TikTok." "The way I think this is going to land is through Instagram Reels or TikTok with people actually showing stuff they do and people say, 'That's amazing. That would make my life so much easier.'" "It really is a show-not-tell sort of technology."