NowRift — Software & Technology News Platform

Data ScienceRIFT LEAD

Terminal-Bench-Science: New Benchmark for AI Agents in Scientific Research

A new open-source benchmark, Terminal-Bench-Science, aims to rigorously evaluate AI agents on complex scientific research tasks.

28 Aug 2026

Newsletter

THE DAILY RIFT

Five shifts. Five minutes. No noise.

•Verified technology intelligence
•Emerging patterns before the hype
•Direct action steps for builders & founders

No spam. Unsubscribe anytime. Powered by Beehiiv.

All Intelligence AI Models Developers Cybersecurity Hardware Founders Creators Data Science

Today's Intelligence

Data ScienceAug 28

Google Lens Visual Search Hits 20 Billion Monthly Searches

Google Lens' massive scale signals a seismic shift, making visual and product data critical for online discovery and SEO.

●Builder Action:Developers must ensure their web applications serve high-quality, well-labeled images. Implement structured data markup (Schema.org) for products and other visual content to aid Google Lens in identifying and ranking your assets. Optimize image alt text and filenames with relevant keywords. Consider how your application's visual content can be queried and understood by machine learning models.

Data ScienceAug 28

575k crop labels from 10 years of book digitization beat AI

Manual Photoshop decisions, not more data or complex models, proved essential for automating Urdu book digitization.

●Builder Action:Developers should reconsider complex AI models for tasks with strong human heuristic components. Instead of training large models on massive datasets, explore minimal annotation strategies that capture operator-specific offsets or styles. This could involve training lightweight models or even using rule-based systems informed by a few key points per document. Consider implementing interactive tools that guide users to provide just a few critical points per book, allowing for rapid, accurate cropping. Benchmark against simple baselines derived from human workflows.

Data ScienceAug 27

Bag-of-Words and Text Vectorization: Turning Text into Numbers

Understand how computers process language by converting text into numerical vectors using Bag-of-Words and other text vectorization techniques.

●Builder Action:Developers can implement Bag-of-Words (BoW) using libraries like scikit-learn's `CountVectorizer` to transform text data into numerical vectors. Understanding BoW highlights the need for preprocessing steps like stopword removal and stemming, which were discussed previously. For more nuanced applications, explore TF-IDF weighting or word embeddings for richer semantic representations.