Henry Farrell argues that social media's real danger isn't fake news but algorithmic distortion of how we perceive other people and public opinion. When journalists and political figures rely on Twitter's skewed representation of the public to inform their work, these distorted views get reinforced across institutions and industries that shape culture.

• The main danger of social media isn't fake news brainwashing people—it's algorithms distorting our perception of what others actually think, warping our mental map of society.
• Journalists, academics, and political staffers are especially exposed to Twitter's skewed picture of public opinion, then embed that distortion into their reporting and analysis, amplifying it society-wide.
• Entertainment and other industries use similar algorithmic sorting, creating the same kind of distorted signals about audience preferences—so the problem isn't unique to political discourse.
• This creates a feedback loop where the institutions meant to help us understand society are all working from the same algorithmically-mangled data, undermining the ability to solve collective problems.

This article discusses a library of stochastic streaming algorithms designed for fast approximate analysis of big data. It highlights the library's ability to handle complex queries efficiently, reducing processing times significantly while maintaining mathematically proven error bounds. Adaptors for various platforms and languages are included to facilitate integration.

• Yahoo cut data processing times from days/hours down to minutes or seconds using these sketch algorithms.
• The library provides sketches (approximate streaming algorithms) with mathematically proven error bounds for queries like count distinct, quantiles, and graph analysis.
• It integrates with Apache Hive, PostgreSQL, and Google BigQuery, and supports Java, C++, Python, Rust, and Go.
• Built-in Theta Sketch set operators allow complex set expression calculations, outperforming traditional Include/Exclude methods.

This article explores the challenges of performing exact queries on large datasets and introduces data sketches as a solution. Sketches provide approximate answers quickly and efficiently, allowing for scalable data analysis without the need for massive storage. The piece outlines how these probabilistic structures work and their advantages in handling big data.

• Exact COUNT DISTINCT on billions of daily events can turn a simple task into a multi-hour or multi-day query nightmare
• Data sketches like HyperLogLog trade a bit of precision for massive speed and memory savings by hashing and summarizing data instead of storing every unique value
• The technique traces back to Philippe Flajolet's 1980s streaming algorithm research
• Spark and BigQuery already have sketch-based functions built in, so engineers can use them without extra libraries

Recent research indicates that building a quantum computer capable of breaking elliptic-curve cryptography (ECC) is becoming easier and faster than previously thought. One study shows it could crack ECC in just 10 days, while another demonstrates breaking ECC-secured blockchains in under nine minutes. Both papers highlight significant progress in quantum computing capabilities, though neither has been peer-reviewed.

• A neutral-atom qubit approach could crack 256-bit ECC in 10 days using 100x less overhead than prior estimates.
• A Google team claims to break ECC securing Bitcoin-like cryptocurrencies in under nine minutes, a 20-fold resource reduction.
• Neither whitepaper has been peer-reviewed, so the claims remain unverified.
• Experts say the results don't pin down a timeline for practical quantum decryption, but confirm steady, unslowing progress toward it.

This article explores an unconventional method for classifying text by leveraging compression algorithms. The author demonstrates how to concatenate labeled documents, compress them, and use the compressed sizes to predict labels for new texts. While the method shows promise, it is computationally expensive and generally underperforms compared to traditional classifiers.

• Compression-based classification (concatenate labeled texts, measure compressed-size increase) hits a 0.749 macro F1 with gzip on 4 categories of 20 Newsgroups, versus 0.88 for multinomial Naive Bayes.
• lzma pushes accuracy up to 0.897, beating Naive Bayes, but takes 32 minutes versus over 5 minutes for gzip on just 1,353 test cases—far too slow to be practical.
• The technique reframes text classification as an information-theory problem, showing compression algorithms implicitly model word probability distributions.