Post

HN Top 10 β€” October 16, 2026

HN Top 10 β€” October 16, 2026

πŸ“Š View styled HTML version

Today’s Top 10 on Hacker News

1. Yet another way to invoke functions without parentheses

⭐ 30 πŸ’¬ 8 πŸ‘€ aragonite πŸ”— Discuss on HN

The article describes a JavaScript technique for invoking arbitrary functions without parentheses by overriding Error.prepareStackTrace so that accessing an error’s stack creates a function from the stack trace. It then makes the stack array stringify into attacker-controlled code by replacing Array.prototype.toString or using prototype pollution, and finally triggers execution through valueOf or string concatenation.

2. Meta’s new commercial terms will allow the scraping of news sites: experts

⭐ 318 πŸ’¬ 115 πŸ‘€ 0in πŸ”— Discuss on HN

Meta’s updated commercial terms would permit the company to use web scrapers to collect content from websites linked to its business accounts, including news publishers. Legal experts say the language could allow Meta to use that material to train its AI products, and publishers currently have no way to opt out without giving up access to Meta’s business tools.

3. Woman who rescued injured crow keeps getting β€˜thank-you gifts’ from birds [video]

⭐ 359 πŸ’¬ 121 πŸ‘€ johnnyApplePRNG πŸ”— Discuss on HN

A video tells the story of a woman who rescued an injured crow and has since been receiving small objects from birds, apparently as tokens of gratitude. The Hacker News post likely discusses the video and whether this behavior suggests crows can recognize and reciprocate toward humans.

4. Muse Spark Open

⭐ 827 πŸ’¬ 415 πŸ‘€ bigwheels πŸ”— Discuss on HN

Muse Spark Open is an open-weight, natively multimodal AI model with 816 billion total parameters and a 1 million-token context window, designed for coding, agentic workflows, and complex multi-step reasoning. It is notable for running on desktop hardware using two Apple M5 Ultra Mac Studios with the MLX framework, while also being available through major API platforms and model repositories. The release includes optimized inference support, safety evaluations, and a license that allows many developers to use the model for free.

5. The Economic Complexity of the Western Roman Empire

⭐ 26 πŸ’¬ 7 πŸ‘€ mikk14 πŸ”— Discuss on HN

Researchers used a large dataset of Latin inscriptions to estimate the economic complexity of the Western Roman Empire by identifying the professions recorded on funerary and honorific texts and mapping them to provinces. They corrected for biases such as status exaggeration, underrepresentation of women and farmers, and uneven archaeological recovery. The resulting analysis reveals how diverse economic activities were distributed across the empire and provides a new way to study ancient economic development.

6. Two-Stack Sliding-Window Aggregation

⭐ 33 πŸ’¬ 1 πŸ‘€ surprisetalk πŸ”— Discuss on HN

The article explains how to maintain sliding-window aggregates efficiently when the summary operation has no inverse, such as minimum, quantiles, or approximate unique counts. It presents a simple two-stack algorithm that achieves amortized constant-time updates for a broad class of associative aggregations by abstracting them into empty, unit, combine, and finalize functions. The post also traces the algorithm’s origins to earlier queue implementations and a more advanced low-latency paper.

7. RPKI now protects 70% of internet routes

⭐ 168 πŸ’¬ 57 πŸ‘€ emot πŸ”— Discuss on HN

RPKI, which lets IP address holders sign which networks may announce their routes, now covers 70.3% of routed prefixes, up sharply from 54% a year earlier, with China accounting for most of the increase. The system is visibly working: invalid BGP routes are seen by only a handful of monitoring points, while valid and unvalidated routes propagate widely. This suggests RPKI is reducing the risk of accidental or malicious route hijacks without adding meaningful instability to Internet routing.

8. The insider trading you’re allowed to do

⭐ 50 πŸ’¬ 30 πŸ‘€ rohanm93 πŸ”— Discuss on HN

The article compares legal career β€œinsider trading” to illegal stock trading, arguing that employees often know more about their company’s strengths, weaknesses, and opportunities than outsiders do. It urges people to act on that private knowledge by moving toward growing areas, building new skills, or targeting jobs where real opportunities exist rather than waiting for public confirmation.

9. Neural networks are clean

⭐ 705 πŸ’¬ 266 πŸ‘€ themgt πŸ”— Discuss on HN

The author argues that neural networks are far more interpretable than previously believed, because apparent polysemanticity and superposition were largely misread as arbitrary concept packing. Instead, what looked like tangled neurons often reflects ordinary learning failures or recursive generalization, and modern architectural improvements make even deep layers broadly legible. This suggests mechanistic interpretability can study networks directly without first extracting or unfolding hidden features.

10. The Folders Programming Language

⭐ 84 πŸ’¬ 11 πŸ‘€ gurst πŸ”— Discuss on HN

Folders is an esoteric programming language that represents programs as nested directory structures, using only the containment and alphabetical order of folders rather than file names or text. Inspired by Whitespace, it treats empty folders as code and can store temporary data in folder names, offering a conceptual model of dematerialized code and potentially unlimited storage. An emulator and interpreter allow users to run Folders programs by loading directories and executing commands based on folder counts.

This post is licensed under CC BY 4.0 by the author.