Published on January 10, 2026 at 06:01 CET (UTC+1)
“Erdos problem #728 was solved more or less autonomously by AI” (326 points by cod1r)
Mathematician Terence Tao reports that a longstanding Erdős mathematical problem (#728) has been largely solved by AI tools. The AI worked autonomously or with minimal human guidance to find a solution. This represents a significant milestone in the application of artificial intelligence to pure mathematical research.
You probably don't need Oh My Zsh (11 points by fla)
This article argues that the popular shell framework Oh My Zsh adds unnecessary bloat and significantly slows down shell startup time. The author advocates for a minimal, manually configured Zsh setup instead, providing a starter configuration to improve performance, especially for users who open many terminal sessions daily.
OLED Not for Me (17 points by c0nsumer)
The author shares a personal experience with a QD-OLED monitor, detailing initial excitement followed by visual discomfort and perceived clarity issues during normal use. Despite the monitor's high contrast and sharp specs, the author ultimately decided to return it within the 90-day window, questioning the suitability of current OLED technology for their productivity-focused workflow.
Maine's black market for baby eels (15 points by noleary)
This article describes how Maine's illegal baby eel (elver) fishing trade, a high-stakes black market, has inspired a new subgenre of crime thriller novels. It focuses on a specific novel, "The Glass Eel," written by a father and son duo, using it as an example of how this unique regional crime story is capturing the imagination of writers and readers.
Flock Hardcoded the Password for America's Surveillance Infrastructure 53 Times (352 points by fuck_flock)
A security researcher discovered that Flock Safety, a major surveillance technology company, hardcoded a default API key in 53 public-facing instances. This key granted unrestricted access to sensitive real-time data layers, including license plate reads, drone telemetry, and police locations from thousands of deployments nationwide, representing a massive privacy and security failure until it was remediated.
JavaScript Demos in 140 Characters (210 points by themanmaran)
Dwitter is a creative coding platform that challenges users to write compelling visual or audio demos using JavaScript, with the severe constraint of fitting the entire code within 140 characters. It showcases the artistry and cleverness possible within extreme limitations, serving as both a technical challenge and an entertainment site.
RTX 5090 and Raspberry Pi: Can it game? (185 points by scottjg)
This is an experimental hardware blog post testing whether a high-end RTX 5090 GPU can be used for gaming when attached via an external enclosure to low-power single-board computers like a Raspberry Pi 5. It compares gaming performance across the Pi and other small computers, highlighting the severe bottleneck created by the limited PCIe bandwidth on these devices despite the powerful GPU.
Greenland sharks maintain vision for centuries through DNA repair mechanism (36 points by pseudolus)
New research reveals that Greenland sharks, the longest-lived vertebrates, maintain functional vision for centuries despite their age and parasitic eye infestations. Scientists discovered a unique DNA repair mechanism in their eye tissue that prevents the accumulation of damage, challenging previous assumptions about their blindness and offering insights into aging and longevity.
How Markdown took over the world (188 points by zdw)
Anil Dash chronicles the unlikely rise of Markdown, a simple plain-text formatting syntax, from a niche tool for bloggers to a ubiquitous standard powering documentation, notes, and even AI system prompts. He credits its success to its simplicity, readability, and open nature, representing a grassroots, human-centric approach to technology that scaled globally.
How will the miracle happen today? (396 points by zdw)
Kevin Kelly muses on the nature of technological and scientific progress, arguing that "miracles" (major breakthroughs) are a regular, almost predictable occurrence in the modern era. He suggests we should actively anticipate and look for these transformative advances daily, framing progress as a constant unfolding of unexpected solutions rather than a linear path.
AI as an Autonomous Research Partner in Hard Sciences: The solution to Erdős problem #728 by AI demonstrates a shift from AI as a mere analysis tool to an autonomous agent capable of discovery in structured, logic-based domains like pure mathematics. This matters because it opens new frontiers for accelerating research in fields with well-defined rules. The implication is the need for new interdisciplinary roles (e.g., "AI-guided research scientist") and frameworks to validate and interpret AI-generated discoveries.
The Primacy of Data Infrastructure and Security: The Flock Safety breach is a stark reminder that the most advanced AI/ML systems are built on data pipelines that are often vulnerable. As AI integrates into critical infrastructure (surveillance, healthcare, finance), the security of the data supply chain becomes paramount. The takeaway is that AI development must be paired with rigorous "AI supply chain" security audits; a sophisticated model is only as strong as the weakest link in its data ingestion process.
The Democratization of AI Development Through Simplification: The pervasive use of Markdown as a lingua franca for AI prompts (e.g., in ChatGPT) highlights a trend towards simplifying human-AI interaction. Complex instructions no longer require formal programming languages. This lowers the barrier to entry, allowing domain experts without coding skills to leverage AI. The implication is that future AI tool success will heavily depend on intuitive, natural interfaces and robust prompt interpretation.
Convergence of AI and Biological Inspiration (Bio-AI): The discovery of DNA repair mechanisms in Greenland sharks that preserve function over centuries provides a novel blueprint for AI system design, particularly around longevity and resilience. This matters for developing more robust, self-repairing, and energy-efficient AI systems and algorithms. It suggests a growing trend where biological research directly inspires new AI architectures and maintenance paradigms, moving beyond neural networks based solely on simplistic brain models.
The Expectation of Continuous Miraculous Breakthroughs: Kevin Kelly's framing of daily "miracles" sets a cultural expectation within the tech and AI community that disruptive progress is the norm. This influences investment, research focus, and public perception, creating a pressure to deliver transformative results. A key takeaway for developers is to foster a mindset of anticipating and leveraging sudden, disruptive shifts in capability (e.g., new model architectures, scaling laws) rather than solely focusing on incremental optimization.
Creative and Experimental Benchmarking Pushes Hardware Limits: The experiment of pairing an RTX 5090 with a Raspberry Pi, while impractical, reflects a trend of using creative, extreme benchmarks to understand system bottlenecks and the practical limits of hardware for AI/ML workloads (like gaming or local model inference). This matters for edge AI and cost-effective deployment, highlighting the critical importance of balanced system architecture—where a powerful GPU can be nullified by poor CPU or I/O—guiding both hardware design and model optimization for specific platforms.
The Rise of Constrained and Efficient Coding for AI Tools: Platforms like Dwitter, though artistic, reflect a broader emphasis on efficiency and creativity under constraints, a skill highly relevant for optimizing AI model inference, writing efficient data processing scripts, or deploying models in resource-limited environments. The trend points to a growing appreciation for elegant, minimal code that does a lot with a little, which is directly applicable to building leaner, faster AI pipelines and micro-services.
Analysis generated by deepseek-reasoner