Vibe coding, the practice of using AI language models to write software with minimal manual coding, is drawing scrutiny after Georgia Tech researchers linked dozens of security vulnerabilities directly to AI-generated code in just a three-month sample. Despite these risks, a survey of over 1,100 professional programmers found that 72 percent use AI coding tools daily, with AI-generated or -assisted code making up 42 percent of their codebases and expected to exceed half by next year.
engadget.com
· 2026-09-06
A software developer published a personal essay cataloguing mixed emotions about AI: awe at how neural networks produce reasoning without explicit algorithms, fear over unresolved existential-risk arguments, disgust at crawlers overwhelming open web communities, sadness for artists undercut by AI-generated content, and anger at political systems failing to check corporate AI power and resource consumption. The author concludes AI feels promising as technology but troubling for society.
beza1e1.tuxen.de
· 2026-09-06
In a LinkedIn post republished on his blog, engineer Bryan Cantrill criticizes the growing use of LLMs to write LinkedIn content, arguing the results are stylistically obvious and often poor quality. He points to telltale patterns like excessive emojis, choppy paragraphs, and 'not just...but also' phrasing as giveaways of AI authorship. Cantrill argues LinkedIn's own AI rewrite prompts encourage this behavior, worsening the platform's content quality.
bcantrill.dtrace.org
· 2026-09-06
A tech commentator recounts waking up with the observation that modern AI systems like GPT 5.6 Pro and Fable can discuss self-reference and Gödel's Theorem fluently, yet nothing about self-referentiality was deliberately engineered into their transformer architecture, training process, or hardware. The author contrasts this with Douglas Hofstadter's Gödel Escher Bach and Roger Penrose's The Emperor's New Mind, both of which treated self-reference as central to whether machine intelligence could ever emerge. The piece notes these self-referential capabilities simply emerged as a byproduct of general pretraining, alongside unrelated knowledge like Pokémon trivia or plate tectonics.
scottaaronson.blog
· 2026-09-02
Academa is a new project that treats lecture videos like software: instructors describe what a teacher says and draws using a code-like script, which a compiler then renders into a video with text-to-speech narration and generated graphics. Because the lecture exists as editable source rather than a finished video file, mistakes can be patched and new versions shipped, much like updating a codebase. The team is now applying large language models to this text-based format, aiming to let AI generate or refine entire STEM lectures automatically.
academa.ai
· 2026-08-30
A research paper compiles evidence that widespread adoption of large language models is homogenizing written language, drawing on decades of linguistics and computational research into personality, gender, regional and cultural markers in text. The authors argue that as LLMs become the dominant tool for writing assistance, they risk flattening distinctive dialects, sociolects and individual voice patterns documented in prior sociolinguistic studies.
nature.com
· 2026-08-30
A new guide explains how everyday users can download and run large language models like those from Meta and Google directly on their own Windows, macOS, or Linux machines instead of relying on cloud-based apps like ChatGPT. Running LLMs locally offers offline access and greater privacy since no data is sent to outside servers, and it avoids subscription fees, though these free models are generally less powerful and slower than paid cloud versions. The guide notes that macOS is often preferred due to Apple Silicon's unified memory architecture, and that at least 16GB of RAM is recommended for smooth performance.
wired.com
· 2026-08-29
Researchers from Google and the University of Illinois Urbana-Champaign tested large language models on variations of the classic Knights and Knaves puzzle, where truth-tellers and liars must be identified from their statements. The models frequently failed when puzzles were slightly altered from familiar training patterns, defaulting to memorized answers instead of reasoning through the new version. A similar pattern showed up on a benchmark called SimpleBench, where humans easily spot subtle twists that trip up even leading AI systems.
technologyreview.com
· 2026-08-26
A commentary piece argues that AI coding assistants primarily benefit developers with years of pre-AI experience, since expertise built through manual coding provides the judgment needed to supervise and correct AI output. Newer developers, by contrast, are being pushed to rely on these tools without having built the foundational skills required to use them safely and effectively, creating what the author calls a paradox where novices need expert-level skills to keep pace.
larsfaye.com
· 2026-08-24
An opinion piece argues that a 17-year-old today should focus on building large language models from scratch rather than just using them, in order to understand their capabilities and limits. The author notes LLMs excel at tasks like writing essays but fail at physical, real-world tasks like cleaning a room, and suggests pursuing college and graduate study aimed at closing that gap. The piece frames this as a path toward developing new architectures beyond current LLMs for physical-world learning.
twitter.com
· 2026-08-23
GoKawiil