An AI-detection tool called Pangram flagged an August Washington Post op-ed by Dartmouth provost Santiago Schnell, which criticized AI cheating, as likely LLM-generated, according to a Semafor report. A follow-up investigation by student paper The Dartmouth found the same tool identified nine other recent works by Schnell as largely AI-written. Schnell confirmed to The Dartmouth that he uses AI for 'language refinement, copyediting, improving clarity and organization,' while maintaining he develops the ideas, research and structure himself.
thefloatingcontinent.com
· 2026-10-16
The PostgreSQL Global Development Group has delayed PostgreSQL 19 from its planned late September/October launch to around November 19, cutting between eight and ten planned features. EDB's Peter Eisentraut, a core PostgreSQL committer, said many of the dropped features had architectural flaws and that newly adopted LLM-assisted code review surfaced a long list of smaller edge-case defects that developers couldn't fix in time for beta.
theregister.com
· 2026-10-16
Torvalds posted a ten-patch series that improves folio lookup by marking 'hot' folios for delayed release under an RCU grace period, allowing both the lookup and the actual data access to happen within an RCU-protected region without taking a refcount on the fast path. The change reportedly yields an order-of-magnitude speedup on a read()-heavy benchmark, though reviewers have already found a race condition requiring further fixes.
lwn.net
· 2026-10-14
Simon Willison analyzed outputs from his recurring 'generate an SVG of a pelican riding a bicycle' benchmark test, noting that SVGs produced by different large language models tend to converge on similar visual characteristics. He compiled original SVGs from his test runs using each model's highest reasoning setting that returned a valid SVG.
arcfu.com
· 2026-10-14
Chris Siebenmann, who runs the Wandering Thoughts blog and CSpace wiki, says he is now attempting to block requests from clients whose HTTP User-Agent header is too generic or suspicious. He attributes the move to a surge in high-volume crawlers, some reportedly gathering data for LLM training, that strain his site's resources.
utcc.utoronto.ca
· 2026-10-14
A developer writing about AI-assisted coding argues that the non-deterministic nature of large language model output, long seen as a barrier to reliable software development, is not actually a problem. The author explains that code is simply one possible output among many valid paths to a correct result, and that LLMs function as actors within a system rather than as a replacement for precise, deterministic instructions.
sklar.rocks
· 2026-10-14
A developer has published a tutorial on Hacker News that walks through building a large language model from scratch using Node.js and TypeScript. The guide aims to teach the underlying mechanics of LLMs rather than relying on existing Python-based frameworks.
book.humanexperiences.io
· 2026-10-14
Consus, a startup from Y Combinator's Fall 2024 batch, has launched an LLM routing service that directs AI requests through infrastructure compliant with ITAR and controlled unclassified information (CUI) rules. The company says it has completed a FedRAMP High equivalency audit covering 410 controls, the U.S. government's strictest cloud security baseline, and can integrate with existing tools like ChatGPT, Claude and open-source AI platforms. Founder Eric previously worked at SpaceX and Castelion in aerospace and defense roles.
news.ycombinator.com
· 2026-10-13
A Hacker News 'Ask HN' post from a hobbyist woodworker who currently uses Shapr3d to manually design projects asks whether large language models can speed up 3D modeling for woodworking. The poster is soliciting recommendations from the community for AI tools or workflows suited to this use case.
news.ycombinator.com
· 2026-10-13
A project called Janas-LLM has published benchmarks showing it can run large mixture-of-experts models, including a 48GB Qwen3-Next-80B model, on a 32GB laptop by streaming model 'experts' from an NVMe SSD rather than keeping everything in RAM. Tests on an Intel Core Ultra 9 185H laptop with its integrated Arc GPU showed Janas-LLM generating text at 24 tokens/sec versus llama.cpp's 2.8 tokens/sec on that model, with similar gains across a dozen other open models from Qwen and Gemma families. The developers released a benchmarking tool, janas-bench, and installation instructions for Linux x86-64 so others can reproduce results on their own hardware.
github.com
· 2026-10-12
In an October 12, 2026 post, writer Huahai responds to Rich Hickey's essay arguing that LLMs count as automation rather than tools because they can't be 'wielded.' Huahai counters that LLMs can be directed, corrected, and refined by users just like saws, chisels, compilers, or spreadsheets, and that unpredictability doesn't disqualify something from being a tool. The post argues that using an LLM well, through iterative feedback and judgment, requires genuine skill.
yyhh.org
· 2026-10-12
Andrej Karpathy argues that public understanding of large language models' capabilities is diverging sharply rather than converging. He outlines a rough tier system: roughly 75% of people have barely used LLMs, 20% use free chatbots casually, about 0.2% of professionals in coding and math see agents completing weeks of work in a prompt, and an estimated 5,000 people with internal access to frontier systems are reportedly seeing swarms of agents handle multi-week software projects, cyberattacks and research-level math.
twitter.com
· 2026-10-11
GoKawiil