Bookmarks: Week of Aug 10 - Aug 16, 2026

Gray Partridge Archives - Birds Calgary

birdscalgary.com

game birds who live in the ravine behind our house. I always thought they were grouse. who knew?

Tags: birds, calgary, via:Maria Hoiss

Taalas - LLM as hardware

taalas.com

AI models are changing the world. As software, they are extremely computationally demanding. A 1000x improvement in efficiency is needed. This goal is far out of reach of our best general-purpose computers.

The company that made chatjimmy.ai - just, of course, acquired by AMD.

Tags: AI, LLM, hardware, chips

chat jimmy

chatjimmy.ai

a chatbot running on an LLM that is implemented in silicon. A hardware LLM. Holy moly it’s fast. FAST. So fast that the response is displayed before I look up after hitting RETURN. So fast that I sit there staring at it, assuming it hasn’t happened yet, until I realize I’m looking at the response rather than my prompt (or the previous response). FAST.

Imagine, if something like Fable could fit on a chip and be run as hardware…

Tags: via:Craig Mod, LLM, AI

Claude’s new Scarlet Letter watermark is invisible—for now - Ars Technica

arstechnica.com

The mark flags anything Claude processed, even human writing it only edited.

Selected text

The approach described by the EU and implemented by Anthropic is unfortunately trivially easy for bad actors to bypass, while potentially punishing users who trust the system to accurately label their outputs. Text watermarks work by biasing the model’s word choices in a pattern spread across the entire document, only detectable in aggregate by the right tool. The catch is that “invisible” can also mean the model occasionally trades the best word for a slightly worse one, just to keep the signal intact.

Tags: Ars Technica, Claude, AI, watermark, person:Ashley Belanger

5 things instructors should know about generative AI at McGill – Teaching for Learning @ McGill University

teachingblog.mcgill.ca

As you prepare for the start of the Fall term, you might be considering how generative AI (gen AI) fits into your teaching and course design. To help you make informed decisions, here are five things you should know about using gen AI at McGill, including opportunities to connect with Teaching and Academic Programs (TAP) […]

Tags: McGill University, AI

5 Rules for a Second Brain You’ll Actually Use

joanwestenberg.com

Outdated and duplicated, vague and unused notes just create huge amounts of cognitive noise, and they turn your “PKM” into a hulking mass that gives you enormous anxiety and actively discourages you from poking at it. My attempt to counter this has been to actually spend at least an hour a week pruning - reviewing notes, removing and merging items, and either processing or deleting at least one object for every five I add. I aggressively delete anything my brain marks as “someday useful.”

If there’s a helpful analogy, it’s probably to think about your own knowledge as tending to a garden, rather than filling a warehouse.

connects to my recent post on pruning my obsidian vault and deleting the automated notes

Tags: person:Joan Westenberg, notetaking, second brain

How Claude marks AI-generated content | Claude Help Center

support.claude.com

Claude uses two complementary techniques to mark content generated and processed by Claude: (1) watermarks embedded in text, and (2) signed provenance metadata attached to files.

but their description of “watermarking” in plaintext sounds kind of handwavium-ish. Is it invisible characters distributed throughout the text (doing creative things with UTF-8 or something), or is it a form of text-based steganography that subtly modifies the actual text itself to include a pattern that it can recognize? em-dashes, special filler words that start creeping in? something else? “Deep dive” and “delve” as AI-tells-turned-watermarks?

Tags: Claude, Anthropic, watermark

Student’s Perception of UX Design on an LMS Course: Designing Online Courses Students Can Actually Navigate - Faculty Focus | Higher Ed Teaching & Learning

facultyfocus.com

Learn how effective LMS course design and UX principles improve online learning by enhancing navigation, reducing cognitive load, and supporting student success.

Tags: Faculty Focus, person:Mark Savignano, person:Scott Page, UX, design, LMS, online learning

Anthropic puts hidden watermarks on Claude text under new EU rules

interestingengineering.com

New Claude models will embed invisible machine-readable watermarks under Anthropic’s plans for complying with EU AI Act transparency rules.

Tags: Claude, Anthropic, watermark, via:Lorelei Anselmo

Post Secondary Mini-Cases to Apply the HUMAN Framework

docs.google.com

When AI Sounds Human Postsecondary and industry cases for applying the HUMAN Framework Case P1: The First-Year Writing Coach The situation A university introduces an AI writing coach in several large first-year courses. The coach remembers each student’s earlier drafts, identifies recurring writ…

Tags: AI, person:Nadia Delanoy, framework