Looking forward to when Claude projects become the equivalent of repos for skill environments, with the ability to synchronize skill files and project knowledge without having to perform manual downloads/uploads, ie not burning tokens, and also do the equivalent of adding notes to Claude.md

It’s just ridiculous how many sad lonely man-children are rushing to create their own version of Samantha

18th century agentic RAG

Claude + Skills perform admirably on @simonwillison.net‘s Pelican on a Bicycle test, by generating an svg then visually inspecting it and iterating until satisfied. Could probably tweak the skill to be more critical: this was its 3rd attempt claude.ai/share/d94dfb...

I may have missed it but the computer as a service implications of Claude Skills seem to not have received the deserved attention yet: Eg the Python environment (currently) has 184 different libraries your skill scripts can import claude.ai/public/artif...

I don’t know when it happened but noticed Claude can now move conversations between projects and the project instruction is changed with the move. This is great for “manual” multi-agent like work (use 2 different ICL experts, potentially armed with different project RAG) in 2 turns of same convo.

Behold my first Claude Skill -- a @preactjs.com Preact Developer skill github.com/oaustegard/c... Very first attempt, asking Claude to create something awesome: austegard.com/fun-and-game...

I -- ok Claude -- created a browser extension that shows your current Claude usage as a handy gauge in your extensions bar github.com/oaustegard/c...

CLawd (and claude.ai) now has Skills -- including an animated Gif creator -- SKILLS transcript here: gist.github.com/oaustegard/9...

If you ever need programmatic access to anything in the claude.ai interface... chances are you'll find it _somewhere_ in the massive claude.ai/api/bootstrap response -- looks like most of the interface is driven from that one file!

Sonnet 4.5 on different prompting strategies between it and Haiku 4.5: claude.ai/public/artif...

Kudos again to the @prefect.io crew on doing documentation right for the AI era. I love everything about this -- mcp server, llms.txt and copy as markdown shortcut behavior

How might we coax an LLM, like Claude to more readily consider cross-domain parallels to a given concept? I generated a userStyle that can be applied in a two-turn exchange with Claude seeking to broaden its search before synthesizing its answer: (README in comment) gist.github.com/oaustegard/2...

Another reason to ban golf ⛳️

jules.google has been pretty handy in documenting all my independent tool scripts, but it has an odd minor bug in its PR code review design: It encourages users to mention it by @jules -- problem is GitHub resolves that as the actual user github.com/jules ... not associated with jules.google

Another itch, another bookmarklet: The Claude Last Week's Conversations bookmarklet makes use of the underlying GET api callable from any Claude page to list your recent conversations, across your whole account, including each conversation's summary, and renders it in MD github.com/oaustegard/b...

3 years ago I was asking text-davinci-002 to use phonetic markup to entice it to create simple rhymes. Now, Sonnet 4.5 not only gets but can respond in style to a spatially structured poem like the below

It took Claude.ai a frustrating 13 tries to *recreate* this (just from the four lines of code below) though arguably 11 of those were duplicate ID/styling/color related more so than actually getting the WebGL and shader code right... Added scroll to zoom for fun: austegard.com/fun-and-game...

Hmmmm

My github-toc web component is becoming quite useful... Here listing inventory of two different repo paths, with different settings for include/exclude and link-prefix Source: github.com/oaustegard/o... README: github.com/oaustegard/o... Example: github.com/oaustegard/o... => austegard.com/biking/

Features missing from Claude Enterprise (@anthropic.com). I fear this will become a lengthy thread... - Enterpise users can't use Claude Code without a Premium license!! What? - A Team concept within the Enterprise. Surprising omission -- or is this the undocumented nested Organization feature?

Plausibly overkill, but here is a drag and drop instruction you can give jules.google to generate a brand new AGENTS.md file for your Github repository. (Iteratively generated with Claude, of course...) WOMM, YMMV gist.github.com/oaustegard/a... Example result: github.com/oaustegard/o...

With Sonnet 4.5 (and plausibly GPT-5 Thinking) my chart below is too pessimistic. My latest vibe-coding tool is 1436 lines of html/js/css and went through 56 internal Claude revisions with only two minor error cases -- all the other iterations was me dreaming up scope creep.

It’s cute how the Anthropic UX folks think 5 lines is an appropriate size for Project Instructions…

For the Claude Enterprise "owner" users out there -- I crafted a log viewer that lets you simply parse the zip formatted audit log file you can download via the management console: github.com/oaustegard/o...

jules.google doesn't get enough attention. It's really quite good, and nicely implemented, the way it runs in a vm with a clone of your project, thus able to locally test its contributions before issuing the PR. Not sure when they added memory functionality? I tend to use it for clean up/docs

You’d think communications professor at a prestigious business school could form an argument cogent enough not to be defeated by the very technology he derides: Point: www.dn.no/innlegg/kuns... Counterpoint: claude.ai/public/artif... #norge #ki

New from 404 Media: Apple just removed ICEBlock, the app for reporting sightings of ICE, from its App Store after DOJ pressure. ICEBlock's developer tells me "we are determined to fight this." "Capitulating to an authoritarian regime is never the right move." www.404media.co/iceblock-own...

Sonnet 4.5 is a hell of a web developer, even when just vibe-coding in Claude.ai; this took a few iterations, but only because I kept scope-creeping: every iteration worked as I had requested up to then: bsky.app/profile/aust...

App stores bending over to authorities and refusing unicorn tracking? 🦄 Good thing the web is unstoppable! Install this PWA straight from your browser to your Homescreen. Community-powered magical creature spotting 🌈✨with GPS🛰️and notifications 🔔 enabled! austegard.com/fun-and-game...

Holy hell. DHS just openly stated that its agents use “reasonable suspicion” to make ARRESTS. This is blatantly unconstitutional. The Supreme Court is clear that you need probable cause to arrest someone, as opposed to stop them to ask them a question or two. This is huge b/c…

🦀

Hello, fellow 5-hr shift workers

PSA! Absent a system prompt, both Sonnet 4 and Sonnet 4.5 believe they are actually Sonnet 3.5... I wonder what this signifies? Is it training data timing related -- that the current model at the time of training 4 was 3.5?

Hey at least now I can tell prior to hitting 98% token allowance usage...

I should create a hat that says You're absolutely right. Even Claude 4.5 tells me I am...

I'm not vibe coding, I'm flibbertigibbiting

Ok, this is fascinating. Claude 4.5's document creation tooling uses an external SKILL.md file for explicit instructions on how to use PowerPoint. Presumably the same for other office types. See gist.github.com/oaustegard/4...

PSA: The recent_chats tool in CLaude.ai is ok, but far from complete. It does not capture any context (file/fetch result, etc) or extended thinking by the model. If you truly want to continue a conversation you're better off copying manually or via a tool e.g Claude Pruner github.com/oaustegard/b...

Is there an active bug in Claude.ai and/or Opus 4.1? At times when Opus 4.1 is chosen it identifies itself as Sonnet 3.5, other times as Sonnet 4.0 (And here I was surprised at the superficial answers I was getting from Opus, compared to Sonnet) Paging @anthropic.com (is there anybody out there?)

I mean yes this is funny, but the fact that these unconstitutional stormtroopers are so directionless that they literally stand around waiting for random nonwhite delivery guys to bike past them is just pathetic. No mandate, no mission, no clue. Just racism.

Pardon the efficiency of my method of presentation: but what if @anthropic.com tweaked Claude Desktop into a recursive agentic framework? claude.ai/public/artif...

Scam ads are getting stranger and stranger

This is madness.

Taken just a few minutes ago outside the ICE facility in Portland that Trump claims is under siege. My message to Donald Trump is this: we don’t need you here. Stay the hell out of our city.

Again I don’t want government by AI but OMG is AI better than some in government: claude.ai/share/932dd6... 🍊👬Misrepresented 2014 Article: jamanetwork.com/journals/jam...

I spent the week at the police surveillance convention and let me tell you my biggest observation: The name of the game now is consolidating as much information as humanely possible from surveillance devices, the internet, other governmental data, and literally a million other places. 🧵

🤢