This warms my heart.

That was a fun podcast! player.captivate.fm/episode/380b...

Where I explain what GitHub Agentic Workflows are, why I used them for docs, and what I've learned so far by producing them: passo.uno/agentic-work...

Where I explain what GitHub Agentic Workflows are, why I used them for docs, and what I've learned so far by producing them: passo.uno/agentic-work...

Whereof one cannot speak, thereof one must file a noop.

All hardships bear fruit of some kind. Not sweet, but certainly nourishing.

Yesterday, my daughter (4), bewildered by its rainbow hues, brought me a mysterious object that came with her book: a CD. I put it into an old reader and then music came out. It was a magical moment. In an age where wonders are losing their physicality, it felt like a wonderful outerworldly ritual.

Getting smart LLMs and then slowly losing them due to cost and infrastructure problems reminds me of the narrative arc from Flowers for Algernon, my favorite SF short story.

I'd love to use AI that thinks and speaks like a mildly annoyed John Cleese.

There's only one feature I miss from Twitter/X: the ability of muting only retweets (or boosts).

Something in this tooling inflation troubles me deeply. passo.uno/tools-slop-i...

I've put in place seven agentic workflows that run quality checks (style, coherence, typos, etc.) in our docs. They use a mix of deterministic (Vale, Lychee, Codespell) and non-deterministic approaches, as well as our public docs MCP server. Basically a mix of all I've used throughout the years!

Something in this tooling inflation troubles me deeply. passo.uno/tools-slop-i...

My avatar is one of those GenAI paintings from 2023, when the tech was still in diapers. It's vintage now.

Ok ok but what if APIs are the shared marriage rituals that facilitated the exogamy of our software.

I know many outstanding technical writers outside of the US. It just pains me to see how they're shunned because of that.

These days I'm using GPT 5.5 for most serious, no-nonsense work (and 5.4 and 5.3) and use Claude Opus for philosophical or architectural planning. And Gemini to create cute pictures for the kids.

🗓️ Over a year in the making 📈 One of the fastest-growing projects in #OpenTelemetry Introducing the Ecosystem Explorer The Ecosystem Explorer helps you understand OTel before you start. We’re starting with Java, with plenty more to come. Got ideas? Come build with us. buff.ly/EAfFJZg

Opus 4.7 is insufferably picky and contrarian when reviewing prose. That's what I'd call a bad editor.

If AI is going to destroy security, the future will be either air-gapped or everything is a static page, like in 1994.

Flowers, 2024 Christian Gralingen (German, dob unknown) illustration MA-g Museum of Avant-garde, Switzerland #art #painting #artist #BlueSkyArt

Some notes on talkie, a new "vintage language model" from a team including Alec Radford (yes, that Alec Radford) "trained on 260B tokens of historical pre-1931 English text" simonwillison.net/2026/Apr/28/...

Oh, all right... Reverting.

Good night. (Today's session made me realize that Sonnet 4.6 and GPT 5.3 are all I need most days.)

I feel playful as of late. I also feel nostalgic. I miss the warmth and politeness of ancient systems. They were calmer. They didn't yell at your face. They didn't vie for your attention. So I changed the theme of my blog to resemble Windows 3.1. passo.uno

I feel playful as of late. I also feel nostalgic. I miss the warmth and politeness of ancient systems. They were calmer. They didn't yell at your face. They didn't vie for your attention. So I changed the theme of my blog to resemble Windows 3.1. passo.uno

Trying to build a Windows 3.1 theme for my blog...

Would redesigning my blog to look like a CDE environment be too much?

Windows 95 came with a README file. <3

You'd be surprised at how few use LLMs to troubleshoot repo problems, including engineers. I often find myself saying "Well, my pal Claude says..." It's the new "Well, a cursory Google search says..."

TIL PowerPoint parties are a thing. Have you ever attended one?

In general, what I've found to be helpful with LLMs is knowing what you're talking about. And this is true for most things, really. Do your homework.

My obscure read of the day is this 2010 article about Soviet canned meat and how it shaped a nation. www.journal.media-culture.org.au/index.php/mc...

Wrong question, pal. You don't do it for the money, nor for the portfolio. You do it because you need to.

If people believe writing is the main thing a technical writer does, the profession is doomed. Same applies to engineering === coding.

Fragment of a chat between a friend and I about engineers: Friend: You've a clearer embrace of their technologies than a lot of us do. I'm not calling you complicit, just pointing out that you move more easily in their spaces. Me: I'm like one of those half-orcs from high fantasy. Or half-vampire.

I'm wary of prompt engineering / word twisting practices, as if the LLM was a magical black box that does its best with the right choice of words. Even if it's like that, it should not; solid AI should understand instructions without resorting to wordplay.

GSV

Hark, learned reader! 've created a Chrome extension that rewrites any doc into an Old English manuscript. Hope you enjoy this practical joke. :-) github.com/theletterf/m...

I came up with a somewhat foolish new benchmark for testing image generation models, to exercise the new ChatGPT Images 2.0: "Do a where's Waldo style image but it's where is the raccoon holding a ham radio" simonwillison.net/2026/Apr/21/...

Hark, learned reader! 've created a Chrome extension that rewrites any doc into an Old English manuscript. Hope you enjoy this practical joke. :-) github.com/theletterf/m...

I'm so bleeding edge that I'm trying stuff that apparently hasn't been tested yet and doesn't work. github.com/github/gh-aw...

Gordon Ramsey, technical writing, and the importance of people who care rachelandrew.co.uk/archives/202...

You know you're a huge nerd when the references in nerdy tshirt shops look extremely boring to you.

Our craft has hyperfocused on function for so long that the discourse has gotten brutalist. We argue about tools and commas but don’t ask what makes users return to the docs. passo.uno/what-makes-d...

Our craft has hyperfocused on function for so long that the discourse has gotten brutalist. We argue about tools and commas but don’t ask what makes users return to the docs. passo.uno/what-makes-d...

What makes documentation beautiful? Not necessarily talking about visuals or design, though those could be aspects. Also not seeking to establish whether beauty matters or not; just wondering what beauty means to you in docs.

I feel quite excited about this. Better life through checkboxes + reusable agentic workflows. Been working on this for a while.

Claude 4.7 is great, but it still comes with a pile of em dashes.