AOL Keywords were a thing in the 90's, and in traditional fashion some enterprises copied this approach internally for their intranets. But in the modern era the keyword clashes have been replaced by semantic ones. www.spletzer.com/2026/10/the-...
Ryan Spletzer
Ryan Spletzer
@ryan.spletzer.com
Distinguished Engineer at Autodesk | spletzer.com
129 followers277 following58 posts
I've been seeing some, uh, shall we say, *misguided* decision making on the part of some in the industry who decide, for whatever reason, to put an LLM behind an MCP server. I discuss the reasons why this is a bad idea, and what to do instead, in my latest post. www.spletzer.com/2026/07/dont...
AI coding tools have gotten remarkably good at a great many things. Producing tidy Markdown is not one of them. My latest post covers how I tame it: a linter as the source of truth, semantic line breaks, and teaching the agent to clean up after itself. www.spletzer.com/2026/07/the-...
"Keep your eyes on the road, your hands upon the wheel" โ "Roadhouse Blues" by The Doors I can think of no better advice when it comes to vibe coding, and I don't mean the throwaway kind; I mean the serious kind with real rigor and engineering practices around it. www.spletzer.com/2026/06/its-...
ccccccgkgdkvvvjvfblbbdihehndrvjkutduvftrgubfc If you're working at any organization that uses YubiKeys, we've all seen something like this emitted accidentally in a channel somewhere. In my latest post I illustrate how to disable this! www.spletzer.com/2026/04/how-...
I've owed this post to some security friends of mine for a while. I've been really into GPG commit signing with a YubiKey for many years, and I finally delivered on my promise to write an in-depth guide about it. www.spletzer.com/2026/04/a-no...
Humans make mistakes, and not only can we learn from them, but we can learn how to put AI agents to work in ways that prevent them from making the same mistakes as we would. In my latest post I delve into a mistake I made when I accidentally dropped a database. www.spletzer.com/2026/04/the-...
When things get cheaper, in many scenarios we often use more of them, not less. A Victorian era economist figured this out about coal in 1865, the cloud proved him right again, and I (and others) think AI is next. www.spletzer.com/2026/03/jevo...
I wrote a follow-up to my post on information diet, and this time I got specific. I use RSS with Feedbin and NetNewsWire, subscribe to about 125 feeds, and don't have any social media apps on my phone, giving me more time for writing, thinking, and building. www.spletzer.com/2026/03/cura...
I opened a fresh Claude Code session, typed one prompt, and 40% of my context window was already gone. I call this "dead context," and if you're running AI coding tools, you probably have the same problem without realizing it. www.spletzer.com/2026/03/shed...
In 2024 setting up Python needed pyenv, pip, pipx, shell init scripts, and pain on Windows. In 2026 it needs one tool: uv. Same commands on every OS, 10-100x faster installs, and way simpler. A new no-nonsense guide to setting up Python environments: www.spletzer.com/2026/03/a-ne...
A Tale of Acceleration and Compound Engineering www.spletzer.com/2026/02/a-ta...
Software has always felt like sculpture to me: you start with a block of possibility and chip away until the shape reveals itself. AI coding tools are the fastest chisels (and the best shop assistants) Iโve ever hadโbut they donโt replace taste. They amplify it. www.spletzer.com/2026/02/the-...
Skeletons are in your closets. Bats will fly out of your caves. The question is whether you let them out now, or wait until your competitors do it themselves and find the hidden treasure on the other side before you do. www.spletzer.com/2026/02/fear...
I spent a little bit of time with Claude Code this weekend and refactored something that had been on my backlog for a long time: my website. www.spletzer.com/2026/02/one-...
The laziest strategy in fast food may be the smartest strategy in AI. In my latest post, I spend 1,500 words on a burger metaphor that probably could've been a tweet. www.spletzer.com/2026/02/mcdo...
New Blog Post: Tidying Your Home for Your AI Guests www.spletzer.com/2026/01/tidy...
It was my absolute pleasure to be a guest on The PowerShell Podcast, where I had a great conversation with the illustrious Andrew Pla. Get it wherever podcasts are sold! YouTube: www.youtube.com/watch?v=ryZ7... Podbean (or grab it in your favorite podcast app): powershellpodcast.podbean.com
Ryan Spletzer reposted
Is it safe to write a blog post that isnโt about AI? In this slightly humorous and irreverent reflection, I look at AI hype, technological amnesia, and why we keep pretending the rest of technology doesnโt exist. (No agents were harmed in the writing of this.) www.spletzer.com/2025/12/is-i...
Ryan Spletzer reposted
MCP pushed me back into looking at OAuth extension specs, and I found myself in a spider web of browser tabsm so I decided to make a Mermaid graph diagram in a GitHub repo to visualize the specs and their reference relationships.
Weโre living in the age of "vibe coding"โwhere itโs easier than ever to generate something that looks real. But just because Pinocchio can sing and dance doesnโt mean heโs a real boy... www.spletzer.com/2025/09/pino...
I wrote a post about applying CQRS (Command Query Responsibility Segregation) principles to AI agents. www.spletzer.com/2025/08/ask-...
I wrote a blog post to try to disambiguate what MCP can and can't do, and what it is and is not. www.spletzer.com/2025/08/mcp-...
If you haven't had a chance to watch this Pragmatic Engineer interview with Gergely Orosz and Laura Tacho, I highly highly highly recommend it. So many pull quotes from this discussion that resonated with me, thinking "Yes! Finally! Someone is saying it!" ๐ www.youtube.com/watch?v=xHHl...
I wrote a post about your information diet in the age of AI. www.spletzer.com/2025/07/your...
I wrote about the hidden complexities of running Model Context Protocol (MCP) servers locally vs remotely, and why deployment context shapes everything from auth to architecture. ๐ www.spletzer.com/2025/05/the-...
Ryan Spletzer reposted
The curse of social media: not just for software engineering but other areas as I observe (say, hiring, investing, product management, venture capital etc) Those the best at their craft are usually busy building not writing/posting. Viral posts often wrong in many ways etc Via @icooper.bsky.social
I wrote a post about the area of enterprise search, where I spent a great deal of time over a decade ago in my career, and whose lessons Iโm realizing are more relevant than ever today in the age of AI, RAG, and agents. www.spletzer.com/2025/05/ente...
I wrote a post about the SPIFFE standard and its accompanying open-source implementation SPIRE. www.spletzer.com/2025/03/zero... (Image courtesy of ChatGPT)
I wrote a post about the SPIFFE standard and its accompanying open-source implementation SPIRE. www.spletzer.com/2025/03/zero... (Image courtesy of ChatGPT)
I wrote a post about how to build an engineering team from scratch. www.spletzer.com/2025/02/how-...
I wrote a post about how to get your client_id and client_secret from an app registration in Entra ID (formerly known as Azure AD) since this regularly trips many people up. www.spletzer.com/2025/01/how-...
I wrote a post about my fun side quest to get SpinRite to run on my Apple Silicon-based Mac. www.spletzer.com/2024/12/runn...
Ryan Spletzer reposted
Ryan Spletzer reposted
Ryan Spletzer reposted
I wrote a post reflecting on my participation in Autodesk's COO Leadership Development Program (LDP) over the past six months. www.spletzer.com/2024/11/san-...
I wrote a post reflecting on my recent experience in Tokyo as part of Autodesk's COO Leadership Development Program. www.spletzer.com/2024/06/tokyo/
I wrote a post that is a no-nonsense guide to setting up Python environments for local development on various platforms. Or rather, there's still _some_ inherent nonsense involved in the process, but I've tried my best to minimize it. ๐
I wrote a post on the importance of experimentation and proofs of concept in de-risking technology decisions and system designs.
I wrote a quick post about a tool that I believe does not get enough hype: Git Credential Manager.