Tonight I’ve been testing something really cool: using Claude, Craft MCP connections, and Craft Agents to write a blog post to Scribbles. The blog post was written by Claude under my instructions but I could tweak the workflow to get the text of the blog post from a Craft Collection instead. Something to experiment during the upcoming Christmas holidays.

Saved 50 Minutes

Realmac Software shared their latest dev talk video. The video title mentions conversations about future plans for Elements. I was curious because I want to know where they are going with the CMS and RSS. I headed to YouTube and asked AI the following question: did they mention CMS? In a few seconds, I got my answer: yes, and they also talked about better support for RSS, which is something I’ve been waiting for. I didn’t watch the video; I already had the information I wanted. I saved 50 minutes of my time.

Now, one question: how is this good for Google? We’re so accustomed to being manipulated by platforms designed to increase our engagement with them; with this AI feature, it’s the opposite—I’m less engaged. Is this another subversive move by Google, part of a master plan that escapes my awareness?

What a night & day difference between Claude Sonnet 4.5 and Opus 4.5 models. The latter is much more effective at working through my MCP connection to Craft. But boy this baby consumes a lof of credits. 😅

I’m trying something new this year for my year-in-review blog post. Using all my monthly post digests stored in Craft1, I’m using ChatGPT to look at those digests, take into account last year’s year-in-review article, plus this year’s document personal milestones to suggest ideas. This is now possible because Craft now supports MCP. Using the best ChatGPT models, I get a lot of material to consider but somehow I feel this is really overwhelming. I should be more directive like limiting the number of possible scenarios, their length, etc. I’m still exploring this workflow.

  1. I subscribe to my own blog post digest newsletter for archival purposed. ↩︎

Time to renew Kagi Search. I’m on the starter tier. Upon examination, I’m doing about 45 searches per month. The trend is decreasing. I use their Article Summarizer on the iPad and iPhone much more often. I want to support them because I believe we need diversity when it comes to searching the web. But it’s not cheap, and my search requests trend is going downward, thanks to AI. Decisions, decisions, decisions. Any Kagi users in the room?

I spent the last few days putting together a workflow to help me write those year-in-review articles. As we get closer to the end of the year, it’s the right time to start working on this. I documented my workflow in this Craft document and submitted to the #winter_challenge on Slack. Hint: my first real use case for MCP.

I don’t like the idea that a service like ChatGPT tries to present itself as a friend, an agent, or a human. With instructions, I can ask it to avoid this type of behavior. But I like the service’s appearance to be warm. Perhaps that’s why I prefer Anthropic’s Claude. Claude, even though it’s a proper name, presents itself better; its visual style seems friendlier to me. I feel like this is a contradiction.

I asked Claude about this possible contradiction. Here’s what he responded:

“I don’t think it’s actually a contradiction. You’re distinguishing between two different things: false intimacy (an AI pretending to be your friend or a person) versus pleasant design (a warm, welcoming aesthetic and tone). Wanting the latter without the former is perfectly coherent—it’s the difference between a tool that’s agreeable to use and a tool that’s trying to deceive you about its nature.”

Thanks, Claude.

I’m experimenting with Claude for a month because I want to test MCP connections with Craft. ChatGPT does support MCP connections but the experience is not as mature as configuring an MCP connection with Claude. Claude is not cheap: pricing is similar to ChatGPT’s but plan usage limits are easy to reach.

Three Years Already

Three years of ChatGPT. Time flies. A few thoughts on that are mandatory it seems. ChatGPT certainly turned things upside down not only for me as a writer but for many creators. The entire software industry also was turned things even more upside down. It’s hard to imagine what would have been software roadmaps or new features pipelines without generative AI. See? We almost forgot about what it was like before ChatGPT. Can’t imagine where things are going. Generative AI integration into apps and user experience is the next frontier where things will be interesting to watch.