I tried to find a place for Ollama in my automation stack. I wanted two things: a lower Anthropic bill, and Ollama running somewhere in my workflows. The first reason didn’t survive contact with the data.

The number

I have a workflow that pulls my Anthropic cost report every morning and posts it to Discord and my Craft daily note. So I checked it.

Thirty days, 56 workflows, every LLM call I make: $7.48. Twenty-five cents a day, already down 60% over that window.

Moving everything off Claude would save about $90 a year. Keeping Claude for work where quality matters brings that closer to thirty.

I standardized most of my stack on Haiku months ago. That’s where the savings already went. I went looking to cut a bill that was already at its floor, because reducing API costs sounds like good engineering whether or not the number supports it.

Ollama Cloud costs more than I spend

Two things I had confused: open-weight models and locally-hosted models are separate choices. I don’t want models running on my Mac. That’s not an argument against open models — Ollama Cloud runs them on their hardware behind a normal API. I’d even set up the credentials back in December and forgotten.

But Ollama Cloud doesn’t bill per token. It’s a flat subscription metered by GPU time, with session limits every five hours and weekly limits every seven days. Pro is $20 a month — nearly three times my current bill, for smaller models.

The free tier might cover my volume. I can’t confirm that, because Ollama has no account usage endpoint. Checking your quota means logging into the web dashboard. Several open feature requests ask for an API; none have shipped. You find out you’re near the ceiling when they email you at 90%.

I went looking to reduce a cost I track with a daily automated report, and the alternative can’t be tracked with one.

The reason that holds up

Cost is out. Privacy doesn’t apply — I’m summarizing my own RSS feeds and my own billing data.

What’s left: I don’t know which of my workflows actually need a frontier model.

I have a dozen-plus workflows making LLM calls. Some summarize articles or pull signal out of a week of notes. Others do arithmetic and pattern-matching dressed up as reasoning. I’ve never tested the difference. Everything points at the same model because at twenty-five cents a day there was no reason to check.

An open model is a cheap way to test that. Not to replace Claude — to find out where the line is. Any task where a small open model matches Haiku is a task I’ve been over-provisioning.

The experiment

I picked the most mechanical node I have: the one that asks four questions about thirty numbers. Trend, spikes, projection. 570 tokens in, no nuance required.

Both models now run in parallel from the same prompt, joined by a merge node so the message builder doesn’t fire twice and double-post. Both answers print back to back, labelled with their model. The Ollama arm fails soft — a rate limit on the test shouldn’t break the report.

One thing I nearly got wrong: I reached for the largest open model first, because it was the name I recognized from the docs. Wrong instrument. A big model will obviously handle a trend summary, and confirming that tells me nothing. If you’re looking for the floor, start at the floor.

If you’re considering the same thing

Spending real money on inference? Do the migration. The math works.

Bill is a rounding error? Then cost isn’t your reason and privacy probably isn’t either. The useful question is which of your automations need a frontier model and which ones you’re over-provisioning out of habit.

I’ll report back either way.

This morning, I delved into integrating Craft with Claude Cowork. Using the Craft MCP connector, I built a dashboard that sources data directly from Craft tables. This setup generates a Live Artifact in Claude Cowork, which dynamically updates as more data is added. Although Lovable could have been used to achieve this, the absence of custom domain support in its free tier led me to choose Claude Cowork, which offers more flexibility for my needs.

Today, I’m excited to share my latest digital ecosystem update, previously known as “The Numeric Citizen Content Creator Workflow”. The diagram sports a new title: “The Numeric Citizen Digital Ecosystem” to better reflect its scope: to enumerate services and apps at the center of my digital presence. It was entirely rebuilt in Apple Freeform, leaving Apple Keynote behind. Freeform is a joy to use; it’s much more flexible and powerful than Keynote, too. The content was also expanded to include a new set of tools I have built over the last few months since publishing the previous update. Let's dive in for all the details.

The entire ecosystem diagram was rebuilt from the ground up. Full resolution version available here.

At a high level, both Micro.blog and Ghost remain the two hosting and publishing pillars for most of my *.numericcitizen.me websites, while a few other websites, like my Ko-fi page and my new about website, Who Is Numeric Citizen, are hosted elsewhere. On the diagram, my websites are positioned on the left and in the top-left portion.

As you might notice, a lot has happened since my last update. Thanks to Claude AI and Claude Code, I made a major shift toward web app to meet two fundamental needs: a brand new bookmark manager (replacing AnyBox), and a lightweight RSS reader that integrates with my bookmark manager. Together, they support my reading workflow and help me produce the Ephemeral Scrapbook newsletter. See both web apps in this YouTube video.

My custom-built RSS reader web application

What’s in, what changed since last fall

AI is now playing a more prominent role in my digital ecosystem since the beginning of 2026. It’s mainly used for vibe coding, maintaining my web applications, and content summarization. You’ll see little light bulbs (💡) in the diagram; they indicate where AI summarization is involved. Anthropic’s Claude AI replaced OpenAI’s ChatGPT in all my workflows. Plus, Claude AI and Claude Code enabled many new possibilities for helping me build small digital services and web apps for my personal needs. It’s something I couldn’t have imagined a year ago.

Claude Code’s main window with some of my projects on the left sidebar and the DIFF view shown on the right

Another milestone was to replace IFTTT with n8n automation. It’s a major upgrade in functionality and usability. My n8n instance is running on the Digital Ocean hosting service, inside a pre-packaged “droplet” with 1 GB of RAM and 10 GB of disk space for $6 a month. Digital Ocean provides a one-click install of n8n which really made a difference in selecting Digital Ocean as a hosting service. I created more than a dozen automations, far more complex and useful than what I could build with IFTTT. Claude AI was also involved in optimizing many n8n workflows. It was possible because n8n supports the MCP endpoint that is exposed to Claude AI. In the following table, I present a summary of my n8n automation workflows. As you will see, Discord is a new entry, as it provides a custom server to receive n8n workflow outputs and some status reports from Tinylytics. Also good to know: all my n8n workflows are automatically documented in a Craft document, thanks to Claude Code, which connects to both the n8n and Craft backends. To get a well-structured documentation, I created a Claude Skill.

A portion of an automation workflow shown in the n8n canvas

My automated n8n workflows

Here’s a list of utilities running as n8n workflows on my self-hosted DigitalOcean instance.

• API Credit Health Check — Pre-flight check on Anthropic credits; alerts Discord if depleted
• Today's Forecast — Pulls weather data into the Craft daily note
• Tinylytics Insights — Injects site analytics summary into the Craft daily note
• Apple Daily News — Fetches MacSurfer headlines into the Craft daily note
• Setup Craft Daily Note — Creates the day's 8-section journaling scaffold in Craft
• Anthropic Daily Cost Report — Posts 7-day and 30-day AI spend breakdown to Discord

Example of automatically propulated Craft Daily note, thanks to n8n automation.

RSS & News Aggregation

A few workflows are used for content processing in the context of news consumption and Micro.blog timeline processing.

• RSS Daily Summary — Aggregates RSS feeds and delivers an AI-summarized digest
• Micro.blog Timeline Highlights — Summarizes Micro.blog activity via Claude
• Apple News Flash Alerts — Real-time Apple news alerts from RSS
• Apps & Services Release Highlights — Monitors release notes feeds for notable updates

Content & Publishing

Other workflows for supporting my content creation include:

• Craft → Ghost — Two workflows (specific and generic) push Craft documents to Ghost as drafts
• Ghost Publish Proxy — Triggers final publication on Ghost via webhook
• Craft Fetch Proxy — Fetches Craft document content on demand for downstream processing
• Poke to Craft / Peek at Ghost — Manual/chat-triggered utilities for quick content inspection

Newsletter & Email

A special workflow automation is processing email summarization on demand: simply forwarding an email to a dedicated Gmail account and after a few seconds I’ll receive a summary on my Discord server.

• Email Summarizer — Polls Gmail every minute; summarizes newsletters and posts to Discord

Maintenance

Finally, backing up my workflows is a must, just in case something really bad happens on my n8n instance.

• GitHub Backup — Backs up all n8n workflow definitions to a GitHub repo every Sunday

---

The Craft Agents main window

The updated diagram illustrates this: Craft consists of two parts—Craft content stored in the Craft cloud backend and various Craft clients, including the native general-purpose client and Craft Agents. Additionally, Craft now supports MCP and offers a comprehensive set of APIs, emphasizing the division between data storage and client applications. These latest additions to Craft are game changers and enable new use cases.

A few thoughts and observations about Craft Agents, a sort of spin-off of Craft, are worth sharing here. Think about Craft Agents as an orchestrator of AI agents. In many ways, Craft Agents can be compared to the Claude Desktop app. There is a high level of affinity between Craft Agents and Craft, connected via APIs or MCP. Consuming Craft content through Craft Agents is an obvious use case. Yet, I’m still not sure why the team behind Craft is spending so much effort on this, but it’s interesting to follow nonetheless. I’m still looking for clear use cases for it, hence the yellow dot (🟡) on its icon in the diagram. There are a few aspects of the app’s design and structure that I don’t like. I don’t like using the task manager to organize conversations. I don’t think task management belongs in this app. Plus, Craft Agents is a very hungry consumer of AI credits. It’s a major concern unless you subscribe to the highest tier of any AI provider (OpenAI, Anthropic, etc.). It’s worth noting that Craft also offers an assistant working inside Craft itself.

Also added to my digital ecosystem is CleanShot Cloud for hosting publicly shared screenshots or short screen recordings. I use that to share bugs or feedback. CleanShot X is sort of the client to CleanShot Cloud and is the best Mac utility you can get if you tae a lot of screenshots.

One last automation that prevented me from closing my IFTTT account was the process of automatically archiving my publications in DayOne. This process was briefly described in this blog post. Each time I share a blog post or an article, it is saved in DayOne. Micro.blog added a send-to-DayOne feature, which was the last needed nail in the coffin of IFTTT. It was such a relief.

Speaking of Micro.blog, I created a new website for sharing automated blog post digests on a monthly basis. Most of the process is supported by the use of AI summarization. It’s not fully automated but very effective. Each month I send my Micro.blog postss digest to Craft fo later processing using Craft Agents. The process takes less than 15 minutes to complete. I know there is a potential for a fully automated process. It’s on my todo list.

What’s out since the last update

Chillidog Web Hosting is out (the service hosting the Who Is Numeric Citizen website). It was replaced with Realmac Software Elements’ hosting solution. You can read the doc here. The transition was seamless. I prefer integrated solutions, which Realmac is now offering. Plus. Chillidog was purchased last year by Exact Hosting, and the purchase has been reported to degrade customer service.

Last fall, I added the ChatGPT Atlas browser to my ecosystem, but I removed it after switching to Claude AI. OpenAI Atlas was mainly used for summarizing articles, but now I do this within the Claude Desktop app. I still prefer using the ARC Browser, even though it might eventually be discontinued. The new browser, Dia, could take its place, but I’m interested to see how things unfold. Dia appears to require a Pro subscription to enable AI features in the browser. At $20 a month, I’ll pass.

Gone is Google News for cross-posting from Micro.blog. It never took off. Also gone is AnyBox, the very well-done bookmark manager available on all Apple platforms, but not available on the web, like Raindrop.io. I never liked Raindrop.io, so I built my own bookmark manager as a web-only app. It’s one of my new custom-built web apps.

Finally, the numericcitizen.io domain will be retired when it expires later this year, and the content hosted on a dedicated Craft subscription has already been moved to my main Craft subscription, helping me reduce costs. The meta.numericcitizen.me is taking over (my previous content was migrated there during 2025). Some of my content was also simply removed from public access.

My newly redesigned metablog as seen in dark mode

Still pondering about these

You’ll find yellow dots on the diagram which points to services or apps that I’m reconsidering. Let’s review a few of them.

Plausible vs Tinylytics: I use two web analytics. Plausible has been around in my ecosystem since I left Google Analytics. Then came long Tinylytics and I wanted to support its developer plus it provides a great design compared to Plausible. Yet, I have a lot of historic data in the latter. Plus, I started using Tinylytics APIs to feed one of my workflows who inject data inside Craft daily notes and dashboard web application. I could replace this with Plausible’s APIs but I don’t feel it’s worth the effort.

Things 3, a popular and highly regarded task manager, is currently being reconsidered. I like managing my to-dos, but I don’t have many. If I want to eliminate Things 3, I could switch to Craft’s basic task management, but it’s quite limited. For now, my most likely option is to create my own app, much like I did with my RSS reader or bookmark manager. However, I’m not in a hurry to switch away since it doesn’t require a subscription, giving me plenty of time to decide and develop a replacement.

Ulysses is another case where I’m pondering its long-term usefulness. If I can post to Micro.blog or Ghost directly from Craft (using my custom-made n8n workflows), why should I continue paying for this app? The next renewal is a year from now.

Finally, my Flipboard account hosts a magazine that pulls content from my main website via its RSS feed. I have 8 followers, apparently. I never get any reactions or comments from this place. The bigger question: is Flipboard still a thing these days?

Risk management

In this edition of the diagram, I included red dots (🔴) to indicate potential disruption risks. For example, Anthropic recently adjusted Claude's credit consumption, leading to a fivefold increase in the number of credits needed to perform the same task; something I've noticed too. Additionally, Anthropic conducted an experiment in which new subscribers could access Claude Code only if they chose the Max subscription tier at $200 per month. Although they quickly reversed this change, it raises the concern that they might eliminate the $20 plan altogether. If that happened, I could face serious difficulties and might stop maintaining my web applications built with Claude Code. Vercel is another risk potential: I’m using the free tier, which is rather generous. If they decided to remove this free tier, I would have to reconsider my position: either drop my web application or accept paying the monthly price they charge.

Concluding thoughts

This evolution reflects a fundamental shift in how I approach my digital life. By embracing AI assistants like Claude and building custom tools tailored to my specific needs, I've moved away from cobbling together disparate services toward a more cohesive, intentional ecosystem. The transition from automation platforms like IFTTT to n8n, coupled with the ability to code solutions with Claude Code, has given me both flexibility and control—rare commodities in the world of Software-as-a-Service.

The introduction of AI summarization across my workflows has transformed how I consume information, while the consolidation of my web presence under fewer, better-managed domains simplifies my digital footprint. Yet this ecosystem remains dynamic and somewhat fragile, dependent on the sustained availability of key services and the pricing decisions of AI providers. The risk management considerations I've outlined aren't just technical footnotes—they're essential guideposts for maintaining digital independence in an increasingly interconnected world.

As I continue to iterate and refine this ecosystem, the principles remain constant: intentionality, control, and the ability to evolve. The next edition will likely bring even more changes, but I'm optimistic that the foundations I've built will prove resilient enough to adapt to whatever comes next.

Previous update to my digital ecosystem can be found here.

Recent readings about n8n and with the help of Claude AI, I finally found a way to replace IFTTT for archiving my RSS feeds content: it won’t be by using an n8n automation but a simple feature that was sitting right in front of my eyes on Micro.blog: cross-posting. The trick is simple: each of my RSS feed is added to the Sources panel in Micro.blog, from there, I configure cross-post the RSS feed article to other places like Day One journal using email-to-dayone. Voilà. I’ll be closing my IFTTT account later today after so many years of service. The only rather big downside: Micro.blog doesn’t copy images from the source into Day One while IFTTT could. 🫤

Update #1: I submitted my issue on the Micro.blog help forum, and a few hours later, a fix was on the way. Thanks @manton

Update #2: And now it is available; photos are now supported, too! Thanks to Manton!

It has been a while since my last update in March 2025. Here’s a summary of the changes.

• I removed Brief.news because I no longer think it will replace Mailbrew.
• I removed Mailbrew because I no longer depend on it to consume Internet content. I tried to replace it with Inoreader email digests, but it didn’t work as I wrote here.
• I decided to add ChatGPT Atlas because I now have a solid use case for it: articles summarization and analysis, as I explained in this YouTube video. This means Perplexity didn’t stay from my previous update. I’m focusing and want to settle on OpenAI for the foreseeable future.
• My new personal landing page, which is mostly complete, has replaced the one previously hosted on Craft public documents.
• I also made several visual tweaks to make it cleaner and more visually appealing.

The pace of updates slowed considerably in the last two years. It’s a good thing, and it means I can focus more on content and less on tooling.

The Who Is Numeric Citizen? website has moved from Cloudflare’s Workers app to Chillidog Hosting. This change was necessary to support PHP, which is required for Elements CMS. Additionally, the publishing workflow is now more straightforward since I no longer need to use a GitHub repository, which was necessary for hosting a static website on Cloudflare. It is rare to simplify a workflow while adding more features, in this case, PHP to implement a CMS component on the News page.

The migration process was relatively simple, thanks to Chillidog support, which was fantastic BTW. The process involved republishing my website using a new configuration publishing setup, which pointed to Chillidog. The setup was straightforward because Chillidog is familiar with Elements, which allows you to download a configuration settings file directly into Elements for setting up the FTP connection. Once republished, I made a DNS change at CloudFlare to replace a CNAME record with an A record to one of Chillidog’s servers. After 15 minutes, it was a done deal.

I’m trying out the new generative AI-based analytics from Tinylytics, and I’m very impressed. Here’s an example for my main website. To my knowledge, that’s a first for website analytics and I hope that Plausible will take a look at this because it’s really useful. Plus, Tinylytics allows you to create your own prompt and see a different take than the one generated by default.

Some cleanup: Readwise is gone. Supporting services are now grouped at the bottom. Corrected a few typos. I made some visual adjustments to make things a little bit cleaner and easier to visualize, especially for website miniatures. I renamed the diagram to reflect the notion of an ecosystem instead of a workflow.

Many additions: each enhanced service with generative AI features is marked as such with a little brain icon. That’s the case for Inoreader, Craft and Grammarly. All my Micro.blog-hosted websites are now indicated. Since adhering to POSSE principles, I added the Fediverse and Bluesky icons and drew the cross-posting arrow lines to them.

A high-resolution version of this diagram is available here.

What is POSSE

The POSSE principle stands for “Publish (on your) Own Site, Syndicate Elsewhere.” It is a content distribution strategy often recommended for writers, bloggers, and publishers. The primary idea is to first publish your content on a platform you control, such as your personal website or blog, and then syndicate or share that content on other platforms like social media, Medium, or different online communities.

Here are some key points about the POSSE principle:

1. Ownership and Control: By publishing on your own site first, you maintain control over your content and ensure it exists in a space you own. This helps protect your work from the risks of platform changes or shutdowns.
2. Centralized Content: Your website becomes the central hub where all your content is stored and can be accessed by your audience.
3. Traffic and SEO: By driving traffic to your own site, you can improve your website’s SEO, increase your audience, and potentially monetize traffic through ads, affiliate links, or direct sales.
4. Syndication: After publishing on your own site, you can share your content with a wider audience by syndicating it to other platforms. This strategy helps reach readers who might not visit your site directly.
5. Preservation: Content publishing on third-party platforms may be subject to their rules and policies. Publishing first on your own site ensures your content is preserved and remains unchanged regardless of policy changes elsewhere.

The POSSE principle is popular among creators who value long-term control over their work and want to build a sustainable and direct relationship with their audience.