I Just Automated My Engineering Friend

A weekly digest. A regex pattern. A Claude prompt that translates engineering jargon into something a marketer can actually use. Built in thirty minutes. Bobby has not yet been notified.

By Shannon Ropelato, Founder, You Digital Co.

Bobby Ulrich is one of my favorite humans on the internet. He runs BDigital Media , builds production-grade developer tooling, and writes a weekly tech blog that I love receiving and cannot read.

I am not exaggerating. His most recent post is titled "A Three-Bar Statusline for Claude Code." The one before that explains "How LLMs Generate Text" using phrases like "logits" and "softmax distributions" and "temperature, top-k, and top-p." When Bobby links me an article, I open it, scroll, nod respectfully at the words I do not understand, and reply "this looks awesome!" because that is a true statement and also because I have nothing more specific to offer.

This is fine. We are not in the same lane. Bobby is a high-tech engineer who builds AI tools for engineers. I am a strategic marketer who builds AI tools for marketers. He recently launched a training product for developers who want to construct production AI workflows. I recently launched a marketing audit product for businesses that need a senior marketer to actually look at their numbers and tell them what to do. We are using AI in completely complementary ways for completely different audiences. The Venn diagram is two circles touching at a single point and the point is "I have used Claude today."

But here is the thing: I want to read his stuff. Not because I'm going to start writing code (don't count me out), but because the strategic implications of what Bobby is building affect how I think about marketing infrastructure. If I understood it better, I'd make better calls for clients. The friction was never that I didn't care. The friction was that the explanations are written for someone with a CS degree, and I have a marketing degree, and the gap between those two is exactly where I get stuck.

So tonight I told Claude: automate Bobby.

What I asked for

The brief was twenty seconds long. Pull articles from Bobby's blog once a week. Summarize them in one to two paragraphs for a marketing professional with no engineering background. Skip the jargon. Tell me what it's actually about and why it might matter. Email me the digest every Monday morning.

I did not write a single line of code. I did not configure a single Make.com scenario by hand. I did not parse a single piece of HTML.

Claude built it. The data store to track which articles had already been seen. The HTTP module to fetch the homepage. The regex pattern to extract article URLs. The filter to skip ones already summarized. The Anthropic API call to do the actual translation work. The aggregator to bundle everything into one email. The Gmail module pointed at my inbox. Nine modules total, all wired together, scheduled for Mondays at 8 AM.

What this actually changes

The output is small. One email a week. A digest of whatever Bobby has posted, in language I can read.

The implication is bigger. I used AI to bridge a gap between two professionals who otherwise speak completely different dialects of the same language. I did not have to ask Bobby to dumb anything down. I did not have to pretend I understood what I didn't. I did not have to add "read tech blog and slowly google things" to my weekly to-do list and then never do it. The translation layer just exists now, and it costs me almost nothing but the inference spend to run it (about a dollar a month, max).

This is the part of AI that does not get written about enough. Most of the AI conversation is enterprise. "AI will replace this department." "AI will transform that workflow." Big, capital-letter words about big, capital-letter shifts. Meanwhile, the actual revolution is small and embarrassingly personal. I just automated my friend. I built something that lives entirely between Bobby's writing and my reading, serves an audience of one, and cost me almost nothing. It exists because I had a small, persistent annoyance and the tools to fix it without needing to hire anyone or learn anything I did not already know.

That is the version of AI that should be more interesting to most working professionals than the doomer takes or the breathless productivity ones. Not "AI will run your business." Closer to "you have permission to automate the small frictions in your life that you've been tolerating."

We are using AI in opposite directions and meeting in the middle

Here is what I find genuinely funny about all this.

Bobby is using AI to build tools for people who build tools. His training product teaches developers how to construct production AI workflows: skill libraries, slash commands, evaluation pipelines, the whole disciplined-engineering stack on top of LLMs. It is excellent.

I am using AI to build tools for people who run businesses but are not technologists. My audit product takes fifteen years of marketing pattern recognition and packages it into a deliverable a small business owner can actually act on. Bobby has been through it (and loved it, for the record). It turns out we are not just building for different audiences. We are also, on a small scale, each other's audience.

Different professional worlds. Different deliverables. Same underlying technology. And, when it actually matters, we use each other's stuff.

And then on a random Monday night, I use the same technology to solve the gap between us. I use Claude to translate Claude. I use the engineer's preferred toolkit (Make.com, an Anthropic API key, a regex) to build a marketer's accommodation for the engineer's blog. The recursion is sort of beautiful if you stand back from it.

What is actually live

For people who care about specifics:

• A Make.com scenario named "Bob's Tech Blog Weekly Digest" running in my account
• A data store tracking which articles I have already been emailed about
• A weekly schedule pointed at Monday morning at 8 AM
• A Claude Haiku 4.5 prompt doing the layperson translation
• A Gmail module dropping the digest in my inbox
• Total ongoing cost: under one dollar per month
• Total build time: thirty minutes, mostly conversation

The first email lands next Monday and I am genuinely excited about it.

Who this is for

If you are a working professional who has been told "just learn AI" and felt vaguely paralyzed by the size of that instruction, this is the version of AI that I think actually matters for you.

You do not need to replace your job with AI. You do not need to become an engineer. You need to identify the small, recurring frictions in your week ("I get sent things I cannot easily process," "I have to manually compile this report from three sources every Friday," "this client question keeps coming up and I keep typing the same answer") and ask yourself whether AI could remove the friction without removing you.

That is what I just did. I removed the friction of not being able to read my friend's blog. I am still the marketer. Bobby is still the engineer. The content of his work is still over my head when I open the raw post. The translation layer between us just runs automatically now, every Monday, before I have my first cup of coffee.

Bobby, if you are reading this: I automated you. I am very sorry. Also, your blog is still excellent.

The rest of you: figure out what your version of "I cannot read my friend's blog" is. Then figure out whether thirty minutes of Claude time could fix it. The answer, surprisingly often, is yes.