10 things I learned from building a newsletter agent

There’s certainly a lot of hype around the idea that anyone can build an agent these days in the matter of a few hours. But what I’ve learned from building a newsletter agent is that whatever you build at a rapid speed with AI, will require time iterating and training.

One year ago I was hired to create an AI-powered email newsletter for a client, pulling from sources such as LinkedIn, Google calendar, news sources and more. But I hit a wall. Zapier could only take me so far. That’s when I met my now Spiral co-founder, Xavier. Within hours of explaining my issue we had an MVP agent running on N8N.

One year ago, N8N delivered the cheapest and most flexible solution for functionality, self hosting and protecting data.

In a few weeks we will release a v2 that connects a wider range of data inputs, including note-taking apps and will even include a Chrome extension for capturing newsletter ideas. We wanted to build something that keeps a human in the loop while cutting down the hours and time it takes to curate, write and publish an email newsletter.

We’ll also be moving off N8N and using our own system. We’ve learned that third-party platform updates often break integrations and meant we had to keep reconnecting APIs manually. 

What we’ve learned along the way: A great newsletter is essentially a small online publication. The ideas that can make your newsletter stand out can get lost without a system to capture and develop them.

Most people want AI to do things faster. But AI is only as good as your inputs. This is the problem Spiral solves: an all-in-one system for idea capture, editorial strategy and consistent publishing.

Spiral's bet is that the bottleneck isn't AI capability; it's idea capture, editorial strategy, and consistency. That's the gap we’re building to close. Check out our origin story and what we’ve learned from a one-year build in this week’s HyperFix breakdown on our YouTube channel here. 

10 Things We Learned Building an AI Newsletter Agent From Scratch

1. Start messy, build fast. The goal of your first build is to prove something is possible, not to build something perfect. Our first MVP was rough, but it validated the whole concept and gave us something real to iterate on.

2. Choose your tools based on your ceiling, not your starting point. The tool that gets you moving fastest isn't always the one that gets you where you're going. Know roughly where you want to end up before you pick your stack.

3. Owning your data means owning your platform. Building on someone else's infrastructure means their updates become your emergencies. The moment we moved to our own backend, we stopped being a third party to our own product.

4. API access is not guaranteed. We applied for LinkedIn API access over a year ago. One of us got denied. The other is still waiting. Just because you need access to data doesn't mean anyone is obligated to give it to you.

5. "I can build that in a day" is a starting point, not a finish line. Yes, AI raises your floor dramatically. But UX, security, copy, authentication, and database architecture are all still legit problems that take real time to solve. Shipping a product and shipping a prototype are two different things.

6. The bottleneck isn’t writing a newsletter. For most busy founders and CEOs, the real problem is idea capture. The insights are happening all day in meetings, conversations, and articles but they disappear without a system designed to catch them.

7. Human in the loop isn't a limitation. AI handles the middle: research, assembly, compression of time and effort. But humans need to drive the beginning (strategy, voice, ideas) and the end (review, approval). 

8. Expect many versions before you have something real. We built a Telegram bot, then a custom chat interface, then a V1 web app, and now V2. Each one was necessary. Each one taught us something the previous version couldn't so that we could make something that actually solved problems. 

9. A newsletter is a media asset, not a content calendar. Done well, a newsletter builds an audience, generates leads, launches products, and becomes a standalone revenue stream. That requires editorial strategy, not just a publish button. The tool matters less than the thinking behind it.

10. Tech is a living thing, tend to it accordingly. Especially right now. AI capabilities are shifting fast enough that what felt cutting-edge a year ago feels basic today. Whatever you build, plan to keep building it

I’m currently part of a team building a tech product for a consulting company. I noticed a pattern of dumping meeting transcripts into Claude and asking it to make a prompt based on what was said in the meeting.

However, upon closer look at the transcript and the notes, even AI couldn’t figure out what to derive. Ideas were floated in a product meeting without reaching a definitive output or timeline.

Instead of letting Claude Cowork run rampant with the fuzzy meeting notes and build a prompt, a team lead and I sat there for a few hours making sure we understood the new product requirements. It took going back to leadership and saying things like “when you stated this, did you mean x or y.”

It resulted in responses like: “Now that I’m really thinking about it, why don’t we try this instead.” 

More investigation and discovery was needed to fully understand and give AI the right directions. In this Substack piece titled  "AI Raises the Cost of Weak Agency":
Alex de Carvalho argues that AI amplifies whatever is already present in a person or organization. It lands inside overloaded teams and distracted minds and turns up the volume on what's already there. Without true knowledge AI drift toward wrong goals accelerates, and avoidance of real decisions gets hidden behind beautifully formatted options.

Actress Reese Witherspoon may have stepped into an AI controversy she didn’t intend to create. Over the weekend she published an IG video essentially stating that women weren’t adopting AI as fast as men and should be because women’s jobs are more likely to be taken by AI.

Cue the angry backlash. Including author Sophia Benoit who wrote: “There’s something particularly insidious about seeing that women — the group you have built your brand on — have not adopted something and instead of assuming it’s out of wisdom, infantilizing them with ‘we’re falling behind.’”

My take: Women are tired of being told they need to do more or be more in a world that places never-ending demands on them. I think most of us know by now that AI is coming for our jobs, but the tone of her messages felt like another scarcity-based canned AI message that every tech bro is broadcasting with a megaphone.

It was tone deaf on the way that AI has impacted the creative industries that Witherspoon represents. A lot of people are asking if she was paid by AI companies to make that post or if she’s investing in AI startups on the side.

Either way, I think this is an important cultural moment. The “AI revolution is here” mantra is on its way out. How about more empowering messages, especially for women who work in creative industries and are committed to authenticity?

🔥 EXTRA HYPE

📨 P.S. If you are building a newsletter and want to eliminate the weekly scramble, that is exactly what we help our customers do.

 👉 Book a consult with our team

And if this sparked something, forward it to an AI-curious friend.