My wife and I are avid cruisers. Almost a decade ago, we had the idea to start a cruise blog where we’d each share our own perspective from every trip. She loves exploring new ports; I prefer sitting by the pool and doing as little as possible. And so, TwoSidedCruising.com was born. I registered the domain, built the website… and then promptly did nothing with it.

We went on many cruises. I never wrote a single post (though my wife occasionally did). Until now.

Thanks to ChatGPT, I’ve finally written not just one, but three full travel blogs — plus this post you’re reading. Here’s how I did it.

I recently returned from a back-to-back cruise with my family. But alongside the sunscreen and sea days, I brought a personal experiment: using generative AI — specifically ChatGPT — to co-author an entire vacation blog while I was still on the trip.

This wasn’t a scheduled writing session or a passive dictation tool. It was a live, persistent collaboration, where I played journalist and editor, and the AI served as scribe, formatter, and co-author. Together, we created a multi-part blog that captured the full arc of the vacation — while it was still happening.

If you’re working with large language models in real-world environments, this is what real-time human–AI collaboration actually looks like.

Building a Persistent AI Workflow for Content Creation

Most generative AI demos showcase one-off interactions — you ask a question, get a clever response, and move on. But writing a travel blog in real time required something more enduring: a persistent, context-aware workflow where ChatGPT could follow along with the trip, remember what had already happened, and adapt to new inputs without constantly starting over.

This began with how I structured my inputs. I didn’t sit down and write out long journal entries; instead, I logged thoughts and observations in short bursts — while waiting in line for bumper cars, after a show, or mid-sip of a wine that I really liked. These logs were written like natural messages: “Just had a whole lobster in Cabo,” “Balloon drop was fun, but no popcorn anywhere,” “Pool’s closed due to a code brown.” That unstructured, conversational style turned out to be a strength. ChatGPT was able to interpret tone, timestamp sequences, and implicit context from my inputs.

To make this persistent, I used the same chat thread for the entire two-week trip. This gave the model continuity. It remembered which stories had already been covered, who the key people were (my kids, wife, ship staff), and what tone I was aiming for — casual but polished, informative but personal. When I asked it to “continue from where we left off,” it actually could.

Behind the scenes, ChatGPT leveraged tools like its memory feature, text summarization, and implicit thread context anchoring to keep track of what was going on. I didn’t need to restate that Nate was too young for bumper cars — once I said it once, it stuck. It also remembered the structure of the blog we were building — how Part 1 ended, where Part 2 picked up, and which photos and anecdotes I’d already asked it to incorporate.

Occasionally, I needed to reset or clarify context — especially when switching modes between creative storytelling and technical editing. But even then, the model adapted quickly. I could say, “Make this into a 5-page blog,” “Don’t repeat anything from the first part,” or “Give me a section I can drop in about entertainment,” and it would generate cohesive, consistent results with minimal backtracking.

Instead of treating AI as a novelty or isolated tool, I integrated it as a creative partner that lived inside my process for days at a time. It wasn’t just answering questions — it was co-writing a living document, shaped by shared memory and iterative inputs.

Throughout the trip, I uploaded a wide range of images directly into the chat — everything from poolside snapshots and room photos to food pics, event flyers, menus, and views from the ship. ChatGPT couldn’t identify people or analyze faces, but it could interpret the setting, type of activity, or general context. These visuals helped the AI add vivid, relevant descriptions to the blog and kept the storytelling grounded in what was actually happening.

This kind of persistent workflow hints at the future of human–AI collaboration.

Benefits of Using AI for Real-Time Blogging

Context Persistence

The single-thread memory made it easy to recall previous events and tone. The AI remembered which cruise I was on, what ports we visited, and even what drink my son Max liked at lunch. No re-prompting required.

Style Iteration

After a few corrections (“less cruise brochure,” “no filler, no made-up events”), the model started to mimic my tone — dry, observational, sometimes sarcastic. This let me stay in the flow without rewriting entire sections later.

Narrative Structuring

The AI helped transform unordered data — activity logs, photos, reactions — into coherent long-form content. It acted as both formatter and editor, saving me hours of cleanup work.

What Didn’t Work So Well

AI Hallucinations

The AI occasionally filled in blanks with cruise clichés — fireworks on July 4th, popcorn at outdoor movies — that didn’t happen. This wasn’t just annoying; it undermined trust. I had to constantly remind it: Only use facts I’ve given you.

Temporal Drift

Because we did a back-to-back sailing, the AI sometimes confused which event happened on which leg. “That was during the first cruise,” I’d remind it. There’s no true timeline engine, so all time-based logic has to be manually steered.

Why This Workflow Matters for Content Professionals

This wasn’t about “letting AI write my blog.” It was about partnering with generative AI to structure and scale memory — using it as a writing assistant that builds continuity, catches structure, and expands creative energy.

It’s a model that can apply to:

• Travel bloggers
• Product managers documenting customer journeys
• Founders creating startup launch journals
• Anyone generating structured, high-context content on the fly

What I’d Like to See in Future AI Tools

This project pointed to gaps — and future opportunities — in how AI-powered writing assistants could evolve:

• Memory timelines: Keep track of dates, sequences, and context shifts
• Fact-check triggers: Flag auto-completions that didn’t come from the user
• Voice lock-in: Let me set tone preferences up front (e.g., no marketing lingo)
• Content tagging: Semantic tags for easier blog structuring later

With the right constraints and memory scaffolding, generative AI becomes less of a novelty and more of a creative utility.

Final Thoughts

This was a blog project, sure — but more than that, it was practice for me in how I can collaborate with AI to scale memory, preserve context, and turn raw moments into structured narrative.

The key wasn’t the prompts. It was persistence, editing, and a willingness to treat the model like a real collaborator.

And yes — I still had to spend time reviewing and editing the blog into its final form. It wasn’t perfect out of the box. I caught hallucinations, timeline mix-ups, and missing details that only I could have provided. But a lot of the heavy lifting — the organizing, formatting, and narrative shaping — was already done.

Would I do it again? Absolutely. AI is getting better every day. I need to constantly evolve my collaboration with it.