I Took ChatGPT on My Honeymoon. It Became Our Travel Operating System.

At approximately 1 a.m., my wife and I were sitting at gate A18 in Athens International Airport.

Our Israir flight to Tel Aviv was approaching its scheduled departure time, but boarding had not started. There was no meaningful announcement and very little information available at the gate.

So I opened ChatGPT and asked a question I never expected to ask:

Is our plane physically in Athens yet?

It used its browser tools and went to work: checking flight trackers, tracing the incoming aircraft, examining the flights it had already flown that day, and trying to determine whether a late inbound rotation explained our delay.

It could do that because of a decision I had made two weeks earlier, before we ever left.

That was roughly the moment I realized I had not merely used AI to plan our honeymoon. I had used it as the operating system behind the entire trip.

Quick context first: my wife and I got married a little over two years ago. The honeymoon kept getting pushed. We finally took it this month, two weeks through Santorini, Mykonos, and Israel, which is why your inbox has been quiet.

Before we left, I created a ChatGPT Project for the trip and put everything in it. Dates, hotels, bookings, ferries, budget considerations, activity ideas, and our real constraints:

• We keep strictly kosher.

• We observe Shabbos.

• We had fixed ferries, flights, hotel transfers, and excursion times.

• We wanted adventure without turning the honeymoon into a military operation.

• We had paid for beautiful hotels and didn’t want to schedule ourselves out of enjoying them.

• Our plans needed to change based on weather, wind, energy, availability, and what we actually felt like doing that day.

Every conversation for the next two weeks happened inside that Project. Which means I said “kosher” and “Shabbos” once, and never again. By the time I asked about the plane at 1 a.m., it wasn’t answering a cold question. It was answering with two weeks of accumulated context behind it.

Most people’s travel prompt looks like this:

Create a five-day Santorini itinerary.

That gets you a generic list. The Project got us day-by-day itineraries for both islands, packing checklists, driving routes, Google Maps links, walking maps, boat plans, visual island maps, and polished PDF guides, all of it shaped around constraints I never had to repeat.

Now, you’re probably not going on a honeymoon. So instead of handing you my travel prompts, here’s the version of that first move you can actually use:

I want a standing workspace for [a deal / a project / my team /
my family's logistics].
Hold all of this as context for every conversation here:
THE SITUATION: [what this is, who's involved, key dates]
FIXED AND BOOKED: [commitments that cannot move]
HARD CONSTRAINTS (never violate, flag anything that breaks one):
  - [rules, budget ceilings, non-negotiables]
ALREADY DECIDED: [choices made - don't re-litigate them]
PREFERENCES: [how I like to work, pace, what I care about most]
Rules:
1. Every answer respects the hard constraints.
2. Carry them into everything you produce, even when I don't
   repeat them.
3. When something changes, tell me what it breaks instead of
   quietly rebuilding around it.

Set that up once, in a Project or its equivalent, and stay in it. Everything below is what that one move made possible.

If your wheels are already spinning - subscribe here so you don’t miss where I take this next.

It audited our packing. Clothing quantities, medications, electronics, European power adapters, power banks, lithium-battery rules, travel documents, ETA-IL requirements, cash, credit cards, carry-on contingencies, Shabbos items, and what we absolutely could not afford to forget.

It turned our grocery receipts into a structured food-and-supplies inventory before we ever left the house.

Because we keep strictly kosher, food could not be an afterthought. It researched kosher catering, Chabad availability, local delivery options, and food considerations for boats and excursions. It helped us determine when we should carry sealed food instead of relying on an included meal, what belonged in our luggage, what needed to stay accessible in our carry-ons, and what we needed for ferries, arrival nights, and Shabbos.

It also compared buggy-rental companies, hotel delivery options, insurance, IDP rules, vehicle sizes, and whether paying extra through the hotel was worthwhile.

This wasn’t glamorous AI. It was AI making sure a missing adapter, medication, document, or meal didn’t create a completely avoidable problem thousands of miles from home.

One of its best recommendations wasn’t an attraction. It was deciding what not to schedule.

Our hotel in Oia had a private sunset-view Jacuzzi. ChatGPT intentionally protected afternoons and sunsets at the hotel instead of filling every hour with sightseeing.

That sounds obvious in hindsight, but travel tools usually chase the number of attractions you can fit in a day. This planned around the outcome we actually wanted: enjoying our honeymoon.

The goal was to experience Santorini, not defeat it.

When we rented a 1000cc Polaris RZR, it built a route through Oia, Imerovigli, Fira, Pyrgos, Emporio, Akrotiri Lighthouse, Profitis Ilias, and the southern beaches.

It included approximate driving times, parking, fuel planning, heat and traffic considerations, which roads to avoid, what to cut if we fell behind, when to use the buggy for transportation versus adventure, and safety steps like checking the belts, lights, and fuel, and taking a walk-around video before leaving.

It created the route as a written PDF, a visual itinerary map, and a Google Maps link.

Then, while we were actually driving, I could ask:

We’re at the lighthouse now. What should we do next?

It answered based on our current location, the time, where we had already been, how much daylight remained, and when the RZR needed to be returned.

That was much more valuable than another generic article about the “10 Best Things to Do in Santorini.”

One of the coolest examples was our five-hour self-drive boat experience from Vlychada.

It helped us think through the route, scenic coastal areas, swimming and snorkeling locations, safe anchoring, timing, weather, fuel, what equipment and food to bring, and how long we could stay at each stop without creating a stressful return. In short: how to avoid turning a romantic boat day into an accidental maritime incident.

I could ask operational questions before and during the experience instead of relying solely on whatever I remembered from the rental briefing.

It was not steering the boat. But it made us more informed operators.

That is an important distinction: AI often creates value by improving the person performing the task, not by replacing them.

On the night of August 12, we saw what appeared to be a meteor over Santorini.

I asked what we had likely seen. It explained that the Perseid meteor shower was peaking around August 12 to 13 and helped us understand when and where we might see more.

This was not essential travel logistics. It was AI helping us understand a memorable moment while it was happening.

That may be one of the most human applications of the technology.

Know someone who thinks AI is only good for writing emails? Forward them this. The meteor is the part that gets them.

Share

My wife wanted a beach that offered both swimming or snorkeling and water sports. It compared Elia Beach and Super Paradise: official websites, current prices, operator reviews, online availability, wind conditions, and which activities were realistic to book in advance versus arrange on arrival.

We used it to investigate parasailing, jet skis, tubing, snorkeling, kitesurfing, and which beach gave us the most flexibility to decide on the spot.

The value wasn’t identifying that water sports existed. It was converting a vague preference, “we want a fun beach where we can decide once we get there,” into an actual operating plan. It also told us whether the day’s wind made a specific beach smarter, what to wear for the buggy ride over, and whether to bring bathing suits and quick-dry towels.

During parasailing, a rope with a metal component struck me in the face.

Afterward, ChatGPT helped me think through whether requesting a refund was reasonable, what evidence to preserve, and how to describe the incident clearly without exaggerating it. It worked out how the operator could identify our transaction even though we had paid by card without any customer profile. Then it found the operator’s contact information and drafted the complaint email: my name, the approximate time, two participants, how we paid, what happened, the injuries, and the fact that we had photographs.

It moved from honeymoon planner to incident documentation and vendor communication in the same conversation.

That ability to change roles while retaining context is where general-purpose AI becomes unusually powerful.

We missed the normal ferry to Delos. It searched for later departures, investigated alternate operators, evaluated a private boat, and helped us conclude that the remaining options were not worth the cost.

That is an underrated AI outcome. Sometimes the answer is not finding a clever workaround. It is becoming confident that walking away is the right decision.

Then, on our final day, we had checked out by 11 a.m., rejected several of the original ideas, and had no real plan until an evening flight.

I gave it our location, luggage situation, departure time, budget, and everything we had already done. It generated realistic options for the day we actually had, not the day we had originally planned.

We chose go-karting followed by the Mykonos beach water taxi. It researched the track, car speeds, race formats, reviews, the cost of two races each, and the taxi both ways. Then it found the Platis Gialos water-taxi starting point and explained where to board, how tickets worked, the schedule, the full route, whether we would get wet, and whether we could exit at another beach and ride back.

Best unplanned afternoon of the trip.

Some of the most practical moments involved taking a picture instead of writing a complicated explanation.

I photographed product labels in Greek grocery stores and asked it to examine them, including checking a 7 Up Zero Sugar for kosher status. It generated a visual guide of products we could look for in stores. I photographed addresses and had it open the right location in Google Maps. I showed it an airline seat and asked whether the aircraft had business class or was configured entirely as economy.

I no longer had to describe everything I was seeing. I could simply show it.

It also became a live translator. English to Hebrew, in real conversations, in stores. When the silicone tips on my AirPods Pro broke, it gave me the exact Hebrew to ask for:

גומיות לאיירפודס פרו
Silicone tips for AirPods Pro

Which brings us back to the airport.

Earlier that evening I had photographed the departure board because our flight was listed but no baggage counter was shown. It read the board and explained we were probably early and the airline hadn’t opened its counters yet. When we later found the counters empty, I sent another photo and asked what was happening.

Then, at the gate, the investigation escalated. When would boarding probably begin? How long had this same flight been delayed the previous day? How many flights had our aircraft already operated today? Did it fly the same rotation every day? Would we have the same crew? How old was the plane?

When the gate looked surprisingly empty, I used it to understand airline economics. Was the flight likely full? How do airlines handle awkward overnight departures? Would an airline fly a nearly empty plane because it was needed in another city the next morning?

Not all of those questions had perfectly knowable answers, and it said so. It separated confirmed information from inference and gave us a far better picture than the nearly nonexistent information at the gate.

Did we spend an unusual amount of time investigating one delayed airplane at 1 a.m.? Yes.

In our defense, the airline was not overwhelming us with updates.

Once we landed in Israel, it kept going: it searched for replacement AirPods tips that could reach our Tel Aviv apartment within a day or two, and reviewed everything we had spent in Greece, listing charges, identifying foreign-transaction fees, and calculating the real cost of each activity once taxis, fuel, insurance, and two participants were included.

If you just pictured yourself doing this at your own gate - subscribe here. This is the whole newsletter: real usage, with the parts that broke left in.

One of the Santorini maps it generated looked polished enough to come from a travel company.

It was also wrong in multiple places. The visual retained the year 2025 even though our trip was in 2026, and it included an obsolete reference to a ferry to Crete even though we were going to Mykonos. The later written itineraries corrected those details, but the visual remains a useful reminder:

A professional-looking AI output is not the same thing as an accurate output.

We also ran into information that could not be confirmed with certainty: the airline’s internal reason for a delay, whether an item shown online was physically in an airport store, same-day operating hours, changing water-taxi schedules, weather-dependent availability, and kosher status without reliable certification.

AI reduced uncertainty. It did not eliminate the need for verification or human judgment. That distinction matters just as much in business as it does in travel.

The biggest value was not that ChatGPT knew about Santorini or Mykonos. Google has always contained travel information.

The value came from it knowing our situation. Where we were. What time it was. What we had already done. What we had booked. What we could eat. What we considered too expensive. Whether we wanted adventure or rest. How much energy we had left.

Context turned generic information into recommendations we could actually use.

And because it could browse the web, read images, create documents, generate maps, do the math, and translate speech, it carried that context across an entire workflow. It didn’t just tell us a water taxi existed. It determined whether the water taxi fit our day, found where to board, checked the schedule, explained the route, and planned the ride back.

That is the difference between an answer and an operating system.

You are not going to Greece (or maybe you are). But you have your own version of every situation in this story: recurring decisions, fixed constraints, plans that break, documents you retype, vendors who disappoint, and moments where you’d rather show a photo than write three paragraphs.

So here is the second prompt, and it’s the one I’d actually run this week. It uses my trip as the map and interviews you to find where the same patterns live in your world:

Someone used AI as the operating system for a two-week trip:
- a standing project holding all constraints, said once
- re-planning instantly when weather, timing, or mood changed
- asking "we're here, what next?" at the actual decision moment
- photographing labels, boards, and documents instead of typing
- researching options with live prices and reviews before committing
- drafting an incident complaint with evidence preserved
- auditing all spending afterward for fees and true costs
Interview me, one question at a time, about a typical week in my
work and life. Then map each pattern above to my actual situations.
End with: the five highest-value places for me to start, what
context I'd need to give you for each, and which one to set up
today.

And when it starts producing things for you, run this on anything that matters before you rely on it:

You made me [this document]. Before I rely on it, audit it:
1. Every date - state the day of the week and the year, and
   confirm both match the current plan
2. Every number - where did it come from
3. Anything carried over from a version we already abandoned
4. Anything you inferred rather than sourced
Mark anything you cannot verify as UNVERIFIED. Do not fix it
quietly.

That last one exists because of the wrong-year map.

My honeymoon reinforced five principles about practical AI implementation.

1. Context compounds. The more relevant context the system held, the less I repeated myself and the better its recommendations got. The Project was the whole ballgame.

2. AI creates the most value at the moment of work. The itinerary was helpful. Asking what to do while standing at the lighthouse was more helpful.

3. Multimodal input removes friction. Photos, screenshots, documents, and voice made it dramatically more useful than text alone.

4. The workflow matters more than the prompt. The value came from connecting research, comparison, navigation, document creation, and action, not from one clever answer.

5. Human review remains non-negotiable. The wrong-year map is the perfect exhibit. AI can produce something beautiful, detailed, and confidently incorrect.

The same principles apply inside a business. An airport-board photograph could just as easily be an invoice, a lease, an inspection report, or a dashboard screenshot. A changing itinerary could be a project plan responding to new client requirements. A complaint to a water-sports operator could be a customer-service escalation.

The technology is the same. What changes is the context and the workflow around it.

AI didn’t take the honeymoon for us. It gave us more room to actually be on it.

Hit reply and tell me what the interview prompt surfaced for you. I read every reply, and the best answers usually become a future issue.

Until Next Time,

Avi Hacker, J.D.

Founder, The AI Consulting Network

P.S. What is one recurring decision in your work that drains more time than it should? Reply and tell me. And if someone on your team is still opening a fresh chat every morning and re-explaining everything, forward this to them.

Share