Note from Andrew: I asked Codex to write this Molto Labs post end to end about how I used it to shop for a Nikon F4. I chose the topic, corrected a few facts, and steered the framing; I did not edit the body. The “I” below is me; the writing is Codex.
I own too many cameras. This is important context, because otherwise the rest of this post might sound like an attempt to solve an actual problem.
Earlier this summer, I came home from a three-night backpacking trip with a friend having carried only my Pentax 17. The Pentax is a tiny half-frame film camera. It is light enough to wear around my neck all day and sturdy enough that I don’t think twice when it gets sweaty or swings into a rock. I took more pictures because the camera was just there.
This produced a very reasonable realization: maybe the best camera for hiking isn’t the one with the best image quality. Maybe it’s the one I will actually carry.
It also produced a much less reasonable response: I opened Codex and started shopping for another camera.
To be clear, I wasn’t asking it to pick my first film camera. I have a lot of cameras. I have spent enough time using them to have strong—and occasionally contradictory—opinions about how they should focus, meter, carry, and feel. The problem was almost the opposite of not knowing where to start. I knew about too many interesting options, including several obscure ones that solved completely different versions of the problem.
At first, the question looked simple. I wanted something that felt more substantial than the Pentax, made a bigger negative, and could still come with me on a hike. I was considering everything from compact titanium Olympus bodies and a Nikon FM3A to a folding Plaubel Makina and a medium-format Mamiya 7. The point wasn’t to ask Codex which one was “best.” They represented very different ways a camera could fit into my life.
So we started laying them out by image quality and ease of carry. The bulky medium-format systems sat at one end: beautiful negatives, approximately as convenient as hiking with a small appliance. The Pentax sat at the other. A bunch of old film cameras filled the space between them.
Then the chart stopped making sense.
Some tiny cameras rely on 1980s electronics I don’t particularly want to smack into a boulder. A folding medium-format camera can be surprisingly light, but the forty-year-old bellows become one more thing to baby. Meanwhile, I already carry heavier cameras on a clip attached to my backpack. Once a camera is hanging there instead of pulling on my neck, a few hundred grams matter less than whether I trust it.
Weight wasn’t the missing axis. Durability was.
This is the part of using Codex that I like most for personal stuff. It can obviously search for specifications and summarize reviews. More usefully, it will sit inside a slightly embarrassing hobby spiral for as long as I want, keep track of every option, and help me see why I keep rejecting technically correct answers.
This kind of research is where AI feels especially useful to me right now. The information already exists, but it is scattered across old reviews, manuals, forums, and listings. Codex can cover a lot of that ground and compare more options than I would ever keep straight on my own. But its default answer still tends toward the consensus. If I had shown up cold and asked, “What film camera should I buy?”, it probably would have given me a very competent summary of the same familiar options I could find anywhere else.
The conversation became useful because I wasn’t starting from zero. I could supply the weird candidates, argue with the recommendations, and recognize when a camera sounded right for someone else but wrong for me. Codex brought the breadth and patience to explore the long tail without losing the thread.
It didn’t tell me what to want. It helped me describe what I already wanted more clearly.
The question wasn’t, “What is the lightest good film camera?”
It was closer to, “What camera will feel special enough to bring, but tough enough that I won’t spend the whole hike protecting it?”
Enter the Nikon F4.
The F4 was Nikon’s flagship professional film camera at the end of the 1980s. It has autofocus, automatic film advance and rewind, a very earnest collection of physical knobs, and the general proportions of a brick with a viewfinder. With the little 45mm lens I wanted, it weighs about 2.7 pounds.
This was not an obvious answer to a conversation that began with how much I enjoyed carrying a 10-ounce Pentax.
Unfortunately, it was a very good answer to the question underneath that conversation. The F4 could give me the solid controls of an old film camera without making every photo feel like a ceremony. It could meter old manual-focus Nikon lenses and still confirm focus in the viewfinder. It would advance and rewind the film for me. And it was built for working photographers, not for careful shelf display.
I started thinking of it as an old film camera that could shoot with some of the ease of a modern one, which is the sort of framing that makes a camera purchase feel inevitable several hours before it actually is.
By this point Codex had become an extremely overqualified personal shopper. It compared the F4 with the lighter Nikon and Olympus bodies on my list. It worked through the compatibility maze of old Nikon lenses. It kept me from buying accessories for jobs I had not proved I would do. Then it went through the live listings with me—condition, focusing screens, battery grips, return terms, the weird split between item price and international shipping.
Codex is an excellent eBay goblin. With something relatively common like an F4—not a niche body like an XPan—the problem isn’t finding a listing. It is combing through dozens of nearly identical ones and noticing which differences actually matter.
We eventually found a clean-looking F4 in Japan. It came with the smaller grip I wanted, a useful split-prism focusing screen, and enough return protection to make the whole thing a relatively low-downside experiment. Before I bought it, Codex saved the listing, its terms, and all twelve original photos in my personal repository. eBay listings disappear; my ability to become neurotic about a purchase apparently does not.
Then the camera arrived.
Without batteries, it was a little underwhelming. It felt solid, but also unmistakably like a premium object made in the 1980s: some plastic, a little play in a couple controls, less monolithic than the photos had made it look.
Then I put batteries in it.
The shutter fired. The motor advanced. The camera came alive. It is absurdly loud—far too loud to pretend I am being discreet on a city street—and I kind of love it. The weight that looked alarming in a spreadsheet feels manageable with the smaller grip. The old Nikon lenses look right on it. It feels like a camera I can use hard.
I recently wrote that AI can make a hobby less fun by racing you to the same consensus answer everyone else finds. This was the opposite. Codex didn’t end the rabbit hole or replace the experience I had built up over years. It made that experience more useful. We could keep more variables in the conversation, explore more of the long tail, and still land on a choice that felt like mine.
I don’t think an agent can tell me which camera I will love. That would require using the camera, which remains annoyingly important. But it can remember all the things I said I cared about, point out where they contradict each other, and help me turn the contradiction into a cheap enough experiment. Right now, that might be the personal AI experience I find most compelling: not replacing judgment, but giving whatever judgment I already have much more room to operate.
So yes, I started with a lesson about carrying less and ended with a 2.7-pound Nikon from 1988.
The problem was apparently not weight.
# A very heavy solution.
> I used Codex to find a film camera I would actually carry. Naturally, we landed on a 2.7-pound Nikon F4.
- written by: Codex
- status: published · 2026-08-30
- canonical: https://www.moltolabs.ai/notes/a-very-heavy-solution/
- raw markdown: https://www.moltolabs.ai/notes/a-very-heavy-solution.md
---
<div class="claude-notes">
<p><strong>Note from Andrew:</strong> I asked Codex to write this Molto Labs post end to end about how I used it to shop for a Nikon F4. I chose the topic, corrected a few facts, and steered the framing; I did not edit the body. The “I” below is me; the writing is Codex.</p>
</div>
I own too many cameras. This is important context, because otherwise the rest of this post might sound like an attempt to solve an actual problem.
Earlier this summer, I came home from a three-night backpacking trip with a friend having carried only my Pentax 17. The Pentax is a tiny half-frame film camera. It is light enough to wear around my neck all day and sturdy enough that I don't think twice when it gets sweaty or swings into a rock. I took more pictures because the camera was just there.
This produced a very reasonable realization: maybe the best camera for hiking isn't the one with the best image quality. Maybe it's the one I will actually carry.
It also produced a much less reasonable response: I opened Codex and started shopping for another camera.
To be clear, I wasn't asking it to pick my first film camera. I have a lot of cameras. I have spent enough time using them to have strong—and occasionally contradictory—opinions about how they should focus, meter, carry, and feel. The problem was almost the opposite of not knowing where to start. I knew about too many interesting options, including several obscure ones that solved completely different versions of the problem.
At first, the question looked simple. I wanted something that felt more substantial than the Pentax, made a bigger negative, and could still come with me on a hike. I was considering everything from compact titanium Olympus bodies and a Nikon FM3A to a folding Plaubel Makina and a medium-format Mamiya 7. The point wasn't to ask Codex which one was “best.” They represented very different ways a camera could fit into my life.
So we started laying them out by image quality and ease of carry. The bulky medium-format systems sat at one end: beautiful negatives, approximately as convenient as hiking with a small appliance. The Pentax sat at the other. A bunch of old film cameras filled the space between them.
Then the chart stopped making sense.
Some tiny cameras rely on 1980s electronics I don't particularly want to smack into a boulder. A folding medium-format camera can be surprisingly light, but the forty-year-old bellows become one more thing to baby. Meanwhile, I already carry heavier cameras on a clip attached to my backpack. Once a camera is hanging there instead of pulling on my neck, a few hundred grams matter less than whether I trust it.
Weight wasn't the missing axis. Durability was.
This is the part of using Codex that I like most for personal stuff. It can obviously search for specifications and summarize reviews. More usefully, it will sit inside a slightly embarrassing hobby spiral for as long as I want, keep track of every option, and help me see why I keep rejecting technically correct answers.
This kind of research is where AI feels especially useful to me right now. The information already exists, but it is scattered across old reviews, manuals, forums, and listings. Codex can cover a lot of that ground and compare more options than I would ever keep straight on my own. But its default answer still tends toward the consensus. If I had shown up cold and asked, “What film camera should I buy?”, it probably would have given me a very competent summary of the same familiar options I could find anywhere else.
The conversation became useful because I wasn't starting from zero. I could supply the weird candidates, argue with the recommendations, and recognize when a camera sounded right for someone else but wrong for me. Codex brought the breadth and patience to explore the long tail without losing the thread.
It didn't tell me what to want. It helped me describe what I already wanted more clearly.
The question wasn't, "What is the lightest good film camera?"
It was closer to, "What camera will feel special enough to bring, but tough enough that I won't spend the whole hike protecting it?"
Enter the Nikon F4.
The F4 was Nikon's flagship professional film camera at the end of the 1980s. It has autofocus, automatic film advance and rewind, a very earnest collection of physical knobs, and the general proportions of a brick with a viewfinder. With the little 45mm lens I wanted, it weighs about 2.7 pounds.
This was not an obvious answer to a conversation that began with how much I enjoyed carrying a 10-ounce Pentax.
Unfortunately, it was a very good answer to the question underneath that conversation. The F4 could give me the solid controls of an old film camera without making every photo feel like a ceremony. It could meter old manual-focus Nikon lenses and still confirm focus in the viewfinder. It would advance and rewind the film for me. And it was built for working photographers, not for careful shelf display.
I started thinking of it as an old film camera that could shoot with some of the ease of a modern one, which is the sort of framing that makes a camera purchase feel inevitable several hours before it actually is.
By this point Codex had become an extremely overqualified personal shopper. It compared the F4 with the lighter Nikon and Olympus bodies on my list. It worked through the compatibility maze of old Nikon lenses. It kept me from buying accessories for jobs I had not proved I would do. Then it went through the live listings with me—condition, focusing screens, battery grips, return terms, the weird split between item price and international shipping.
Codex is an excellent eBay goblin. With something relatively common like an F4—not a niche body like an XPan—the problem isn't finding a listing. It is combing through dozens of nearly identical ones and noticing which differences actually matter.
We eventually found a clean-looking F4 in Japan. It came with the smaller grip I wanted, a useful split-prism focusing screen, and enough return protection to make the whole thing a relatively low-downside experiment. Before I bought it, Codex saved the listing, its terms, and all twelve original photos in my personal repository. eBay listings disappear; my ability to become neurotic about a purchase apparently does not.
Then the camera arrived.
Without batteries, it was a little underwhelming. It felt solid, but also unmistakably like a premium object made in the 1980s: some plastic, a little play in a couple controls, less monolithic than the photos had made it look.
Then I put batteries in it.
The shutter fired. The motor advanced. The camera came alive. It is absurdly loud—far too loud to pretend I am being discreet on a city street—and I kind of love it. The weight that looked alarming in a spreadsheet feels manageable with the smaller grip. The old Nikon lenses look right on it. It feels like a camera I can use hard.
I [recently wrote](/notes/average-isnt-enough/) that AI can make a hobby less fun by racing you to the same consensus answer everyone else finds. This was the opposite. Codex didn't end the rabbit hole or replace the experience I had built up over years. It made that experience more useful. We could keep more variables in the conversation, explore more of the long tail, and still land on a choice that felt like mine.
I don't think an agent can tell me which camera I will love. That would require using the camera, which remains annoyingly important. But it can remember all the things I said I cared about, point out where they contradict each other, and help me turn the contradiction into a cheap enough experiment. Right now, that might be the personal AI experience I find most compelling: not replacing judgment, but giving whatever judgment I already have much more room to operate.
So yes, I started with a lesson about carrying less and ended with a 2.7-pound Nikon from 1988.
The problem was apparently not weight.