Hi there. This is 服部陽良, founder of coiai Inc., one year into running my own company.
As the founder of an IT company centered on making things — doing system development and product development — I share the reality of running a startup on this show every Monday.
Today's main topics are trying out Anthropic's newly released "Fable 5" model, and a design theory of "structure vs. decoration." In the second half, things wander even further than usual — supply and demand, underpricing your own work, thinking about UI through manual transmissions, and a founder's budget planning.
Main Topic 1: Trying Anthropic's "Fable 5" — a Disappointing Ending
First up. Anthropic released a refined (restricted) version of the model that had been generating buzz as rivaling the level of Mythos — "Fable 5" — to general users. The initial announcement said it would be available until the 22nd, but the situation has since changed. More on that below.
As soon as it was announced, I tried all sorts of things with it.
Building a Business App — the UI Was So-So
First I tried asking it to build a business app, and honestly, my first impression was "so-so."
What was so-so was things like the user interface and the overall course structure/composition. I found that my own skill, personally, was still ahead of it here, more so than our company's overall level. That said, its ability to run on its own — interpreting instructions and pushing forward independently — was fantastic. Is the UI the best it can be? Not quite. I don't think AI is all-powerful, at least not yet.
Refactoring a Unity Project — This Part Was Excellent
On the other hand, I tried asking it to refactor a Unity project I'm involved with, and it came out really well. There was a spot that had been throwing errors and that I hadn't been able to fix cleanly, but once I handed it to Fable 5, it got things running stably, and I found myself thinking "wow, this really is impressive" as I used it.
…Except It Then Got Shut Down
And then, it got suspended at the request of the US government. It's now only usable from within the United States.
Looking into it online, it seems like even connecting through a VPN with a US IP address doesn't work; apparently things like information that reveals whether you actually live in the US (such as credit card details) are also being factored in as part of the judgment. It may even be looking at what IP addresses you normally connect from. It seems to be judging US residency pretty strictly, and right now, accessing it from Japan is essentially impossible.
What was supposed to be happy news turned into disappointing news pretty quickly — that kind of episode.
Personal Update: I'm at Driving School Camp
On to this week's update. It's been a pretty busy week — for one thing, I'm currently at an intensive driving school camp, getting my driver's license.
Ever since finishing high school, I'd either been working nonstop, or, during the periods I wasn't working, laid up with depression, so I never had time to get a driver's license. But since a car is going to be essential once I start working out in the countryside from July, I've finally made time for this intensive driving course.
I brought my desktop PC packed into a backpack-style case to the driving school and I've been working out of the dormitory. It's a pretty disciplined lifestyle — either taking driving lessons or working, nothing else. Development, proposals, invoices, bookkeeping... knocking them out one by one.
Startups Are Rough — the "You Have to Do Everything Yourself" Problem
The rough part of running a startup is that you end up having to do everything yourself. We do have employees, but some are on conditional arrangements, and the other co-founder, Takemura, is more of an investor / volunteer-style presence, so I don't hand him much day-to-day work. I do assign him things, but he's also very much in the "party people" category (laughs).
Even so, my mental state would fall apart without him, so his presence is indispensable.
Main Topic 2: A Design Theory of "Structure and Decoration"
Now, on to a design topic I've been mulling over all week.
I have a fair amount of confidence in what I build, especially when it comes to UI and design. The specific thing I'm confident about is "structure" — the structural/compositional part: where you place which button, and what behavior you attach to it.
But here's the thing: to someone who hasn't studied structure and composition, this is completely invisible. Either they can't see what's good about it, or they end up assuming "that's just normal, of course it's like that."
Structure Can Be Copied Without Anyone Noticing
To put it concretely: if another company copies the structure of an app I built, nobody notices. Decoration, on the other hand, gets caught immediately.
Take the colors of Mario's buttons — you can kind of picture that instantly. Same with things like "oh, that's Kirby's UI," or "that's the Smash Bros. UI" — instantly recognizable. And what you're actually judging there is not the structure itself, but the decoration — the "costume" layer.
Being sued just because your structure is similar to someone else's is pretty rare. Look at RPGs, for instance — honestly, they all share more or less the same mechanics and composition: "there's a story, there are protagonists, they defeat enemies." You'll basically never see the founder of the RPG genre say "stop it, that's the RPG structure."
Compare that to games accused of looking too much like Pokémon. That controversy isn't really about the "structure" resembling Pokémon — it's that the character designs lean too close to Pokémon's, and that's what creates litigation risk. There are plenty of games about capturing or defeating monsters (Monster Hunter is one of them). Copying decoration is easy to spot. It gets noticed, and it also just feels unpleasant.
Copying structure or composition rarely gets noticed, and can even earn you praise like "wow, you really studied this."
The Moment You Show a Good Structure, People Think "I'd Thought of That Too"
This is actually a pretty painful phenomenon, as a designer.
The moment someone hears about a good structure from you, or experiences it themselves, it gets internalized and starts to feel like the obvious baseline. After that, anything of lower quality than that structure just feels "lame" to them.
So, no matter how well-structured an app I build and show to a client, the instant they see it, the client tends to think "oh, I'd have thought of this too." Do you know what I mean? No matter how carefully I designed and presented a genuinely usable app structure, the moment I reveal it, it flips into "I came up with that" in their mind. That's the phenomenon.
Decoration is hard to steal. It's a much finer-grained thing, more about individual ideas, so it's easy for people to say "only that person could pull off this decoration." But it's much harder for people to say "only that person could have come up with this structure/composition."
Traditional Clothing Is Also About "the Beauty of Structure" — Something I Personally Love
I'm not particularly skilled at decoration/costuming. Even in my everyday clothes, I gravitate toward trad style. I think trad is fundamentally a game of structure.
For example, a blazer pairs well with an Oxford cloth shirt, a regimental tie, and grey flannel slacks. Whether the slacks are more pipe-stem cut or an homage to marine trousers — and if you lean toward the marine-trouser side, blue works too. It's an aesthetics of structure built from conditional combinations like these. That's what I think trad style is.
The people who break away from that and do more interesting things — that's the world of fashion more broadly. I think dress clothing is a domain where "the beauty of structure" plays a huge role, and that's the part I personally love.
The Same Goes for App Design — Except Structure Rarely Translates into Price
It's the same story with app design — I love structure and composition. I like decoration too, but I don't consider myself especially skilled at it. In that domain, the people we call artists have a far sharper aesthetic sense than I do; I don't think I've reached that level.
But structure is a domain where the quality is hard to communicate to the other party, so no matter how good the structure of something you build is, it doesn't necessarily get properly recognized and sold at a high price. That's the honest reality.
Supply and Demand — Does Something Good Actually Sell for More?
On a related note, a bit about supply and demand.
The simple model you learn in middle school says that prices rise the more consumers want something and fall for things nobody wants — that as supply increases, prices go down, and as it decreases, they go up, in a mutual relationship.
But this is a model that only holds up in an "ideal world" where consumers make correct judgments. In reality, consumers don't necessarily end up wanting the thing that is objectively correct or good.
Demand gets shaped by marketing, momentum/popularity, trends, and all sorts of other factors. It's genuinely complex, which makes it hard to explain cleanly.
Startups Have No Choice but to Build Up Authority
Even if we propose something better than what other companies offer, whether the client actually perceives it as "better" is a completely separate question. Building something good essentially never translates directly into higher demand or a higher price.
In reality, decisions are often made based on "authority" — whether it's "a major established company, or a startup that hasn't proven itself yet" ends up being the deciding factor.
What's more, consumers and clients generally can't tell whether a product is actually good or not. That's because the person who signs off on a decision worth tens or hundreds of millions of yen is a different person from the one who actually uses the service on the ground. Only the person on the ground can judge "is this easy to use or not," but the sign-off comes from someone much higher up. There's just no way for quality to get communicated properly under that setup.
So lately, I keep thinking that startups need to build up their authority through marketing and presentation.
Another Concern: I Want to Stop Underpricing Our Work
This is a separate topic. I want to stop underpricing our work as much as possible.
Back when I was a sole proprietor, I genuinely took on any job at all. There were good parts to that — I got to experience a lot. But going forward, I'm in a position where it's not just my own profit at stake, but properly paying everyone on the team, so underpricing our work is genuinely not okay.
For example, if we take on a project under ten million yen — website production is completely fine at that range. The people who want a website are usually small companies, sole proprietors, people running their own shop. There's a real social-contribution angle to that kind of work too, so I can offer to build it at an accessible price.
But I've decided to stop taking on business systems or SaaS work as an "underpriced coder-for-hire." Selling off a product we went through the trouble of requirements-defining, designing, and implementing — IP and all — for under a million yen just doesn't add up at all.
Because if we built that same thing for ourselves, it would become our own product, right? Taking that on for a few hundred thousand yen instead feels almost like self-harm, honestly.
Something I Heard from a Friend: Songwriting Has the Same Structure
A quick interesting aside here.
A friend of mine said, "in songwriting, isn't the melody the decoration, and the chord progression the structure?" — and I thought, that's a genuinely great point.
For example, you can copy a chord progression or a song's overall arrangement and basically nobody says anything. Even if the same artist makes a similar-sounding song with the same progression, listeners tend to react positively — "oh, it's got that same structure as the song that was popular, but with a different melody layered on top that builds it up" — that's actually a welcome thing. It tends to become popular, even.
If a modern musician says "I studied the structure of Mozart's or Bach's compositions and applied it to my own songwriting today," the reaction is "wow, you really studied this." But copying a melody directly is seen as lame, and if two modern musicians use the exact same melody, the moment it goes past homage or sampling, it becomes plagiarism.
Line up design and songwriting side by side, and structure/composition versus decoration = melody map onto each other beautifully. I thought that was a really lovely observation.
A UI Theory from Manual Transmissions — "The Design of Everyday Things"
This is another design-related topic. I'm currently getting my license for a manual transmission.
I've come to think manual cars are genuinely great. What makes them great is that the user gets to operate the machine by understanding its actual mechanical structure. Clutch operation, gear shifting, engine RPM and torque... it's a design where an extremely simple interface is handed to the user, but operating it actually requires learning.
Giving the user an interface for operating the machine's actual mechanical structure, and letting them master it through learning — I think that's a genuinely wonderful UI, and that's how I feel about manual cars.
The Refrigerator Example from "The Design of Everyday Things"
This comes from a story in Donald Norman's "The Design of Everyday Things," which anyone who's into design has definitely read. If a refrigerator dial's UI is "the more you turn it, the colder it gets," that's incredibly intuitive. That's because it matches the actual mechanical structure. The mechanism that changes airflow and cooling matches the UI, and that's exactly why it's easy to understand.
I realized manual cars have that exact same structure. The user being able to switch how they operate the machine through learning is genuinely wonderful.
Of course, a UI like an automatic transmission, where "anyone can drive," is also wonderful, and handing over the more engineering-heavy parts to the machine rather than the user is also great design. Both have their own kind of beauty — that was the thought I landed on.
A Founder's Budget Planning — a Brake on the "Creature That Wants to Spend"
Lastly, some management-side work I did this week. Since a fairly large piece of work is coming in, I put together a budget plan.
I realized that a founder actually looking at the budget plan and doing the calculations themselves is a pretty important thing.
A Founder Is "a Creature That Wants to Spend Money"
We founders are creatures that want to spend money. Obvious, maybe, but:
- Wanting to put more into advertising to land bigger jobs
- Wanting to make the office nicer so employees can work in a better environment
- Wanting to pay employees as much as possible
— these thoughts just naturally creep in.
But the only real way to confirm whether that's actually "possible" is budget planning and accounting.
This week I did six months' worth of bookkeeping, along with allocating a budget with an eye on money coming in going forward, and I realized this is the only real brake a founder has.
It was genuinely educational, and it became an opportunity to think about the future, and to make sure I actually walk toward it deliberately. Founders really should sit down and work through the numbers themselves on a regular basis.
About coiai Inc.
That's it for this week.
coiai Inc. works in the following areas:
- Web development
- XR development
- Business system development
- Core business system development (I really want to push this one more — whatever core system you're using now, ours will always be easier to use. Please consider adopting it.)
- On-premises AI adoption support
Whatever your proposal or requirements, we promise to build something with genuinely excellent UI. As I talked about today, I'm confident in the structure/composition side of things.
People on the ground come first — why not work with us to build an environment that's easy to use and easy to work in for the people actually using it day to day. Quotes are always free, so please feel free to reach out about anything at all.
New episodes every Monday. Please subscribe to the channel, and come back to listen again next week. Thank you.