What building an AI workspace taught me about token bills, wasted time, and the real price of software.
Most people think affordability is a number on a pricing page.
After two years of building and launching apps at 123sudo, I think it's much bigger than that. The price is decided by every step we add, every minute we waste, every piece of data we ask for, and every time someone pays for more than they need.
Disclosure: I'm a co-founder of 123sudo, where Vijay and I built 9xchat, an AI workspace with multiple models and credit-based usage. I use it every day, and building it shaped a lot of what I'm sharing here.
The bill you don't notice
One AI subscription feels reasonable. Then you add a second for writing, a third for images, and maybe a fourth because a new model launched last week.
Each one makes sense on its own. Together, they quietly become a monthly habit.
You might be paying for several tools while using each of them for only part of your work. That's not a pricing problem. It's a design problem: the software was shaped around the subscription, not around how you actually work.
Tokens are becoming a utility bill
Here's what I think many people haven't fully noticed yet.
As AI becomes part of everyday work, we're moving toward something closer to an electricity bill than a one-time purchase. Every question, image, and long document uses compute. Compute costs money, and powerful models are not unlimited.
That's why we see usage caps, rate limits, and plans that change as demand grows. Nobody can honestly promise that AI costs will only go down.
So the long-term question isn't just "How cheap is this tool?"
It's "Will I understand what I'm spending, and can I control it?"
Why we chose credits
When we built 9xchat, we chose credits instead of locking everyone into the same monthly plan.
You top up, use what you need, and see where your credits go. Different models and tasks cost different amounts, and I'd rather be honest about that than hide it inside an all-you-can-use promise that eventually comes with limits.
For me, credits respect a simple truth: not everyone uses AI the same way.
Vijay is a developer whose workflow has evolved toward working with his own agent. I use the same workspace for research, images, writing, and communicating with our team. Our usage looks completely different. A single fixed plan would make one of us overpay.
Affordable means paying for your work, not someone else's average.
The right dose for the task
I studied pharmacy before I ever worked in tech. One lesson stayed with me: the right dose matters. More isn't automatically better.
AI works the same way.
You don't need the most powerful and expensive model to fix the grammar in a short message. And you probably shouldn't use the quickest, lightest model for a complex decision with real consequences.
| Task | What usually makes sense |
|---|---|
| Fixing grammar, rewording, short replies | A fast, lighter model |
| Research on current topics | A model with live web search |
| Complex planning, analysis, long documents | A stronger reasoning model |
| Exploring a visual idea | An image model |
| Repeated workflows | An agent you set up once |
Choosing the right model for the task is one of the simplest ways to save money. It's also a skill worth learning, because it makes you think about what the task actually requires.
That's part of why we put many models in one workspace. Switching shouldn't mean opening another app, paying another subscription, or explaining your project again from the start.
Time is part of the price
A simple design change used to take my designer two days.
Now I can explore changes myself in 9xchat's Image Studio, see different directions, and arrive at the conversation knowing what I want. It doesn't replace a designer's skill. It removes the waiting between my idea and the next decision.
Two days of waiting has a cost, even if it never appears on an invoice.
The same goes for re-explaining your context every time you open a new chat. Persistent memory means I can save what matters about my work, update it, or remove it, without starting from zero each time.
If a tool is cheap but wastes your time, it isn't affordable. It's just billing you differently.
Your data shouldn't be the hidden payment
My law degree taught me to read the fine print.
When a product feels unusually cheap or free, it's worth asking how it's being paid for. Sometimes the answer is your attention. Sometimes it's your data.
At 123sudo, we don't have a hobby of collecting people's emails. Our apps use a UID instead, so you don't need to share your email address to get started. We still stay connected with our users through in-app support.
What is a UID, exactly?
UID stands for unique identifier. Think of it as an ID number the app uses to recognise your account, rather than asking for your email address.
At 123sudo, we use UID-based access so you can get started without sharing your email. If you need help, you can reach us through in-app support.
But there’s an important distinction: a UID is an identifier, not a privacy guarantee. It doesn’t automatically make you anonymous or mean no data is processed. Privacy also depends on what information an app stores, how it protects that information, and which services receive it.
For example, when you use a cloud AI model, your request still needs to be sent to that provider for processing. UID-based access removes the need to provide an email address; it doesn’t change how cloud AI works.
For me, it’s one practical design choice: don’t ask people for personal details just to let them begin.
I also want to be honest: when you use cloud AI models, your requests are processed outside your device. Privacy isn't one magic switch. It's a series of decisions about what we ask for, what we keep, and what we tell people clearly.
Software isn't affordable if your privacy pays the difference.
What "affordable by design" means to me
When we build, I keep coming back to a few questions:
- Can someone try it without a big commitment?
- Will they understand what they're paying for?
- Are we asking for data we don't actually need?
- Does it save time, or create more to manage?
- Can people choose the right tool for the task instead of paying for the most expensive one by default?
None of these show up as a single number. All of them decide whether software is genuinely affordable.
The bigger dream
I believe good software shouldn't feel like a luxury.
A student, a craftsperson, a solo founder, or someone working in their second language should be able to use powerful tools without needing a big budget, a technical background, or a willingness to give away their data.
As AI becomes part of everyday life, affordability will matter more, not less. The people who understand their usage, choose the right model, and protect their data will get far more value from the same budget.
I want to build tools that help people do exactly that.
What's one tool you're paying for right now that you're not sure is worth it anymore?
KAT — Co-founder of 123sudo and Pangur Care. Building practical, affordable, privacy-first software.
Top comments (0)