DEV Community: KAT | Co-founder, 123sudo

I Didn’t Want to Quit My Phone. I Wanted My Time Back.

KAT | Co-founder, 123sudo — Sat, 14 Nov 2026 16:59:37 +0000

How I brought my daily screen time from 4–5 hours down to around an hour or less, without giving up the parts of technology I genuinely like.

I’m old enough to remember when people barely knew how to send an SMS.

I’m also part of the generation that can lose hours to a feed before realizing the room has gone quiet.

I’ve lived in both worlds.

I know what it feels like to be completely fine without checking anything. And I know what it feels like to stay awake far too late, telling myself I’ll put the phone down after just one more video.

So this isn’t a lecture from someone who has always been disciplined.

I’ve lost plenty of evenings to a screen.

The difference is that after a while, it stopped feeling like something I was actually choosing.

I wasn't even enjoying it

There is a big difference between watching a movie on purpose and opening a social feed because your hands were empty for three seconds.

With a film, I usually know why I’m there. With a feed, I can go in to reply to a single message from a friend and emerge forty minutes later having forgotten what I opened the phone for in the first place.

Then comes the stranger part:

If you asked me what I watched during those forty minutes, I could barely tell you two things.

That was what finally started bothering me. Not just the lost time, but how little of it felt intentional.

The reflex problem

The hard part wasn’t admitting that I scrolled too much.

The hard part was seeing how automatic the movement had become:

Waiting for a build to finish? Open an app.
A task gets slightly difficult? Tap the screen.
A few seconds of silence? Fill it immediately.

I caught myself closing an app, looking around, and opening the exact same app two seconds later.

My thumb had an agenda my brain hadn't signed off on.

I didn't need another generic productivity tip telling me to "just build discipline." I needed a way to break the loop before I was already twenty posts deep.

What we lose when we can't sit still

I see this restlessness everywhere now, especially in younger people.

A pause becomes uncomfortable. A longer video feels impossible. We open a second screen while something is already playing on the first one.

I don’t believe human attention has magically broken forever. But I do think we get very good at whatever we practice all day. If we practice switching away every eight seconds, that becomes our default setting.

That’s why I have so much respect for anyone who can still sit with a book, read a full technical article, or listen to a long conversation without reaching for a pocket.

Not because reading makes someone morally superior.

Just because staying with something takes practice.

Outside of tech, I paint and play piano. Nothing I’ve ever made that felt meaningful came out of a ten-second burst of focus. You have to stay through the boring, uncomfortable minutes before the good part shows up.

If our tools make it harder to stay, we end up paying for it in the work we actually care about.

I still want social media in my life

I’m a co-founder at a tech company. I’m not about to throw my phone into a river and pretend I live in 1995.

I use social platforms to discover visual art, find new music, follow builders, and stay connected with friends across time zones.

I don’t want to lose that.

I just don’t think keeping up with friends requires four or five hours of every day I’m alive.

My goal became simple: open an app for a specific reason, enjoy what I came for, and leave before it turns into an entire evening.

What helped me break the habit

Along the way, I started using deScroll, an Android launcher designed around intentional phone use.

(Disclosure: I’m involved with the team behind deScroll. This is my personal experience using it.)

A few practical things made a difference:

A calm, text-first interface: removing colorful, high-contrast app icons stopped my eyes from treating the home screen like an arcade menu.
A visible scroll counter: seeing the number of scrolls made a habit that’s usually invisible impossible to ignore.
A pause before opening distracting apps: just enough friction to ask myself whether I actually wanted to open the app or was simply reacting to boredom.

It didn't fix everything by magic. I still had to make the decision to put the phone down.

But that small pause between the urge and the tap gave me back enough room to make an actual choice.

Over time, my daily screen time went from four to five hours down to around an hour or less.

That’s my own experience, not a guaranteed result for everyone. And screen time alone isn't a moral score: reading documentation, debugging on mobile, or calling family all happen on screens.

What changed for me was cutting out the hours I never planned to spend in the first place.

Time back doesn't have to become "productivity"

I don't believe every minute you rescue from a screen needs to be turned into more work.

Sometimes I play piano. Sometimes I work on a canvas. Sometimes I have a conversation without checking a notification halfway through.

And sometimes I just let myself be bored.

Empty time doesn't need to be optimized. It just needs to belong to you.

I still catch myself reaching for my phone without thinking. The difference now is that I notice it early.

And more often than not, I leave it face down on the desk.

Next time you unlock your screen, try asking yourself one quiet question before you tap:

https://play.google.com/store/apps/details?id=com.sudoajay.descroll&pcampaignid=web_share

What did I actually come here to do?

Meet 123sudo: privacy-focused desktop apps for AI, media, screenshots and music

KAT | Co-founder, 123sudo — Mon, 09 Nov 2026 06:53:07 +0000

Most everyday tasks are small. Convert a file. Save a clip. Grab a screenshot. Ask a model a question.

The tools around those tasks often aren't. An account to create, a file to upload, a dashboard to learn, a subscription to remember.

I'm KAT, co-founder of 123sudo. We're a small team from Uttarakhand building desktop and mobile apps for AI, media, and files. This post is a straightforward introduction: what each app does, what runs on your device, what connects to the internet, and how licensing works.

The design rule behind every app

Before we build a feature, we try to answer one question:

What does this task actually need, and what can it do without?

A file conversion needs your file and your CPU. It doesn't need your email. A screenshot history needs storage on your machine. It doesn't need a cloud account.

That rule doesn't mean every feature is offline. Downloading needs a network. Cloud AI needs a network. The goal is to be specific about which is which.

The apps

9xchat: an AI workspace with cloud and local models

9xchat brings multiple AI models, persistent memory across models, notes, and agentic workflows into one workspace. You can start chatting without setting up API keys first, or bring your own.

Where data goes: with a cloud model, your prompt and attachments go to the provider of the model you chose. With a model running on your machine, that processing stays on your device.

9xconvert: local file conversion

9xconvert converts video, audio, and files locally on Windows, Mac, and Linux. The conversion itself doesn't involve uploading your file to a conversion server.

The free version includes 5 outputs per 24 hours. Pro adds unlimited conversions and URL downloads.

9xbuddy: a media downloader without the trap buttons

9xbuddy has been around in various forms for years. On our apps, extraction runs locally in the app, and page and media URLs aren't sent to 123sudo. Downloading supported media still requires a connection to the source.

9xsnap: screenshots, recordings and clipboard history

Our newest desktop app, available on Windows, Mac, and Linux.

9xsnap keeps one searchable history of what you copy, the screenshots you take, and the screen recordings you make. A few things developers may find useful:

Clipboard history for text, links, images, and files
Region, full-screen, and fixed-size screenshots
Screen, window, or region recording with optional mic narration
Annotation: arrows, text, highlights, numbered steps, blur, and pixelate
Trim recordings and export MP4, WebM, or GIF
Exclude chosen apps from your history, and hide detected secrets in previews
Lock your history with a PIN or password

Where data goes: clips, screenshots, and recordings stay on your device. 9xsnap doesn't upload them or sync them to a cloud. Licence checks connect to 123sudo with your UID.

The free version opens your 50 newest retained entries. Pro unlocks the full history, the editors, and export.

9xlyrics: lyrics while you listen

A smaller, lighter app for floating lyrics while music plays.

How licensing works: UIDs, not passwords

Our apps use a UID, a unique identifier, rather than an email-and-password account. The same UID can identify you across supported 123sudo apps, but each app's licence is separate.

To be clear about what this means: a UID is still an identifier. It reduces account friction. It isn't a claim of anonymity, and payment processing has its own requirements. Each app's privacy policy explains what's sent and when.

Try first, then decide

Every app has a free way to start. We'd rather you test a tool on a real task than buy it on a promise.

Bundle offer this 11.11

If you're picking up software this 11.11, our lifetime bundles are live for a short time. Each one includes Lifetime Pro on every app in it: pay once, never renews, one UID per app.

At the time of writing:

Bundle	Apps	Price
Creator pack	9xbuddy, 9xconvert	$25.18 (≈ ₹2,518)
AI + Media	9xchat, 9xbuddy, 9xconvert	$33.57 (≈ ₹3,357)
All apps	9xbuddy, 9xconvert, 9xchat, 9xlyrics, 9xsnap	$51.47 (≈ ₹5,147)

UPI is supported, and there's a 3-day money-back guarantee. The live page has the current prices and countdown.

One distinction worth stating plainly: a lifetime app licence is not unlimited cloud AI usage. Cloud models still have usage costs.

See the bundles on 123sudo

What we're still working on

We're a small team, and every app is still improving. If something behaves differently from what this post describes, or a privacy explanation isn't clear enough, tell us through in-app support. That feedback shapes what we fix next.

Disclosure: I'm a co-founder of 123sudo, the team behind these apps. This is a product introduction, not an independent review.

Written with help from 9xchat, our multi-model AI workspace, and my 123sudo Head of Growth agent. The views are my own.

When you install a new desktop utility, what's the first thing you check: permissions, network access, pricing, or something else?

Why we build apps that don’t ask for your email

KAT | Co-founder, 123sudo — Wed, 04 Nov 2026 08:10:42 +0000

Why we build apps that don’t ask for your email

The product decision behind UID-based licences, on-device file processing, and one honest exception: cloud AI.

A file converter asks you to create an account.

You only want to change a file’s format. But before you can do that, you’re asked for an email, a password, and permission to upload the file somewhere.

I’ve often wondered: does this task really need to know who I am?

I’m Kanchan(, co-founder of 123sudo. We’re a small team building practical apps for AI, media, and files. One decision we’ve made across parts of our portfolio is to use a UID for app licensing instead of requiring an email-and-password account.

That doesn’t make every app anonymous, or mean every feature works offline. It does make us ask a useful question:

What information does this feature actually need to do its job?

A simple task can involve more data than you expect

Imagine a student trying to convert a lecture recording before a deadline.

They find a website, upload the file, and create an account because that’s the path the site gives them. The task is simple. The data journey may not be.

Or imagine a small shop owner trying to save a product video. A misleading download button opens more tabs, and the real task gets lost in ads and redirects.

These are illustrative scenarios, not customer testimonials. They reflect the kind of friction we want our tools to avoid: making a simple job feel complicated, or asking for more than the job seems to require.

Start with the task, then ask what data it needs

A useful design question for any app is:

What does this feature need to know, and what can it do without?

A local file conversion, for example, can process a file on the user’s computer. 9xconvert’s desktop app is designed to convert files locally on supported desktop platforms. That’s different from uploading a file to a conversion server.

But it’s important not to stretch that into a claim about every feature in every app. Some tasks need an internet connection. Some services involve third-party providers. Each product and workflow has to be described accurately.

For 9xchat, if you choose a cloud model, your prompt and any attachments you include are sent to that model provider for processing. If you choose a model that runs locally, that processing happens on your device. The two paths are different, and users deserve to understand the difference.

Why use a UID instead of email and password?

For licensing in supported apps, we use a UID, a unique identifier. It can identify you in supported 123sudo apps, but it does not mean the apps or payments are anonymous. App licences are separate, and payment processing still has its own requirements.

The reason for this design is practical: using a utility app shouldn’t automatically require creating another email-and-password account.

That’s one choice we’ve made to reduce friction. It isn’t a blanket privacy guarantee. A UID is still an identifier, and the details of how it’s used matter.

Privacy is a set of product decisions

“Privacy-first” is easy to put on a landing page. It’s more useful to explain what happens in a real workflow.

For each feature, we should be able to answer:

Does the task run on the device or use a server?
What information leaves the device?
Which service receives it?
What does the app store locally?
Does the task need an account or identifier?
Where can a user read the relevant policy?

If we can’t answer those questions clearly, we shouldn’t hide behind a broad slogan.

That standard applies to our own apps too.

Offline support is about access, not nostalgia

Our team comes from Uttarakhand, where reliable internet isn’t something everyone can take for granted in every place or moment.

An app that can do some work on the device can be more useful when the connection is weak or unavailable. It can also reduce how much needs to travel over the network.

But “offline” shouldn’t be used loosely. Downloading something from the web requires a connection. A cloud AI model requires a connection. The right claim depends on the specific feature.

I’d rather say exactly what works offline than make a sweeping promise.

A small Diwali note

We’re currently running a Diwali bundle offer on 123sudo. The bundles include Lifetime Pro licences for the listed apps, with one UID per app. The page shows the current prices, bundle contents, payment options, and offer terms.

https://123sudo.com/bundle

A lifetime app licence is not the same as unlimited cloud AI usage. Please check the current bundle details before purchasing.

See the Diwali bundles at 123sudo

The question I keep coming back to

When we build a feature, I want us to ask two things:

What is the simplest way to help someone finish this task?
What data does that solution genuinely require?

We won’t always have the same answer. A local converter and a cloud AI model have different needs. The important thing is to be specific and transparent about those differences.

That’s the kind of software I want our small team to keep building: useful tools that respect people’s time, explain their trade-offs, and don’t ask for more than the task needs.

When an app asks you to create an account or upload a file, what do you wish it would explain first?

Disclosure: I’m a co-founder of 123sudo, the team behind these products. This article describes our design approach, not an independent review. AI helped me organise and edit my notes; the experiences and views are mine.

I studied law. Now we're building a workspace to make legal language easier to understand.

KAT | Co-founder, 123sudo — Sun, 01 Nov 2026 06:08:46 +0000

You can read every word in a legal document and still not understand what it means for your life.

A notice. A contract. A court order. A property document passed down through a family.

Recognising the words isn't the same as understanding their implications.

That gap is one reason we're building aidhar, a legal workspace from 123sudo.

It has now moved from research and design into development. For our team, that's a milestone. For me, it's personal.

My connection to law came before this product

I'm KAT, co-founder of 123sudo.

I studied law through a three-year LLB and have worked in High Courts and district courts. My relationship with the legal system comes from education and experience, not just an interest in building a product for it.

The last couple of years have left me questioning the distance between what people expect from justice and what they experience.

I feel frustrated. But I also feel a responsibility to contribute.

I don't believe an app can fix the judiciary. I don't want to promise that software will clear pending cases or resolve problems much bigger than our team.

But I keep returning to a question we can work on:

What if more people could understand the legal language that shapes their lives?

Understanding is a starting point, not a verdict

Understanding a document doesn't resolve a dispute.

It doesn't establish that someone's interpretation is correct. It doesn't replace advice from a qualified lawyer.

But clearer explanations can give someone a starting point for asking questions.

What does this term mean?
Which part of this document needs professional attention?
What information should I bring to a conversation with a lawyer?
What am I still uncertain about?

I want us to build for that moment: when someone is looking at something important and doesn't know where to begin.

Why our team is taking this on now

At 123sudo, we've been building and improving practical software, including 9xchat, our multi-model AI workspace.

That experience has given us more confidence to take on a problem with a different kind of responsibility.

A confusing interface is frustrating. A confident but misleading legal explanation can have more serious consequences.

So this isn't a case of taking a general-purpose chat interface, changing its name, and calling the problem solved.

We're at the beginning of development. There are questions we need to work through carefully.

The design questions I'm carrying into development

These are principles I want us to explore, not a list of features we're claiming to have shipped.

1. Can we make language clearer without changing its meaning?

Simplifying language is not the same as removing nuance.

A short explanation can become misleading if it leaves out a condition, exception, or limitation that matters.

The goal should be understandable language, not simplicity at any cost.

2. Can people distinguish an explanation from legal advice?

A user shouldn't need to read between the lines to understand the limits of a tool.

I want us to be clear about what an explanation does, what it doesn't establish, and when someone needs professional help.

That distinction belongs in the experience, not only in a disclaimer at the bottom.

3. How should uncertainty appear?

An answer that sounds confident isn't necessarily reliable.

We need to think carefully about how the workspace communicates missing information, ambiguity, and points that need verification.

Sometimes the useful response may be a question or an explicit limit, rather than a polished answer.

4. How do we keep the original document in view?

An explanation should help people engage with the source, not encourage them to forget it exists.

I want to explore ways of keeping the relationship between a document and its explanation understandable.

Again, that's a design goal, not an announcement of a completed capability.

Built with people working in law

aidhar began as a personal initiative, but it isn't mine alone.

It's being built by the aidhar team and shaped by conversations with lawyers and the people we hope to support.

I want to contribute something useful to my colleagues in law. I also want someone outside the profession to feel less lost when they encounter unfamiliar legal language.

Those audiences have different needs.

A lawyer working through documents and a person trying to understand a notice shouldn't automatically receive the same experience.

Listening to both matters. So does being honest about what we can build well first.

Language shouldn't become another locked door

India's linguistic diversity is central to how I'm thinking about aidhar.

Understanding a legal concept in one language doesn't necessarily mean someone can understand it equally well in another.

Our ambition is to support more languages over time.

But multilingual support can't just mean adding a language selector. Legal meaning, terminology, and nuance need careful evaluation.

We haven't announced a launch-language list yet. I'd rather start with a scope we can assess properly than promise every language before we're ready.

Why I keep thinking about the next generation

I hope more young people grow up understanding their rights and the institutions around them.

Maybe one day practical legal literacy and fundamental rights will have a stronger place in every student's education.

Until then, there is room for tools that help people begin learning and asking informed questions.

Technology can't create equality before the law on its own.

But we can try to make understanding less intimidating and more accessible.

Where we are today

aidhar is in development. It isn't publicly launched yet.

We're not announcing completed features, a release date, or a promise to replace professional legal work.

We're sharing the problem we care about and the direction we're taking.

For me, this project sits at the intersection of three things: my legal background, our team's experience building software, and a belief that people deserve a clearer understanding of the systems that affect their lives.

From our small team in Uttarakhand, this is the beginning of something we hope will be useful.

aidhar, a legal workspace from 123sudo. Coming soon.

Disclosure: I'm a co-founder of 123sudo, the team developing aidhar. This article shares our motivation and development goals, not legal advice or a description of completed product capabilities.

Written with help from 9xchat, our multi-model AI workspace, and my 123sudo Head of Growth agent. The experiences, views, and direction are my own.

Whether you're a lawyer, developer, or someone who has struggled with legal language, what is one thing a workspace like this must get right?

Starting Is One Skill. Keeping Software Useful Is Another: Meet sudovijay

KAT | Co-founder, 123sudo — Wed, 21 Oct 2026 05:46:23 +0000

A launch tells you that someone built something.

What happens afterwards tells you whether they are willing to keep taking responsibility for it.

That is the part of software development I have come to appreciate most while working with Vijay Singh, known online and on GitHub as sudovijay.

He is the founder of 123sudo, a self-taught full-stack developer, and someone I call VJ.

I’m his co-founder, and I work closely with him. This is my perspective on his journey: how a student building websites became a developer maintaining products for people around the world, and what I have learned from working beside him.

He started before there was a career plan

VJ was born in Chaturani, a small village in Bageshwar district, in Uttarakhand’s Kumaon region. His school years took him to Meerut.

By Class 10, he was already building websites.

He later left college during his first semester and continued learning independently.

It would be easy to turn that into a familiar dropout story. I don’t think that would do his journey justice.

Leaving college did not automatically make him a developer. The years of learning and building afterwards did.

His route was unconventional, but the work still required the same things: understanding problems, learning technologies, making mistakes, and figuring out how to improve.

The meaningful part is not that he stopped attending college. It is that he did not stop learning.

A stack is a toolkit, not a product

Today, VJ works across web, desktop, and mobile.

His publicly listed toolkit includes JavaScript, TypeScript, PHP, Python, React, React Native, Next.js, Node.js, and Laravel, among other technologies.

That gives some context, but it does not explain the whole job.

A user does not experience a framework. They experience whether an installation works, whether a task makes sense, and whether the application helps them do what they came to do.

Working beside VJ has made that distinction clearer to me.

Knowing how to implement a feature is important. So is noticing where the experience becomes confusing, where an assumption fails, or where a seemingly finished feature still needs attention.

Those details are part of development, not something separate from it.

Different tools, different problems

At 123sudo, VJ’s work includes:

9xchat, an AI workspace bringing multiple models and persistent context together.
9xconvert, a practical file-conversion tool.
9xlyrics, another part of the growing portfolio.

More products are coming.

These are not interchangeable projects. Each has its own audience, constraints, and reasons to exist.

What connects them is a practical question:

Can we make this everyday task less frustrating?

That is one of the things I value about his approach. A product does not have to make a grand promise to be worth building. Helping someone complete a task with less friction is a meaningful outcome.

Privacy starts with what you ask for

Another principle VJ cares about is avoiding unnecessary demands for personal information.

That thinking informs the UID-based approach used in supported 123sudo products, reducing reliance on a conventional email-and-password account.

There is an important distinction here: a UID does not automatically make a service anonymous, and it does not mean that no data is collected. Those details depend on the implementation and the individual product’s privacy policy.

The design question still matters:

What information does this product genuinely need to function?

I think that is a useful question for any developer to ask before an onboarding flow becomes a collection of fields nobody has challenged.

It reflects the philosophy behind our work:

Software should solve the problem, respect the user, and get out of the way.

Building internationally means questioning your assumptions

According to our company’s account of its reach, VJ’s software has reached millions of people across more than 200 countries.

These are company-reported figures, not independently audited measurements.

For me, the significance is not only the number. It is the distance between where the work began and where it has travelled.

Someone using a product may have a different device, connection, or workflow from the person building it.

That creates a useful discipline: your own environment cannot be the only definition of “working.”

For a small team, every difference cannot be anticipated. But continuing to learn from those differences is part of the responsibility.

Leadership is also part of the work

I have seen VJ’s dedication extend beyond his own development tasks.

As his co-founder, I believe he has helped raise the quality of our work and helped the people around him grow in confidence and contribution.

That is my experience of working with him, not a quantified performance claim.

In a small team, development does not happen behind a sealed door. Technical decisions meet design questions, communication challenges, and user feedback.

The ability to bring those perspectives together matters.

VJ’s passion for building is visible in the products. His leadership is something I experience in the work we do together.

“I’m more of a private repo”

That line appears on his personal website, and it makes me smile:

“I’m more of a private repo.”

It captures his understated online presence. But it is not the whole person.

Away from software, VJ loves travelling and photography. He has visited multiple countries and enjoys meeting people and connecting with them along the way.

For someone with such a brief online introduction, he has a genuine interest in the people beyond his own world.

He also loves animals. From what I have seen, the affection is mutual.

His mother and family are deeply important to him, too. Those relationships belong in this story because they are part of the person I know, even if they never appear in a developer biography.

And yes, I find him charming and handsome. Consider that an openly biased co-founder observation, not an engineering benchmark.

You may already know his work

There is one product I have deliberately left until this point.

VJ is the builder behind 9xbuddy, including the 9xbuddy website.

If that name is familiar, you may have encountered his work long before you encountered his name.

It also explains why maintenance is such an important part of this story.

A media-download tool interacts with external websites and systems that change. Available formats differ. Previously working behaviour needs investigating again.

For a user, the interaction can be brief.

For the developer, the responsibility continues well beyond it.

That is the background to the newer work at 123sudo: not just experience launching products, but experience returning to them.

Small teams are not limited to small reach

I don’t believe a product used around the world must begin inside a large company.

One person can start something useful. A small team can develop it further. Neither guarantees success, but a modest starting point does not have to set the limit.

For me, VJ is an example of that possibility from India: a self-taught developer whose work has travelled far beyond where he began.

That does not mean he has done everything alone. It means that sustained work from a small team can have an international reach.

The lesson I take from his journey

When I think about VJ’s story, I come back to the difference between starting and continuing.

Starting a website in Class 10 mattered.

Continuing to learn mattered more.

Launching products mattered.

Continuing to maintain them mattered just as much.

There are more products ahead, more things to understand, and more work to improve. His journey is still unfolding.

If you know him only as sudovijay, I hope this gives you a fuller introduction to the person behind the handle.

Explore his work through sudovijay on GitHub and 123sudo.com.

Disclosure: I’m a co-founder of 123sudo and work closely with VJ. This article reflects my personal perspective. Biographical details and reach figures are provided by us; reach figures have not been independently audited.

The Real Cost of AI Isn’t Just the Bill

KAT | Co-founder, 123sudo — Sat, 17 Oct 2026 12:37:41 +0000

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.

I Build What I Wish I Had

KAT | Co-founder, 123sudo — Thu, 15 Oct 2026 12:28:22 +0000

Two years of building useful tools, learning as I go, and finding my place.

In 2012, I sat in my first pharmacy class and watched the professor write on the board.

My education through Class 12 had been in Hindi. Now I was studying pharmacy in English, trying to understand the subject and the language at the same time.

Thanks to my mama (my mother’s brother ) we watched plenty of Hollywood films. I could listen and understand more than I could confidently write.

Unfortunately, Hollywood hadn’t prepared me to write pharmacy exam answers.

I had six backlogs in my first year.

I recovered, completed my degree, and later, after a four-year break, went on to study law in 2020.

Today, I’m building software. 9xchat, the AI workspace Vijay and I built at 123sudo, is now part of how I research, create, and communicate every day.

That sounds like quite a jump. It was.

But the connection becomes clearer when I look at what I’ve spent the last two years doing.

From having ideas to putting them into people’s hands

For the last two years, I’ve been building and launching apps with our team.

Some began with something I needed. Others came from how Vijay worked, a frustration within our team, or a problem experienced by the people we build for.

That’s the part I enjoy: noticing something annoying and asking whether we can make it easier.

Then comes the less glamorous work. Explaining the idea. Discussing it with the team. Making decisions. Testing it. Realising that something obvious to us isn’t obvious to someone else.

Eventually, you put it into someone’s hands.

That changes the conversation. You’re no longer only discussing what the product could do. You’re learning what people actually find useful.

As a co-founder of 123sudo, I work across creative direction, marketing, product ideas, and design. I’m involved in how we present our products and how people experience them.

I don’t code yet. But I’m very much part of building.

We don’t all need the same tools in the same way

Vijay is a developer. His workflow has evolved toward working with his own agent.

Mine looks different.

I might be researching a topic, exploring a visual direction, writing a product explanation, or trying to communicate an idea to our team.

With 9xchat, I use live web search, generate images, and build and deploy agents. I connect favourite apps through MCP and use screenshots, shortcuts, and the floating window to bring the workspace into what I’m doing.

But I don’t wake up thinking about a list of features.

I think about the work in front of me.

Can I understand this topic?
Can I show someone what I’m imagining?
Can I explain this clearly enough for us to move forward?

Vijay and I use the workspace differently because our work is different. Seeing those differences helps us think about the people we’re building for, too.

A small change that used to take two days

I paint passionately, so I often think visually.

Sometimes I can see what I want in my head, but describing it doesn’t quite get me there.

A simple design change used to take my designer two days. Now I can explore those changes myself in Image Studio, see different directions, and work out what I actually want.

It doesn’t replace a designer’s experience. It helps me arrive at the conversation better prepared.

Instead of sending another long explanation, I can show a direction and say, “Something closer to this.”

That’s a small example, but small improvements matter when they become part of your everyday work.

The ideas were mine. The grammar needed help.

The same thing happens with writing.

I have plenty to say. Sometimes too much.

English isn’t my strongest language, and a message can become a mixture of ideas, spelling mistakes, and sentences that made more sense in my head.

My workspace helps me organise that.

Recently, I was talking to someone who runs a furniture business. He described turning something imagined into a physical piece of art.

I understood him because I paint. I wanted to explain how a tool could help him explore a design without replacing the craftsmanship that made it his.

The understanding was already there. I needed help expressing it clearly.

That’s very different from asking a tool to have my opinions for me.

I want help communicating my thinking, not replacing it.

I still read, edit, question, and decide what sounds like me. A polished sentence isn’t useful if it changes what I meant.

I believe useful tools should be affordable

Building something useful is only part of the job. People also need to feel comfortable trying it.

I believe in affordable tools. I don’t want someone to feel they need a large budget just to explore whether a piece of software could help them.

That belief shapes how I think about what we build and how we offer it.

For 9xchat, we use credits that people can top up as needed. Different models and tasks can have different costs. My goal is for those costs to be understandable, rather than another thing users have to guess at.

Affordability isn’t only about the price. I also think about the time and effort it takes someone to get value from a tool.

If people spend more time figuring out the software than doing the task they came for, we still have work to do.

My ambition isn’t to give everyone another app to manage.

It’s to make tools useful enough to earn their place.

The people made the difference

I haven’t done this alone.

Vijay, my family, and our team have been part of this journey. Pangur Care, our wellness brand, also came to life with the support of my family and partner. We started and built it together.

At 123sudo, I’ve found a team that feels like family. They give me space to ask questions, learn, and contribute without pretending I already know everything.

That has changed my confidence.

I’m planning to learn coding and explore vibe coding with Vijay’s help. I’m excited to understand more of how our ideas become working software.

I’m not waiting to know everything before I participate. But I’m not finished learning either.

What I want the next two years to mean

In some ways, these last two years have felt like twenty years of healing within two.

Not because everything became easy, but because I found room to create, work, learn, and feel excited about my life again.

The apps matter to me. So do the conversations, the people, and the confidence that has grown through building them.

My bigger dream is to keep making practical, affordable tools that help people act on what they know and imagine.

Someone shouldn’t need to code to explore an idea. They shouldn’t need perfect English to communicate something valuable. And they shouldn’t need to understand every technical detail before a tool can be useful to them.

They do need room to learn, make decisions, and keep their own judgment.

That’s the kind of technology I want to help build.

Two years ago, I couldn’t have described exactly where this journey would take me. I still can’t.

But I know I want to keep building things we genuinely use, listening to the people we build for, and enjoying the process of becoming better at it.

What’s one everyday frustration you wish someone would build a useful, affordable tool to solve?

KAT — Co-founder of 123sudo and Pangur Care.

I want to compare AI models, not manage another stack of tools

KAT | Co-founder, 123sudo — Mon, 12 Oct 2026 06:39:11 +0000

Sometimes I want a different AI model’s perspective, not another account, billing dashboard, or conversation that starts from zero.

I might be working through a product idea, refining a brief, or asking for a second opinion on a draft.

The task hasn’t changed. But switching tools can turn into its own task.

I’m KAT, co-founder of 123sudo. I work on product experience, creative direction, and growth, and I use our AI workspace, 9xchat, daily.

An older article I wrote focused heavily on token allowances and getting started without API keys. Looking back, I’d explain the problem differently now.

Getting access to a model is only the beginning. Keeping the work connected is the bigger challenge.

Access and continuity are different problems

There are several ways to use AI.

You can use a provider’s own app. You can work through an API. You can run a compatible model locally. Or you can use a workspace that brings multiple models together.

None of those approaches is automatically right for everyone.

For me, the friction appears when one piece of work stretches across several tools:

The project background sits in one conversation.
A useful draft sits in another.
A decision is buried in chat history.
A different model needs the same explanation again.
Usage is spread across separate billing systems.

The problem isn’t having choices.

It’s the work required to move between them.

Why we’re building one workspace

9xchat brings multiple AI models, persistent memory, notes, and agentic workflows into one workspace.

The idea is straightforward: make it easier to continue the work when you change models or move to the next task.

For example, imagine you’re preparing a product brief:

Explore the problem and audience.
Draft a possible approach.
Ask another model to challenge the assumptions.
Save the decisions that matter.
Continue with the next task using relevant context.

That’s an illustrative workflow, not a claim that switching models automatically improves the result.

A second answer can be useful. It can also repeat the first model’s mistakes.

I still need to evaluate the output.

API keys should be a choice, not the opening hurdle

Some people want to bring their own API keys. Others just want to start a conversation without setting up provider accounts first.

9xchat supports both approaches: bring your own keys, or use ours.

You don’t need to log in to try the app. Open it and start chatting.

That doesn’t mean usage is unlimited or free forever. It means the initial setup doesn’t have to begin with an API tutorial.

If you bring your own keys, provider pricing and billing still matter. If you use workspace credits, check the current usage information before starting a large task.

A token balance isn’t a context window

This is something I’d make much clearer than I did in my older article.

These are different concepts:

Usage allowance or credits: what you can spend.
Context window: how much information a model can handle in a request.
Persistent memory: saved information that can support later conversations.

Having a token allowance doesn’t mean every model has a context window of that size.

And a large context window doesn’t mean a fixed amount of money buys the same amount of work across every model.

Costs depend on the model, input and output, and the type of task. Image generation and other operations may use different pricing units.

That’s why I wouldn’t promise that a particular balance lasts a certain number of weeks or produces a fixed number of images.

The useful question is what your actual workflow costs, not how impressive the token number sounds.

Memory should help, not become another pile of old information

Persistent memory matters to me because some context stays relevant across tasks:

Which product we’re discussing.
Who the audience is.
Which decisions are confirmed.
Which constraints the work must respect.

But not everything deserves to carry forward.

An abandoned idea shouldn’t quietly become a requirement. A temporary instruction shouldn’t become a permanent preference.

Memory is useful when the context is relevant and current.

It also doesn’t make model processing free. Depending on the implementation, remembered information may still become part of the input a model processes.

I think of memory as a continuity feature, not a guaranteed cost-saving mechanism.

Cloud and local models need an honest distinction

“One workspace” doesn’t mean “one data path.”

When you choose a cloud model, your prompt and the attachments you send go to that model’s service for processing.

A model running on your own machine can provide a different processing path. But connected tools and other actions may still involve external services.

Before using confidential material, check the relevant settings and policies for the model and tools involved.

I wouldn’t treat the word “local” or “private” as a substitute for understanding what a specific task sends and where.

What I’d check before adopting any AI workspace

Whether you’re considering 9xchat or another tool, I’d start with these questions:

Model access: Does it support the models you actually need?
Context: Can you understand and manage what carries into later tasks?
Costs: Can you see how usage is charged?
Data handling: What stays on your device, and what goes elsewhere?
Actions: What can connected tools do, and what controls apply?
Workflow fit: Does it reduce steps in your real work, or just add another interface?

You don’t need every feature.

You need the combination that makes your recurring work easier to manage.

The goal isn’t collecting more models

It’s being able to use the right tools without rebuilding the workspace around every task.

That’s the direction we’re taking with 9xchat.

Not a promise that one interface eliminates every trade-off. Not a claim that every model is equally suitable.

Just a practical goal: less setup, clearer usage, and better continuity.

Disclosure: I’m a co-founder of 123sudo, the team behind 9xchat. This is a product perspective, not an independent review.

AI helped structure and draft this article.

Explore 9xchat through 123sudo.

Before an AI agent acts, show me what it can change

KAT | Co-founder, 123sudo — Sat, 10 Oct 2026 06:27:09 +0000

“Allow this action” is a difficult question to answer when I can’t see what the action actually does.

Which file will change? Who will receive the message? Is the agent preparing a draft, or sending it?

Before I approve an action, I want those details in front of me.

I’m KAT, co-founder of 123sudo. I work on product experience, creative direction, and growth. A recent conversation here on DEV got me thinking about what an understandable approval screen should show.

This is a product-design perspective, not a security specification or an announcement of existing features in our AI workspace, 9xchat.

Connected doesn’t mean approved

Connecting a tool makes certain actions available. It shouldn’t make the user guess which of those actions an agent is about to take.

An email tool might let an agent:

Read a message.
Prepare a reply.
Save a draft.
Send the reply.

Those are different outcomes.

The same goes for files. Reading a document is different from overwriting it.

I’d want the interface to distinguish between:

This tool is connected.

And:

This action is permitted for the current task.

That distinction matters more to me than a long list of integration names.

Show the outcome, not just the tool

A technical tool name may help a developer understand what’s happening. It may tell another user very little.

Compare these two approval requests:

Allow send_message?

And:

Send this email to Alex at an external address, with the attached project brief?

The second gives me something I can check.

For an email, I’d want to see:

The sender account.
The recipient, including CC and BCC.
The subject and message.
Any attachments.
Whether the action sends immediately or saves a draft.

For a file edit, I’d want:

The target file and its location.
The proposed changes.
Whether it creates, edits, or overwrites something.
Any relevant limits on undoing the action.

The preview should answer a simple question:

What will be different after I click approve?

Build the preview from the pending action

Someone in the DEV discussion made a useful distinction: the preview should come from the tool call itself, not just the agent’s summary.

That changed how I thought about the interface.

An agent might say:

I’ll send the brief to Alex.

But the pending action contains the details that matter: the actual address, message, and attachments.

A reassuring explanation isn’t enough if it doesn’t match those details.

My design preference would be to show a readable explanation alongside the relevant fields from the pending action.

The explanation helps me understand. The fields help me verify.

And that preview alone isn’t a guarantee. The system still needs to enforce that the approved action is the one executed.

Approval should belong to a specific action

Imagine I approve a message.

Then the agent changes the recipient, adds an attachment, or rewrites the body.

Does my earlier approval still apply?

For changes like those, I’d expect a fresh approval.

Otherwise, I’m approving one thing and potentially getting another.

This is a principle I’d want the system to enforce, not just a promise written on the button:

Approve this action, with these details.

Warnings need words, not just colours

A coloured border might draw attention to an external recipient or a file outside the task folder.

But colour alone shouldn’t carry the meaning.

I’d pair it with a clear label:

External recipient
Outside the selected task folder
Overwrites an existing file

Then explain why it matters.

For example:

This attachment will be sent to an address outside your organisation.

That tells me what I’m deciding. A red icon by itself doesn’t.

More prompts don’t automatically mean more control

I don’t want an approval dialog for every tiny step.

If every action gets the same warning, the interface risks teaching people to click through without reading.

I’d rather see understandable boundaries at the start of a task, followed by specific approval requests when the agent reaches an action outside those boundaries.

For example:

You can read the selected files and prepare a draft. Ask before editing files or sending messages.

That’s easier to reason about than an unexplained “full access” setting.

The right boundary will depend on the task, the sensitivity of the information, and the tool. Reading can matter too, especially when private information is involved.

What I’d want to see before execution

My starting checklist would be:

Access: What information can the agent read?
Actions: What can it create, change, send, or delete?
Targets: Which accounts, recipients, files, or folders are involved?
Approval: Which actions require me to review the exact details?
Changes: What happens if the action changes after I approve it?

This isn’t a complete security model. It’s a way to make the user’s decision less vague.

Clear previews still need reliable permission checks and execution controls behind them.

Trust needs something concrete to check

As a co-founder helping shape an AI workspace, I’m interested in interfaces that make an agent’s actions understandable.

Not because the agent should explain every internal step.

Because when it changes something on my behalf, I should understand the change.

“Trust me” isn’t a useful approval screen.

Show me the action. Show me the target. Let me check what matters.

Disclosure: I’m a co-founder of 123sudo, the team behind 9xchat. These are design principles I’m exploring, not claims about features currently available in the product.

AI helped structure and draft this article.

What would you need to see before approving an AI agent’s action?

India Committed ₹10,372 Crore to AI. Here's How to Make AI Work for You.

KAT | Co-founder, 123sudo — Thu, 08 Oct 2026 09:55:06 +0000

A practical guide for individuals using Claude, GPT, Gemini, and DeepSeek, and what a bootstrapped Indian team learned building a workspace around them.

Claude for writing. GPT for brainstorming. Gemini for research. DeepSeek because someone recommended it for reasoning.

Four tabs. Four logins. Maybe four bills.

And another fresh conversation that needs an introduction to your work.

The models are better than ever. The way we use them can still be messy, expensive, and repetitive.

India is betting big on AI

In March 2024, the Union Cabinet approved the IndiaAI Mission with an outlay of ₹10,371.92 crore over five years. It covers compute, datasets, homegrown foundation models, startup financing, and skills.

Homegrown model efforts are growing too. Sarvam AI is one of the organisations selected to build indigenous foundation models, with stated financial and compute support of ₹246.72 crore. BharatGen is building multilingual models for Indian languages.

One clarification: ₹10,371.92 crore is an approved programme budget, not a claim that the full amount has already been spent. It also does not represent all public and private AI investment in India.

National investment builds the foundation. But individuals face a more immediate question:

How do I use these models without wasting money, losing my work, or sharing more information than the task requires?

That is the question our team at 123sudo has been working on with 9xchat.

Disclosure: I am a co-founder of 123sudo, the team that builds 9xchat.

Problem 1: You are paying for tools you barely use

A monthly subscription costs the same whether you use it every day or twice a month.

Try this: list every AI subscription you pay for. Next to each one, write how often you actually used it last month.

You may find overlap or a tool you no longer need.

How 9xchat approaches it: one workspace with cloud models through OpenRouter, local models through Ollama, and the option to bring your own supported API keys.

Its credit-based cloud workflow uses transparent credits, helping you understand where your usage goes.

That does not automatically make AI cheaper. Heavy use of advanced models still costs money. The benefit is being able to assess your spending rather than guessing.

Problem 2: You use an expensive model for every task

Not every task needs the most powerful model.

Reformatting a list does not require the same capability as reviewing a complex code change.

A practical routing habit

Model families contain different versions with different strengths and prices. Test the actual model on your task instead of assuming the brand name guarantees quality.

In 9xchat, you can work with multiple models without opening another app.

Using lower-cost models for straightforward tasks can reduce spending, provided the output still meets your needs.

Problem 3: You keep re-explaining the same project

This is the hidden effort that rarely appears in a pricing table.

You open a new conversation, paste in your project brief, find the earlier decisions, and explain what you have already tried.

I am a co-founder who does not code, and I am close to ₹1 lakh in estimated AI spending this year across the tools and agent skills I use.

That made me pay closer attention to how much work involved repeating context rather than making progress.

How 9xchat helps keep context connected

Persistent memory across models helps carry important information forward.
Separate memory scopes distinguish broader context from individual chat and agent context.
A knowledge base lets you save project material and access it from chats.

Memory does not make context free. Information sent to a cloud model can still contribute to token usage.

The useful goal is less repeated setup and more relevant context, not a promise that every conversation will cost less.

A habit that helps regardless of tool: keep a short, current project brief covering what you are building, who it is for, what is decided, and what comes next.

Problem 4: You are not sure where your data goes

A locally stored chat and a locally processed AI request are not the same thing.

Here is the distinction worth understanding:

Local storage: 9xchat describes chats, notes, and memory as stored on your device.
Local models: Ollama lets you run supported models on your own machine.
Cloud models: using a cloud model requires sending the request and relevant context to the services involved in processing it.
Tools and agents: web access, publishing, and other external actions may transmit information even when part of your workflow runs locally.

A local model can help limit external processing, but review the complete workflow, not just the model selector.

For cloud requests, share only what is necessary and check the relevant provider and routing-service policies.

Privacy is a set of choices and boundaries, not a magic setting.

Problem 5: The answer is not the finished task

AI gives you a plan. You still have to organize the steps, prepare the draft, review it, and move it toward completion.

How 9xchat approaches the gap

Skills provide a repeatable approach to a job, such as reviewing code for bugs and missing tests.
Agentic workflows help carry out multi-step tasks, including publishing a post.
A floating copilot window makes AI accessible alongside your browser, IDE, or terminal.
Voice notes can turn spoken ideas into structured Markdown.

Publishing and scheduling are different capabilities: 9xchat can publish as part of an agentic workflow, but it does not currently schedule posts.

You still need to review facts, permissions, and consequential actions. The point is to reduce the manual handoffs between a useful answer and a completed task.

Your 10-minute AI cost reset

Whatever tool you use, try this:

List your AI subscriptions and how often you used each one.
Review overlapping plans before cancelling or pausing anything.
Write one current project brief and save it somewhere reusable.
Choose a lower-cost default model for straightforward tasks.
Review data handling before working with sensitive information.
Check usage after one week, alongside the quality and usefulness of the results.

Look at cost per useful outcome, not just cost per response.

A cheap answer that requires repeated corrections may not be the better deal.

Built in India, for everyday work

123sudo is a bootstrapped team from India. We have been building everyday internet utilities since 2014, with a simple philosophy:

Software should solve the problem, respect the user, and get out of the way.

Our product page reports that more than 500 people have used 9xchat across 82 countries, with availability listed for Windows, macOS, Linux, iOS, and Android.

We are not claiming that 9xchat currently supports Sarvam or BharatGen. Their work belongs to the model layer. Our focus is the workspace layer: helping individuals use AI in a more connected way.

India's AI future needs compute, research, and homegrown models. It also needs practical products that help a student, freelancer, creator, or founder turn those capabilities into useful work.

Try 9xchat with a 3-day free trial and welcome credits at 123sudo.com.

Sources

I don’t code. My AI usage still reached 145 million tokens.

KAT | Co-founder, 123sudo — Wed, 07 Oct 2026 03:53:15 +0000

When I checked my AI usage dashboard, two numbers made me stop:

145 million tokens. 8,377 requests.

I don’t code.

Vijay, our founder and main developer at 123sudo, handles that side of the work. These were my conversations.

In September, I wrote about those figures on Medium. My 9xchat dashboard showed an all-time cost of about $637. That wasn’t my complete AI spending for the year. Including other AI tools and services, I estimated I was approaching ₹1 lakh.

Those are different totals: one was a dashboard figure, the other my broader spending estimate.

But both raised the same question:

What was I actually getting from all this usage?

It wasn’t one enormous task

I’m a co-founder working on product experience, creative direction, and growth.

My AI usage comes from the ordinary work of building and growing products:

Working through product ideas.
Researching an audience.
Drafting launch content.
Finding clearer language for a confusing feature.
Comparing possible approaches.
Asking another model to challenge an idea.

Sometimes I need a draft. Sometimes I need a second opinion. Sometimes I haven’t understood the problem well enough to ask a useful question yet.

The total wasn’t one dramatic project. It was thousands of small interactions.

That’s what surprised me: you don’t need to be writing code for AI usage to become a meaningful expense.

A big number doesn’t automatically mean waste

Some conversations help me make a decision. Others help turn an unclear idea into something we can work on.

That can be valuable.

I don’t want to reduce AI usage simply to make a dashboard number smaller. A useful conversation can be worth more than several cheap but unhelpful ones.

The question I’m learning to ask is:

Did this interaction move the work forward?

A finished brief, a clearer decision, or a usable draft is different from another page of options I won’t act on.

Starting over is the part that frustrates me

I expect to explain a new task.

What gets tiring is rebuilding the background around it:

What is 123sudo? Which product are we discussing? Who is it for? What have we already decided? Which claims should we avoid?

Then I switch models for another perspective and repeat the setup.

There are two costs here:

The usage cost, depending on the model, billing method, and context processing.
My time and attention, spent reconstructing information instead of doing the next piece of work.

I don’t have a measured savings figure for fixing that. But it’s a real source of friction in my workflow.

Remembering context isn’t the same as making it free

This is an important distinction.

Persistent memory can reduce the amount of background I manually repeat. It does not mean remembered information becomes free for a model to process.

Relevant context may still be included in a request. Its cost depends on the implementation, model pricing, caching, and other factors.

So I think of memory as a continuity tool first, not a guaranteed token-saving mechanism.

The useful question is:

Are we bringing the right context into the task, or just carrying more text?

Three questions I’m using to judge my AI work

Seeing those usage figures made me think more carefully about what a productive session looks like.

1. What am I trying to finish?

“Help me think about marketing” can become an endless conversation.

“Help me draft a product brief for this audience, using these confirmed facts” gives the work a clearer destination.

Exploration still has a place. I just want to notice when I’m exploring and when I’m supposed to be finishing.

2. What context actually matters?

A model doesn’t need every discussion we’ve ever had.

It needs the information relevant to this task: the audience, goal, constraints, and current decisions.

Old plans can be just as distracting as missing information.

3. Is another response helping, or delaying the decision?

This is the uncomfortable one.

Sometimes asking for more options is useful. Sometimes I’m asking because choosing an option feels harder than generating another ten.

AI can help me think. It can also give me a very convincing way to stay busy.

Why this matters to what we’re building

Disclosure: I’m a co-founder of 123sudo, the team behind 9xchat. I use it daily, so this is not an independent product review.

9xchat brings multiple models, persistent memory across models, and transparent credits into one AI workspace.

Those capabilities matter to me because my work doesn’t stay in separate boxes. Research turns into a product decision. That decision turns into a brief. The brief turns into content.

I want the work to continue without rebuilding the background at every step.

But having a workspace doesn’t remove my responsibility to give clear instructions, review outputs, and decide what deserves another iteration.

A tool can support a better workflow. It can’t make every conversation worthwhile.

The lesson wasn’t “stop using AI”

It was stop treating usage as proof of progress.

145 million tokens tells me how much activity happened. It doesn’t tell me how many good decisions I made, what shipped, or which conversations were worth their cost.

That’s the distinction I want to get better at.

I don’t code. AI has still become a substantial part of my working life.

Now I’m trying to judge it less by how much I use it, and more by what the work becomes.

AI helped structure and edit this article. The usage figures and personal observations are mine.

What tells you an AI session was worth it: a finished task, a better decision, time saved, or something else?

I stopped re-explaining my project to every AI model. Here's what changed.

KAT | Co-founder, 123sudo — Tue, 06 Oct 2026 03:09:23 +0000

I'm Kanchan satyal , co-founder of 123sudo. We're a small studio that makes everyday software: 9xchat, 9xbuddy, 9xconvert, and 9xlyrics.

Running a portfolio like that means my days don't stay in one lane. In one afternoon I might write a Threads post, check a product decision, draft a YouTube description, and look at a weekly analytics recap.

For a long time, every one of those tasks began the same way: explaining who we are again.

The ritual

If you use more than one AI tool, you probably know it:

Open a new chat.
Paste in the background on your project, your brand, your rules.
Get a useful answer.
Switch to a different model for a different job.
Start over at step 2.

None of the models were bad at their jobs. The workflow around them was broken. I was the only thing connecting them, and I was doing it by copying and pasting.

What the context actually was

Re-explaining your context isn't one paragraph. For us it includes:

Which products are active, and which ones are discontinued
How we write: no em dashes, one CTA per post, no made-up numbers
What we've learned already, like which post formats have worked for us
Corrections, like "always refer to me as co-founder"

Context like this changes all the time. A product gets dropped. A post format starts working. A rule gets updated. Keeping it current across different tools by hand never worked for long.

What changed

We run 123sudo's growth work inside 9xchat now: strategy, Threads replies, captions, research, and planning. Work is sorted into projects for 123sudo, 9xchat, and marketing.

Here's what changed in practice:

1. It remembers our decisions.
When we removed a product from our portfolio, that decision was saved once. I don't have to bring it up in every new chat.

2. Our wins get written down.
When one Threads format clearly beat the others, the pattern got saved: an industry-wide hook, a quick list of examples, what it costs the user, what we do differently, a strong close. The next draft builds on it instead of starting from zero.

3. I can find old work again.
I once needed a caption we'd written weeks earlier for a 9xbuddy post. I didn't dig through tabs. I asked, and it found the original chat and the exact text.

4. I can let Auto choose, or choose for myself.
For routine work, I use Auto. A GPT model selects a model for the task and routes it there, so I don't have to make that decision every time. When I want a specific model, I choose it manually. 9xchat also lets users bring their own API key.

Because memory carries across models, switching doesn't mean starting from a blank page.

5. My teammate can pick up where I left off.
When I need to hand work over, I share the 9xchat conversation with a teammate. They continue in that same chat, with the discussion already there. I can share agents too, so they use the same setup instead of rebuilding it.

It feels less like passing around a document and more like handing off an active workspace.

What didn't change

Memory isn't magic, and I don't want to oversell it.

I still review everything. A tool that remembers your rules can still get them wrong.
Judgment is still mine. The AI drafts and organizes. Taste and decisions stay with me.
Memory needs maintenance. And I learned that one the funny way.

We changed our welcome offer from $1 worth of AI tokens to 1 million AI tokens. Yet the assistant kept mentioning the $1 offer, even after I corrected it.

Apparently, I'd built a workspace that remembered our brand a little too well.

Persistent memory isn't the same as automatically keeping every fact current. When something changes, I update the memory and check that drafts reflect it. Memory reduces repetition. It doesn't remove the need for judgment.

The real lesson

Better models matter. But for daily work, the bigger difference came from a workspace that remembers what I already told it.

A smarter model that forgets you every time still makes you do the remembering. I'd rather spend that time shipping.

That's why we built 9xchat: one workspace, cloud and local models, persistent memory across them, and transparent credits.

Try 9xchat. Start instantly with 1 million AI welcome tokens, no API key required.