Skip to content
LittleStone— home
LittleStone

What is on the shelf.

We are currently developing our first generation of applications. While nothing is available on the App Store or Google Play just yet, we are actively building the foundation for our upcoming portfolio.

Instead of showcasing mockups or committing to arbitrary release dates, we prefer to let our work speak for itself when it is ready. Below, you will find details on what to expect when our upcoming portfolio of products is officially launched.

Upcoming Releases

This is where our released products will be showcased. We are working hard behind the scenes and look forward to sharing our upcoming portfolio with you.

What each entry will carry

  • The name, and one line saying what it actually does.
  • A plain description, including what it deliberately does not do.
  • The platforms it runs on.
  • Whether its AI runs on your device or calls a hosted model, feature by feature.
  • A link to the store listing, and nothing in its place until there is one.

If you want to hear when the first one arrives, write to us and say so. There is no mailing list to join and no countdown to watch — a person reads it and a person replies, and if what you are waiting for is not something we are building, you will be told that rather than kept warm.

Two lines of work.

Both are ours. We pick the problem, we carry the cost of picking it wrong, and there is nobody standing between us and the person using the result.

AI applications for iOS and Android

Separate, unrelated applications rather than one suite. Each is built around a single task, uses a model where a model genuinely helps, and has to be worth installing on its own to someone who never opens any of the others.

Who it is for: People who want one task done, and who would rather not hand over more of their life in order to get it done.

What is different about it

  • They are not a family. No shared account, no cross-selling between them, and no reason to own more than the one you came for.
  • The model is a component, not the point. Where a feature works better without one, it does not get one.
  • Where a feature can run its model on the device, it does. The work happens on hardware you already own, and it keeps working in a tunnel.
  • Anything that leaves the phone is disclosed in the interface, at the moment it happens, not only in a policy you were never going to read.
  • Accessibility is in the first build. Text scaling, focus order, screen reader labels and contrast are part of finished, not part of a later release.

Where it stands

Nothing is on the App Store or Google Play. There is no public beta, no waiting list, and no date we are willing to give you.

Software for small businesses

Software we build once and licence to the businesses that need it, rather than a project delivered to a brief and handed over. We choose the problem, we own the roadmap, and we carry the maintenance.

Who it is for: Small businesses, across trades rather than inside one industry. The common thread is size and not sector: somewhere important is being run on a spreadsheet, an inbox, and one person who remembers how it all fits together.

What is different about it

  • We are the vendor, not a contractor. There is no handover cliff and no rebuild by whoever arrives after us.
  • The paperwork exists before the sales conversation. The terms, the retention schedule and the sub-processor register are published on this site already, with nothing to sell yet.
  • The same on-device and hosted split applies to business features, and a buyer can see which is which before signing anything.
  • It has to work for the person doing the job at four in the afternoon, not only for the person watching the demo.

Where it stands

Nothing is generally available and nothing is for sale today. We are not taking client briefs either — that is a different business, and not ours.

Where the model runs is the whole question.

Some features run their model on your device. Others send your input to a model hosted by another company. It is not one or the other across the range, and it is not something you should have to guess at, so the answer is given per feature rather than per application.

On the device

The model ships inside the application and runs on your phone. Your input is processed by hardware you own.

When we use it
Whenever a model small enough to run on the device can do the job to a standard we would accept ourselves. Small and good enough beats large and remote.
What it means for you
Your input does not leave the phone for that feature. It works with no connection, and there is no retention question to answer because we never received anything.

A hosted model API

The application sends your input over the network to a model run by another company, and shows you what comes back.

When we use it
When the work needs a model too large to run on a phone. That is a real trade rather than a detail, and we treat it as one.
What it means for you
Your input reaches a third company. That company is named, along with its region, how long it keeps input, and whether it may train on it. Features that work this way say so where you use them.

How the choice gets made

  • On-device is the default. Hosted has to be argued for, feature by feature, and the argument is recorded.
  • A feature that calls a hosted model sends the least it can. Not the whole document when a paragraph answers the question, and not your identity when the task does not need it.
  • The split is disclosed twice: in the privacy policy for the people who read policies, and in the interface for everyone else.
  • We do not describe a feature as private without saying where it runs. That single sentence is the reason any of the rest of this is worth publishing.

Named means named. The hosted provider is [MODEL_PROVIDER] — this detail has not been filled in yet, and the sub-processor register carries its region, how long it keeps input, and its position on training. Where an entry there is still a marked placeholder, it is because we have not signed the answer yet — and a placeholder that shouts is better than a reassuring sentence that is not true.

None of this describes software you can install today, because there is none. It is the standard the products are being built to. When they arrive, check it against what they actually do.

Rules we would rather be held to.

With an empty shelf, this is the part worth reading. It is what a product would be judged against once it is on that shelf, published before there is one so it cannot be quietly adjusted to fit whatever we end up shipping. Every rule below is either already visible on this website or written into the published policies.

  • On the device, where the feature allows

    On-device processing is the first option considered for every feature, not the fallback when a budget runs out. A feature calls a hosted model only when the work genuinely cannot be done well on a phone, and that has to be argued rather than assumed.

    Less of what you type and photograph leaves your hardware, the feature keeps working with no signal, and there is nothing for anyone to retain because nobody received it.

  • What leaves the device is named

    Every piece of data that goes anywhere is written down: what it is, where it goes, which company runs that place, and how long they keep it. Not "selected third-party service providers".

    You can read the sub-processor register before you install anything, and find a list of companies rather than a list of categories.

  • Accessibility is built in, not retrofitted

    Contrast, focus order, keyboard and screen reader support and text scaling are part of the definition of done, and the checks run in CI rather than in somebody's memory. This website is built the same way, which is the part you can verify right now.

    The product works with the settings you already depend on, on the day it ships, instead of two releases after somebody complained.

  • No advertising, and no tracking for it

    We do not sell advertising, we do not embed an advertising toolkit, and we do not build an identifier so that anyone can follow you between applications. That decision is made once, at the start, because it is almost impossible to reverse later.

    There is nothing here to opt out of, because the thing you would be opting out of was never built.

  • Deletion means deletion

    Removing your data is a function inside the product, not a support ticket you have to chase. Anything held on the device goes when you delete the application, and anything we hold goes when you ask.

    Leaving is as easy as arriving. That is the only thing that makes the rest of these promises worth anything.

  • The policies are published before the product

    The privacy policy, the terms, the accessibility statement and the sub-processor register are on this site now, while there is nothing to sell. They are still going through legal review and each one says so at the top.

    You can judge us on documents that already exist, rather than on a page of intentions written by the people asking for your trust.

Things worth asking a company with nothing to show.

Why is there nothing in the portfolio?

Because nothing has been released. This page shows what we have shipped, and today that is an empty list. We could fill it with concepts, mockups and names we have not committed to, and it would read better for about a minute — until you tried to find one of them in a store. An empty shelf you can trust is the more useful thing to publish, and it is the version that is still true next month.

When will something be available?

We do not have a date, and we are not going to invent one in order to have something to say. Nothing is on the App Store or Google Play, there is no public beta and there is no waiting list. When a release date exists, it will appear on this page before it appears anywhere else.

Will my data be used to train models?

We do not train models on your content. For a feature that runs on the device the question does not arise, because your input never reaches us. For a feature that calls a hosted model, whether that provider may use input for training is a contract term rather than a hope, and the answer will be written in the privacy policy in plain words before the feature ships. If it is not an answer we can live with, that provider does not get used.

Will it work offline?

Partly, and which parts depends on the feature. Anything running its model on the device keeps working with no connection. Anything calling a hosted model does not, and it will tell you that rather than failing quietly or blaming your signal. We would rather label the boundary than pretend it is not there.

What happens to my data if you shut down?

A fair question to ask a small company before you rely on it, rather than after. Data held on your device stays on your device and goes when you delete the application. For anything we hold, the commitment is notice, an export in a format you can actually open, and deletion afterwards — written into the terms rather than promised on a marketing page.

Will there be a free version?

Some products will be free with no payments in them at all. Others will be paid, through a subscription or a purchase inside the application, and the business software will be licensed under a contract. We have not settled which model each product uses, so we are not going to imply otherwise. What we can say now is that nothing will be funded by advertising or by selling data about the people using it.

How much will it cost?

We do not know yet, and any figure today would be a guess dressed up as a price. When a price exists it will be visible before you install anything rather than after, and a subscription will be as easy to end as it was to start.

Can I use it in my business?

That is the intention for part of the range, under a contract with terms you can send to your own lawyer. Today there is nothing to sign. If you want to be told when there is, write to us and say what you are trying to solve — it is genuinely useful to us, and we will be straight with you about whether we are building anything close to it.

Do you take on client projects?

No. We build our own software and answer for it ourselves. If you are looking for a studio to build something to your specification, we are the wrong company — and we would rather say so at the top of the page than three meetings in.

Anything about how your data is handled is answered at greater length in the privacy policy, including the per-feature table of what runs where.

Follow along on LinkedIn.

We will be sharing our progress, insights, and updates on our upcoming releases. Follow our page to be the first to hear when our portfolio starts rolling out.

Follow us on LinkedIn