AN INDEPENDENT SOFTWARE STUDIO
We build software for ourselves and for businesses, under one rule: a promise that matters becomes a check the build runs. If it cannot fail, it was never a promise — only a hope.
Three products of our own, and one set of rules behind them.
Three products of our own.
Each states a promise — and each has
a check that fails the build if the promise stops being true.
“No ads, no sign-in, no data.”
Free reading for curious children, seven to twelve. 165 articles across six subjects, 15 games, 261 hand-drawn illustrations. It needs no account, and it keeps working offline.
The check
The deploy fails on any external link, any remote image, any unresolved placeholder. A page that could phone home never ships.
“Instrument, don’t advise.”
A self-directed trading workspace. Connect your own broker, research, scan for chart patterns, build rules without writing code, and test them against history. It screens holdings against the AAOIFI Shariah standard. Your money never leaves your broker.
The check
The legal disclosure lives in one file pinned by a test, so an edit cannot quietly drop the broker attribution. No accuracy figures anywhere: reliability is labelled history, never edge.
In private preview. Signup is closed, and no performance is claimed or implied.
“The reminder arrives showing a photo of the actual thing.”
Photograph the pill bottle, the shelf, the form, the door. Write your own words on the picture, choose a time, and at that time the picture comes back. Nothing to type and nothing to describe. Five languages, right-to-left in Arabic.
The check
A test reads the built manifest and fails if the app has gained camera or internet access. It has neither. And when Android will not allow an exact time, Ramz says “about 6:00 pm” rather than pretending.
Built, signed and verified on a device. Not published yet, so there is nothing to install today.
CAPABILITIES
These are the kinds of system we build. We are not describing anyone else’s project — for each one, the engineering is already running in something of ours that you can open and inspect.
Turning messy inputs — filings, feeds, PDFs — into something a system can act on, with every input it cannot read reported rather than quietly skipped.
Language models placed where they earn their keep, inside a harness that measures the output instead of trusting it.
Rules, scoring and gating that a person can audit, a test can hold to account, and a user can be shown the reasons for.
Accounts, roles, billing and third-party integrations on managed cloud — built so the browser never holds what it should not.
Offline-first, accessible, localised, and taken through store review to the point of being installable.
We are a small studio.
Saying yes to everything is how a small
studio ships nothing well.
Design, build, ship and operate it. You get a working product and the reasoning behind it — not a body in a seat, and not a deck.
Decision systems a model gates rather than decorates: agentic pipelines, evaluation harnesses, and the governance that keeps a model’s output accountable.
Before anything is designed — and often it is not the stated one.
A practical blueprint, with the checks named up front.
Working software, in production, with its gates running.
Measure it, correct it, and keep the correction legible.
“Why hire a studio whose shipped products are a children’s reading site and a reminder app?”
Because the products are small and the constraints are not. A children’s site is where a broken privacy promise costs the most; a reminder app is where a silently dropped alarm costs the most. We chose them because they are unforgiving. Behind them sits a trading workspace carrying a twenty-workstream compliance screen, an evaluation harness for the models inside it, and an infrastructure bill we adjudicate line by line. What you are buying is not the apps. It is the discipline that made them — applied to yours.
A principle nobody can check is a preference. Each of these is one we can show you happening, in products you can open.
A rule that lives in a document is a hope. A rule that fails the build is a rule.
BrightKidz refuses to deploy with a remote image in it. ZadVest’s disclosure is pinned by a test. Ramz’s build asserts that camera and internet access are absent.
We do not take what we do not need, and we build so that we cannot.
No analytics or cookies in BrightKidz. No account or server in Ramz. In ZadVest your money never leaves your broker, and the browser never touches a raw brokerage key.
Where the platform will not let us be exact, we say so — in the product, not the footnote.
Ramz says “about 6:00 pm” when the system will not promise better, and reports a reminder the phone dropped. ZadVest publishes no accuracy figure at all.
Nothing we make is designed to exploit attention, fear, or a child.
BrightKidz has no streaks to break and no timer anywhere; its arcade never takes a star away. Ramz carries no advertising, and a feature that shipped free stays free.
Accessibility is a gate the build runs, not a box a designer ticks.
BrightKidz holds AAA on body text. Every Ramz screen survives the largest system font uncut. In one product an ink colour measured 2.32:1 against the darkest skin tones — caught by measuring before anything was drawn.
We correct things in place and in public: the superseded claim stays beside the fix, dated, so the change reads as a change. A stale warning is worse than no warning. Our own market study this month named its kill criteria before it collected any evidence — and then killed the idea its author liked best.
It is a real one, and you should weigh it. What you get in exchange is that the person who scopes the work is the person who builds it, so nothing is lost in a handoff, and our standards are not a policy somebody else has to be persuaded to follow. If your project needs a team of forty, we are the wrong call, and we will say so in the first conversation rather than the third month.
Not yet. It is in private preview: the system is deployed and running, but public signup is closed while a broker integration goes through review. It is software you operate yourself — it does not give advice, does not manage money, and publishes no performance claims.
Soon, but we are not promising a date. The app is built, signed and verified on a physical device, the store listing is written, and submission is next. We will add the download link the day it goes live and not a day before — a store badge that leads nowhere is exactly the kind of small lie we work to avoid.
Yes, extensively, and we would rather say so plainly. It changes what one engineer can carry; it does not change what has to be true at the end. Every output goes through the same gates as anything else — tests, review, and a check that can actually fail. An AI-written rule with nothing enforcing it is worth exactly as much as a human-written one.
Client work is the client’s, and we sign mutual NDAs as a matter of course. The specifics belong in an engagement agreement rather than on a marketing page — raise it in the first conversation and we will put it in writing.
No cookies, no analytics, no third-party requests — even the fonts are served from this domain. The brief builder runs in your browser, and Download and Copy keep it there. Pressing Send is the one thing that transmits anything: your name, company, work email and what you wrote go into our contact records so that we can reply. That, plus our host’s ordinary server log, is the whole of it, and the privacy policy says how long we keep it and how to have it deleted.
Tell us about the work that could work better.
That is a good
place to start.
Download our brief — what we build, and how an engagement runs.
A few details to give your next conversation a head start. Nothing leaves your browser until you press Send.
Or write to us directly: hello@sanadtech.us