A one-thumb arcade game about never standing still — 600 levels, all offline.
iOS · AndroidPlay
Independent Product Studio
We build apps, games and digital products.
An independent product studio. Fourteen products, and one shared production system underneath all of them.
Fourteen products. Every one of them has a demo you can run here, without an account.
ZUPI
Didi
SPRIVO
Products
What we have built.
Fourteen products across games, learning, wellness, AI and one stubbornly useful household app. Each one is ours — designed, built, published and supported by the studio. Four are below; the rest are one click away.
In your browser
Every product here has something you can run right now.
No install, no account, nothing to download. Two of them are the product’s own code running in your browser; the rest say plainly what they are.
The flagship
Do not take our word for it. Play the product’s own code.
The rule check is Glow Garden’s own code, running unmodified in your browser — including the solver that refuses any placement that would leave the board unsolvable. The board is fixed rather than generated and has exactly one solution, as every board the game makes does. Newer versions of the game add mechanics this board does not use.
Real product codeNo installNo sign-in
ZUPIInteractive recreationKeep the Pulse Core alive for thirty seconds without ever holding still — deflect to charge the Pulse, and save it for the plated threats your shield can’t stop.
VeyraReal product codeTurn one of Veyra’s own drawings into a clean tattoo stencil.
Luma JarsInteractive recreationClear two gardens in a row: sort the glowing spirits until every jar holds one colour.
DidiInteractive recreationAnswer “has the dog been walked?” — then add something of your own.
PivumoReal product assetsLearn four colours by ear, then prove it — in a language you may not speak.
Forgotten TownReal product assetsRepair the front gate — the game’s real first order.
Candy DropReal product assetsRace eleven of Candy Drop’s own racers down Candy Canyon and take the fifty-value Candy Bit on the way.
Swarm KingdomsInteractive recreationTake the grove back: steer the swarm through the first expedition and break the Root Warden.
VelaInteractive recreationPick a language you do not speak — and order a coffee in it.
AurenInteractive recreationLog a day, add a meal, and see how carefully the app refuses to over-read it.
SPRIVOGuided demoTake one sprint from a phone clip to a single coaching cue.
PlicaInteractive recreationWalk a sample page through Plica: filter it, read its text, and see what Plica’s own rules find in it.
WeirInteractive recreationTurn off a surface, then try to open it.
How we build
Fourteen products. One shared production system.
Fourteen products do not run alongside each other unless the expensive parts are shared. Research, production, release and measurement run through one internal system, so the fifth product costs less to launch than the first and is better informed when it does.
- 01ResearchSignals and a per-product idea bank, scored for relevance before anything is made.
- 02DesignThe product decides its own voice, claims and boundaries. The system fills them in.
- 03BuildAssets, narration and captions assembled from each product’s own material.
- 04LocaliseEvery string in every language, with parity between catalogues checked rather than assumed.
- 05TestAutomated gates, then a person. Nothing reaches a queue unreviewed.
- 06PublishPer-platform packaging, with publishing deliberately held behind human approval.
- 07MeasureRelease and performance signals return to where the work was made.
- 08LearnWhat worked on one product is readable by the next one that asks.
- 09ImproveThe fix lands in the shared system, not in one product’s copy of it.
Stage Nine leads back to one. What a launch teaches is available to the next product that asks, which is the whole reason for running one system instead of fourteen.
The studio engine
The hard part of automation is making it safe to re-run.
Producing something is the easy half. The half that decides whether a studio can actually run fourteen products at once is what happens on the retry, the crash and the duplicate trigger.
Nothing publishes twice
Retrying is always safe, which is what makes automation restartable.
A crash resumes, it does not restart
A long pipeline can fail without costing the whole run.
The system never learns a product’s name
A new product is a configuration, not a fork.
Human approval stays in the loop
Scale without handing over editorial control.
This page describes what the system does and what it produces. It does not describe how it is deployed. Hosts, endpoints, credentials, machine names, storage paths and security architecture stay private — not because they are interesting, but because publishing them would be careless. The public site and the internal systems share no credentials and no control path.
Real interfaces, not mockups
Fourteen products, each with its own world, out of the same studio.
- ZUPILIVE
- DidiLIVE
- SPRIVOLIVE
- Glow GardenBETA
- Luma JarsLIVE
- VeyraLIVE
- Swarm KingdomsBETA
- Candy DropBETA
- PivumoLIVE
- VelaBETA
- AurenBETA
- Forgotten TownLIVE
- PlicaBETA
- WeirBETA
Lab
Work that is not a product yet.
Open questions we are testing in the open, including the ones where the answer turned out to be no.
- EXPERIMENT
Borrowing a solver instead of rewriting the rules
When a demo needs a product’s logic, is copying the real code in safer than rewriting a simplified version of it?
- LAB
What a demo may claim when the product is not finished
Can a product that does not work yet still have a demo that is worth running and not misleading?
- EXPERIMENT
Shipping the real engine, not a recreation
Does letting a visitor play the actual game — the same engine that runs in the app — prove more than any screenshot gallery can?
- LAB
Producing twenty-nine voiced locales offline
Can a full multi-language voice set for a children’s app be produced locally on one GPU, with no paid speech API and no per-character billing?
- LAB
Sprint form analysis without a server
Can a phone analyse sprint mechanics well enough to coach from, without uploading the video anywhere?
The studio
A studio is judged by what it ships, not by what it says it does.
NOVEMADE is not an agency. We do not take briefs, bill hours or keep a client logo wall. Every product on this site is one we decided to make, funded ourselves and are responsible for after launch — including the parts that are still unfinished.
That changes what the work optimises for. There is no handover, so shortcuts stay our problem. There is no client sign-off, so a product ships when it is genuinely good rather than when it is approved. And because every product is carried by the same studio, anything done twice gets built once and shared.
- 01Ship the smallest honest thing
- 02Build the system, not the one-off
- 03Own the whole lifecycle
- 04Let each product be itself
Evidence
Eight numbers, and where to check each one.
This is what the studio actually has today. All but two are counted from the site’s own content when it builds, so they cannot quietly go stale.
- 14products in the studio
Apps, games, learning, wellness and AI, at every stage. Seven are live on the App Store, seven are in closed testing and zero have never been submitted to a store — and each page says which.
Check it- 14of them you can run in this browser
One per product, with none missing. No install, no account, nothing to agree to — and each one states what it is running before you start.
Check it- 5demos run the product’s own material
Two run code vendored out of the app unmodified, with the source hash recorded in the repository. Three play the app’s own shipped files — Pivumo’s voice recordings and Forgotten Town’s artwork and content. The other nine say plainly that they are not the product.
Check it- 29languages voiced for Pivumo
2,088 recordings, two speakers per locale, produced offline and licence-cleared file by file. Six of those languages are playable on this site, and Pivumo went live on the App Store on September 22, 2026.
Check it- 600levels in ZUPI
Generated from a difficulty model across six worlds, with a checkpoint every tenth level, so two players on level 412 are playing the same level.
Check it- 100orders a test plays before every release
Forgotten Town’s campaign test finishes the whole game using only ordinary merges and coin refills — no granted items, no premium currency — and reloads the save after every order. It is how "completable without paying" stays a build failure rather than an opinion.
Check it- 9stages in the production loop
Research to improve, shared by every product, with publishing deliberately held behind a person rather than automated end to end.
Check it- 0invented metrics on this site
No download counts, ratings, user numbers or revenue figures appear anywhere — and the build fails if one is added, rather than relying on anybody remembering.
There are still no download, revenue or rating figures here. Seven products are on the App Store, the newest of them since October 3, 2026, and the rest are not public at all — so any number worth printing would cover a few days of one platform. When there is one worth standing behind, it will appear in this table with the same sourcing as everything above it.
How the work is checkedJournal
Notes from the work.
- Technical note
A browser demo for every product, and a four-rung ladder to keep it honest
Eleven products, eleven demos, and the uncomfortable part: two of them run the product’s own source, and saying which two is the whole point.
7 min read
- Studio note
Eleven products, one production system, and the order we read evidence in
The expensive parts of a launch are identical every time. Sharing them is the obvious half; knowing which document to believe is the half that keeps going wrong.
6 min read
A thousand levels, and nobody placed one of them by hand
Ten regions, a hundred difficulty tiers and fifty boss families, generated from a model — and the acceptance run that had to prove every route through it is winnable.
6 min read
Services
We build for other people the way we build for ourselves.
NOVEMADE is a product studio first: every product on this site is one we decided to make, funded, and still own. Fourteen of them, seven live on a store, each with a demo you can run in this browser. These six pages are what that turns into when the product is yours instead of ours.
Collaborate
We are not for hire. We are open to the right partnership.
Publishing, partnerships, investment, or a technical conversation about how we build. Tell us which one and we will skip the rest of the form.