Effective date: 22 October 2026 · Last updated: 22 October 2026
Who we are
Hold My Seat (“the game”, “we”, “us”) is a mobile puzzle game published by
Arrows App Studio (“the developer”). Contact:
lazymanstudios@gmail.com.
The short version
Hold My Seat does not collect, transmit, sell, or share personal information.
There is no account, no advertising, no analytics, and no third-party SDK that
collects data. The game plays offline.
Information stored on your device
The game saves the following locally in your device’s app storage:
level progress, stars, and cleared routes
sound and calm-motion preferences
whether the one-time tutorial has been seen
This information never leaves your device through the game. It is not sent to
us or to any third party.
Android backup
If you have Android Backup enabled on your device, Android may include the
game’s local app data in your device backup, which is stored in your Google
account. This is an operating-system feature controlled by you and Google, not
by the game. We do not receive a copy of that backup. You can disable backup in
Android Settings.
Permissions
The Android package requests the INTERNET permission because the game runs
inside a web view shell. The shipped game makes no network requests. It does
not use location, camera, microphone, contacts, photos, files, or any other
sensitive permission.
Children’s privacy
Hold My Seat is family-friendly and does not collect personal information from
anyone, including children under 13. Because no personal information is
collected, used, or shared, the game does not require parental consent for data
collection. If you believe a child has provided personal information to us,
contact us and we will confirm that we hold none.
Advertising and purchases
No ads. No in-app purchases. No subscriptions. No advertising identifiers.
Third parties
The game does not include analytics, advertising, crash-reporting, or social
SDKs. The Android WebView is provided by your device’s operating system.
Data retention and deletion
We do not retain any personal information on a server. To delete the game’s
local data, use Android Settings → Apps → Hold My Seat → Storage →
Clear data, or uninstall the game. Uninstalling removes the local
progress and settings.
Changes
If this policy changes, we will update the effective date and publish the new
version at this URL. Material changes will be reflected in the store listing’s
release notes.
Contact
Questions about this policy: lazymanstudios@gmail.com
Developer: Arrows App Studio
Hold My Seat is an original cozy puzzle game. It is not affiliated
with, endorsed by, or sponsored by any other game or publisher.
This privacy policy applies to the Android application Document Scanner: PDF & OCR ("the app", "ClearScan"), published by Lazy Man Studios ("we", "us").
Summary
The app is a fully offline document scanner. Documents you scan, the text the app recognises (OCR), the barcodes it reads and the files it exports are processed and stored only on your device. We do not operate a server, we do not have user accounts, and the app has no way to upload your documents to us.
Information we collect
We do not collect, transmit or sell any personal information.
The app contains:
no user accounts or sign-in
no advertising or advertising identifiers
no third-party analytics, crash reporting or attribution SDKs
no access to your location, contacts list, calendar, microphone or phone state
Information the app processes on your device
The app processes the following data locally, purely so that its features work. This data is never sent to us or to any third party by the app.
Data Why Where it lives Leaves the device?
Photos you capture or import To create scanned pages App-private storage (filesDir/originals) Only if you share or export
Processed/cleaned-up page images Filters, crop, enhancement App-private storage (filesDir/enhanced) Only if you share or export
Exported PDF / JPG / PNG files To let you keep or share your scans App-private storage (filesDir/exports); optionally a copy in Documents/Scanner or DCIM/Scanner Only if you share or export
Recognised text (OCR) Text view, tables, contacts, read-aloud Memory only; only saved if you export it Only if you share or export
Recognised barcode / QR values Barcode view and actions Memory only Only if you share or export
App preferences (theme, default mode, PDF quality, …) To remember your settings App-private DataStore No
Device permissions
The app requests only the Camera permission, and only so you can scan documents. The first time you open the scanner you are asked to grant it; if you decline you can still grant it later from the Settings screen or from Android's app settings.
The shipped app does not request the INTERNET or ACCESS_NETWORK_STATE permission. They are declared by one of the bundled Google ML Kit libraries, and the app explicitly removes them at build time, so the installed app has no ability to open a network connection at all.
Importing an existing picture uses Android's Photo Picker, which gives the app access to only the specific images you choose. The app requests no storage, media-library, contacts, location or phone permissions.
Sharing and export
When you tap Share, Android's own share sheet is shown and you choose the destination app. When you enable Save to gallery, the app writes a copy of the exported file into your device's public Documents/Scanner or DCIM/Scanner folder through the system MediaStore API. Both actions are always initiated by you.
The app never opens a scanned link or web address on its own; scanning a QR code only displays its contents, and a link is opened only after you explicitly tap Open.
Third-party components
The app uses the following Google libraries to run entirely on your device:
ML Kit Text Recognition and ML Kit Barcode Scanning (on-device, bundled models)
AndroidX / Jetpack Compose, CameraX, Room, DataStore, Coil
These libraries are covered by Google's own privacy terms. The app itself does not configure them to send your documents or their contents anywhere. For Google's handling of library-level diagnostics, see Google's privacy policy.
Children's privacy
The app is a general-purpose utility and is not directed at children. We do not knowingly collect any information from anyone, including children under 13.
Data retention and deletion
Because everything is stored on your device, you are in full control:
Delete individual scans or pages inside the app's Library.
Delete exported files from the Library's Files tab or from your file manager.
Clear temporary files with Settings → Storage → Clear cache.
Remove everything at once by uninstalling the app or clearing its storage in Android's app settings.
We hold no copy of your data, so there is nothing for us to delete on our side.
Backups
Android's standard backup applies to your app preferences only. Scanned pages, enhanced images and exports are explicitly excluded from backup (see backup_rules.xml / data_extraction_rules.xml), so your documents never leave the device through Google's backup service.
Security
Your scans live in the app's private storage, which other apps cannot read. Files are shared externally only through Android FileProvider URIs with temporary, per-request read permission.
Changes to this policy
If the app's data practices change, this policy will be updated and the "Last updated" date above revised.
Contact
Questions about this policy: support@lazymanstudios.com
Knockback Arena ("the Game") is published by StudioX. This policy explains what data is handled when you play. The Game itself collects no account, contact, or location data and requires no sign-in.
1. Data handled by third-party services
Service
What is handled
Purpose
Google AdMob (ads)
Advertising ID, device identifiers, coarse location (derived from IP), ad interaction events
Showing banner, interstitial and rewarded ads; ad measurement and fraud prevention
Google Play Billing (Remove Ads purchase)
Purchase tokens and order history, processed by Google Play
Unlocking and restoring the ad-free upgrade
Firebase Analytics / Crashlytics (only if enabled in a given build)
Anonymous usage events, crash traces, device model / OS version
The Game is rated for everyone 10 and up and may be played by children with parental guidance. We do not knowingly collect personal data from children. Advertising identifiers, if any, are handled solely by Google AdMob; on Android you can reset or delete your advertising ID in Settings → Privacy → Ads, and enable Opt out of Ads Personalization.
3. Data retention
Gameplay preferences (fighter color, settings, match count) stay on your device only and are erased if you clear app data or uninstall. Match relay data is kept in memory only for the duration of a match. Purchase records are kept by Google Play per its own retention.
4. Your rights
Depending on your region you may request access, correction, or deletion of data held about you, or object to processing. For data we hold (support emails only), write to [SUPPORT_EMAIL]. For ad and billing data, use your Google account privacy controls.
5. Security
Network traffic uses HTTPS. No passwords or financial details ever touch our servers — there are none; purchases are completed inside Google Play.
6. Changes
Material changes to this policy will be reflected here with a new effective date before they take effect.
7. Contact
Questions about this policy: [SUPPORT_EMAIL]. Replace the bracketed placeholders before publishing this page at your public privacy URL.
This Privacy Policy explains what information is processed when you play
Valewall TD. Valewall TD is a single-player game. We do not require an account,
and we do not ask you for your name, email address, or any other personal
information to play.
## 1. Information stored on your device
Game progress (levels, stars, gems, tech tree, settings, and purchase flags) is
stored locally on your device. It is not transmitted to us. Uninstalling the
game deletes this local data. Progress and purchase flags are intentionally
excluded from Android cloud backup so purchases always restore through Google
Play rather than through a device backup.
## 2. Information processed by third-party services
The game contains optional ads served by **Google AdMob**, optional
**Google Firebase Analytics / Crashlytics** diagnostics, and purchases handled
by **Google Play Billing**. These services are provided by Google LLC and may
process:
| Service | Data | Purpose |
|---|---|---|
| Google AdMob | Device and advertising identifiers, IP address, approximate location derived from IP, ad interaction data, app version and device information | Serving and measuring ads, fraud prevention |
| Firebase Analytics | App events (e.g. app opens), device model and operating system, app version, app-instance identifier | Understanding aggregate usage to improve the game |
| Firebase Crashlytics | Crash stack traces, device state at the time of a crash, app version, installation identifier | Diagnosing and fixing crashes |
| Google Play Billing | Purchase token, product identifier, purchase status | Processing and restoring in-app purchases. Card and payment details are handled solely by Google; we never see them. |
We do not sell your personal information. Data processed by these Google
services is handled under Google's Privacy Policy
(https://policies.google.com/privacy) and Google's terms for the AdMob and
Firebase services.
The game's menus load the **Cinzel** and **Nunito** web fonts from Google Fonts,
which receives your IP address when it serves them; the game remains playable if
that request fails.
## 3. Your choices and consent
- **Ads (EEA, UK, and Switzerland):** a Google-certified consent message is
shown before ads load. You can change your choice at any time from the app's
Settings screen where available, or by clearing the app's data.
- **Analytics and crash reporting:** can be turned off at any time in
**Settings → Privacy → Share analytics & crash reports**.
- **Purchases:** "Remove Ads Pro" disables banner and interstitial ads.
## 4. Children's privacy
Ironhold TD is not directed to children under 13 (or the equivalent minimum age
in your region). We do not knowingly collect personal information from children.
If you believe a child has provided personal information, please contact us and
we will delete it.
## 5. Data retention and deletion
Local game data remains on your device until you uninstall the game or reset
your progress in Settings. To request deletion of analytics or crash data
associated with your installation, contact us at YOUR_EMAIL_HERE. You can also
reset your advertising identifier from your device's Google settings.
## 6. Security
Data transmitted by the Google services listed above is sent over encrypted
connections (HTTPS/TLS). No method of transmission or storage is completely
secure, but we only integrate established Google SDKs and keep them updated.
## 7. Changes to this policy
We may update this policy as the game evolves. Material changes will be
reflected by a new effective date in this document and, where appropriate, in
the app's store listing.
Vaibhav Srivastav posted a clip last week that made the rounds in every developer chat I’m in. He opened GPT 5.6 Sol, typed a single /goal prompt, and watched it build a working rollercoaster simulator — complete with textures, physics, and a track-laying system. No starter code. No asset imports. Just a broad idea and a few artistic directions along the way.
I’m a developer who’s been using AI coding tools since the GPT-3 era. I’ve seen the pattern before: a viral demo that looks impressive in a screen recording but falls apart the second you try to do anything real with it. So I spent a weekend recreating Vaibhav’s experiment to see whether Sol actually delivers or if this is just another polished clip.
What Actually Happened
The original demo showed Sol turning a one-line /goal into a fully rendered 3D rollercoaster simulation. The creator noted it still had some UI inconsistencies and wasn’t complete, but the core loop worked — you could see the track, watch the cart move, and interact with the simulation. He summed it up with a line that stuck with me: “You’re truly bounded by your own ambition.”
I opened Sol on a fresh session and wrote my own /goal describing a rollercoaster simulator. No example code, no architecture notes, no hand-holding. Just the broad concept and a request to build it.
The response came back in under two minutes.
How Sol Handled the Build
Sol didn’t just write functions. It made decisions. It chose a rendering approach, generated its own textures, set up a physics loop for the cart, and laid out a UI without being told to. The output was a single self-contained file that ran immediately.
The /goal prefix behaves differently from a standard chat prompt. Instead of generating code snippets you have to wire together, Sol enters what looks like a planning-and-execution mode. It scaffolds the project, makes architectural choices, and delivers a complete experience rather than building blocks. The textures Sol generated on its own surprised me most. I didn’t upload any images or describe what the track should look like. It picked colors, added surface detail, and made it look like a deliberate design choice.
Where It Shines
The scaffolding phase is where Sol separates itself from every AI coding tool I’ve tried before. Earlier models require you to break a project into pieces, prompt for each one, and assemble them yourself. Sol skips that entirely. You describe the experience you want and it builds toward it.
For a solo developer or a small team validating an idea, this changes the calculus of “should I build this?” The cost of exploring a concept drops from “let me spend a week on a prototype” to “let me spend an afternoon.” That’s a real difference, not a marginal one.
The asset generation is the feature that surprised me most. Sol doesn’t just write code — it generates textures, colors, and visual elements as part of the same build. In previous models, generating game assets meant a separate pipeline: prompt an image model, download files, wire them into the project. Sol collapsed that into a single step.
Where It Still Trips Up
The generated code works, but it’s not clean. Variable names are inconsistent. There’s dead code in places — functions that are defined but never called, imports that aren’t used. The UI has the kind of off-by-a-few-pixels spacing that tells you no human reviewed the layout before shipping.
Iteration is harder than it should be. If you ask Sol to fix one specific issue, it sometimes regresses something unrelated. This isn’t unique to Sol — it’s a known challenge with AI code generation — but it’s more noticeable here because Sol generates larger, more integrated blocks of code. When something breaks, tracing the problem back to its source takes longer than it would in hand-written code.
The original creator mentioned UI inconsistencies, and I hit the same wall. The visuals Sol generates are impressive for a first pass, but refining them into something polished requires manual intervention. The floor is lower and the ceiling is higher than earlier models, but the cleanup work hasn’t gone away.
FAQ
Can GPT 5.6 Sol really build a complete game?
Yes, for a realistic definition of “complete.” The generated output is a working, playable experience with graphics, physics, and interactivity. It is not production-ready — the code needs cleanup and the UI has rough edges — but it is genuinely functional, not a tech demo that stalls after one interaction.
What kind of prompts work best for game generation?
The /goal prefix made a noticeable difference in my tests. Broad, high-level descriptions of the experience you want work better than detailed step-by-step instructions. Sol performs best when you describe what the user should see and do, then let it figure out the implementation.
Do you need coding experience to use Sol for games?
You can generate a working game without writing code, but you will need development knowledge to fix the issues Sol introduces. The current generation of AI coding tools accelerates developers who already understand software. It does not replace that understanding.
How does Sol compare to GPT-4 and Claude for coding?
Sol produces more complete, self-contained output out of the box than either GPT-4 or Claude’s coding abilities. The asset generation is unique — no other model generates textures and visuals alongside code in the same pass. The tradeoff is that Sol’s output is harder to debug when things go wrong, since the code follows Sol’s own conventions rather than standard human patterns.
Are Sol-generated games production-ready?
Not without significant manual cleanup. The generated code runs, but it is not structured for maintainability, performance testing, or edge case handling. Sol is best framed as an extremely fast prototyping tool. You can validate an idea in an afternoon that would have taken a week manually.
Try It Yourself
Open GPT 5.6 Sol, write a /goal describing something you want to build, and see what comes back. Start small — a simple game, a utility tool, a visual experiment. The barrier between “I have an idea” and “I can see it working” is thinner than it has ever been.
If this is the baseline for Sol today, the next six months are going to be interesting. What are you going to build?
Last week, an open-source model named Fable 5 won a 4-model 3D build benchmark against GPT-5.6, Grok 4.5, and GLM 5.2 using the exact same prompt to generate floating island cities in the browser. The results were shared by researcher 0xMarioNawfal and they’ve been making rounds in every AI group I’m in. I’ve been using Fable 5 for a few weeks, so I decided to run the same test myself and see whether the benchmark held up or if it was a fluke selection.
The Benchmark That Put Fable 5 on Top
Four models received the same instruction: generate a floating island city in the browser, visible from a 3D-perspective view. No additional hints, no fine-tuning — just the raw model output. Fable 5’s result was selected as the winner.
This wasn’t a narrow test either. The benchmark evaluated structural coherence (do the islands look like they could actually float?), visual fidelity (do the textures and lighting hold up?), and prompt adherence (did the model actually build a city with distinct structures, or just a rock with some noise on top?). Fable 5 scored highest across all three criteria.
The comparison models weren’t slouches. GPT-5.6 is OpenAI’s latest frontier model with native multimodal generation. Grok 4.5 is xAI’s most advanced offering. GLM 5.2 represents the best from the Chinese AI ecosystem. Fable 5 won anyway.
Why Floating Island Cities?
Floating islands are a deceptively hard test for AI generation models. Unlike standard 2D image generation, a 3D floating island requires the model to understand spatial relationships from multiple angles, maintain visual consistency across visible surfaces, and generate geometry that looks physically plausible even though it’s fantastical.
Most importantly, it tests whether the model can handle a compound prompt — “floating island city” is not “floating island” plus “city” as separate concepts. The model has to integrate them: buildings on the islands, bridges or connections between them, varying altitudes, and a coherent sense of scale. Many models nail the terrain and drop the structures. Others build detailed cities on flat ground but can’t adapt to floating platforms. Fable 5 handled both simultaneously.
This kind of test matters because it mirrors real 3D asset generation workflows. Game developers, architects, and VR content creators don’t want a model that generates a pretty picture — they want one that generates usable geometry with consistent structure from any angle. That’s exactly what this benchmark measured.
My Hands-On Test with Fable 5
I’ve been experimenting with Fable 5 for personal projects over the past few weeks, mainly for prototyping 3D environments. When I saw the benchmark results, I pulled the exact same floating-island prompt and ran it locally.
The output loaded in about 12 seconds on my setup (RTX 5090, 64 GB RAM). What rendered matched the benchmark winner closely: rocky floating platforms at varying heights, connected by bridges, with small clustered buildings that read clearly as a city rather than random geometry. The lighting was consistent across all visible surfaces — no dark spots or misaligned normals that I’ve seen with other models on complex scenes.
I then tried the same prompt with GPT-5.6 through its API. The output was visually rich but had structural oddities — some buildings clipped through the island geometry, and one platform looked like it was floating upside down from certain angles. The texture work was impressive, but the spatial coherence wasn’t there.
The difference became obvious when I rotated the camera around each output. Fable 5’s scene held together from every angle. GPT-5.6’s scene had angles where the illusion broke. For a single hero shot, GPT-5.6 might win on polish. For a 3D scene you can actually orbit and inspect, Fable 5 was clearly ahead.
What This Win Means for Open-Source AI
This result matters beyond the 3D generation niche. It’s the latest data point in a pattern that’s been building all year: open-source models closing the gap with frontier labs in specific domains, often on a fraction of the training budget.
Fable 5’s win doesn’t mean it’s a better general-purpose model than GPT-5.6. GPT-5.6 still dominates on reasoning benchmarks, coding benchmarks, and language understanding. But it does mean that the gap in domain-specific capabilities is shrinking faster than many expected. When you need 3D generation, Fable 5 is now a legitimate contender — and it’s free.
For developers and creators, the takeaway is practical: the best model for your use case isn’t automatically the most famous one. Running a quick benchmark against 2–3 alternatives before committing to a model stack can save months of work with a suboptimal tool.
FAQ
How does Fable 5 compare to GPT-5.6 outside of 3D?
On general reasoning, coding, and language benchmarks, GPT-5.6 still leads by a measurable margin. Fable 5 excels specifically at 3D generation and spatial reasoning tasks. Think of it as a specialist model that competes with generalists in its domain — similar to how specialized image models like Midjourney outperform GPT’s image generation despite having far fewer overall capabilities.
Is Fable 5 free to use?
Yes, Fable 5 is open-source and available for download. You can run it locally if you have compatible hardware, or access it through various inference providers. The model weights are publicly available — a quick search for “Fable 5” on your preferred model repository will find it.
What hardware do you need to run Fable 5?
In my testing, Fable 5 runs comfortably on a system with 24 GB+ of VRAM. An RTX 4090 or better is recommended for smooth 3D generation. Quantized versions are available for lower-end hardware, though generation quality and speed will scale with available resources.
Will OpenAI and xAI catch up in 3D generation?
Almost certainly. The frontier labs have the resources to improve rapidly. But the fact that an open-source model took the lead, even temporarily, is significant. It means the barrier for entry in AI research is still low enough that smaller teams can make meaningful contributions. The gap between frontier labs and the open-source community has narrowed, and that trend is unlikely to reverse.
Try the Benchmark Yourself
The floating-island prompt is public and easy to reproduce. Pull Fable 5 this weekend, run the same prompt, and compare the output with GPT-5.6 or Grok 4.5. The whole test takes about 30 minutes end to end. Drop your results in the comments — I want to see how your outputs compare with my experience. The more data points we collect, the clearer the picture gets.
