The information was already there. Now it’s usable.
Short & clear:
Files instead of pipelines · Templates & rules · Free for private use
Why scanOS exists
Information is everywhere! In apps, dashboards, PDFs, screenshots, photos, highlights, exports… and obviously in the world around us. Today, it gets read, interpreted, briefly summarized — and then disappears again. Information stays fleeting. And we’ve gotten used to that.
But what if you could capture that information cleanly and keep working with it?
The idea behind scanOS:
What you can see, you can photograph. With scanOS, you keep it. Information is given a stable, explicit form — and stays usable over time.
What is scanOS?
scanOS lets AI see what you see. It is an ingestion and normalization tool for anything that is visible.
It reads content from:
- photos
- screenshots
- PDFs and other files
and turns it into a clean, structured form that both humans and machines can work with further.
This means scanOS is not just a one-off OCR tool and not just a chat feature (although yes — it can do that too 🙂). It shines in situations where the same kinds of visual information are processed in the same way over and over again.
What you use scanOS for
Everyday life & work
- digitizing handwritten notes and sketches
- capturing standardized documents (e.g. invoices for taxes 😉) (*)
- reusing content from training, sleep, or tracking apps
Research & knowledge work
- photos and scans of papers and other reading
- capturing and digitizing highlights and annotations
- creating structured literature and citation data
(e.g. for reference management software)
Documentation & archiving
- machine-readable capture of states, situations, and objects
- normalized output of processed information
- building traceable, searchable archives
(*) LLMs can make mistakes. For sensitive areas like tax declarations, always verify what you digitize!
What you can actually do with scanOS
Ingest & extract
scanOS reads visually encoded information from images, documents, and exports. It doesn’t just consider text, but also layout, emphasis, and structure.
If you can look at it, scanOS can process it.
Normalize & structure
Captured information is transformed into a fixed, consistent structure:
- unambiguous field names
- stable output format
- comparability over time
Structure is not an extra.
It’s the prerequisite for analysis and reuse.
Reuse & integrate
Structured data can be reused directly:
- passed into logs (e.g. trainingOS, sleepOS, nutritionOS)
- used to build stable data flows
- used independently of the MetaMemoryWorks stack
Read once — usable forever.
Example: structured training data extracted from a simple app screenshot
Templates & rules
scanOS works with templates. You define a template for recurring use cases and think through a few rules for processing. Templates can include, for example:
- different structures
- diverging priorities
- custom abort conditions and safety notes (*)
Example:
You process the same kind of information every day, for instance from a project management tool. You pass an image of it to scanOS — and get the information back exactly in the form you defined. Templates and rules make scanOS extensible, adaptable, and transparent.
Templates are optional, by the way. You can also just start right away and tell scanOS directly what you want.
(*) Note:
Rules are semantic guardrails inside an AI. They help steer processing, but they are not hard technical barriers or guarantees.
They can be “argued away” — keep that in mind when using them.
What changes for you
At first, you just upload an image — and get text or structured data back. Or you build a few templates. Practical. Task done.
But over time, something more fundamental changes:
- You stop capturing information in an unstructured way. It gets a fixed form.
- Normalized information can easily be archived, for example in another MMW module.
- Recurrent information suddenly looks the same every time — making it comparable and ready for data analysis 🙂
- Individual screenshots, photos, or scans gradually form context.
- Pipeline integration: scanOS is ideal as an entry point into a data pipeline. Build it into your own workflows or use it, for example, to pass data into trainingOS.
scanOS no longer feels like a one-off helper, but like a stable perception layer you can rely on.
What is recognized stays.
How scanOS works
- You upload an image, document, or export.
- scanOS detects the type or uses an existing template.
- The information is output in structured form.
- The result exists as a file.
- You can reuse it, store it, or integrate it.
Setup
Setup takes about 2–3 minutes:
- download the ZIP
- unzip
- create a new project in your LLM
- upload the files
- start using it
No installation.
Tested with: ChatGPT
What exactly is included
The scanOS ZIP contains files only. No installation, no dependencies, no hidden code.
Specifically:
-
scanOS Engine
The engine definition. It specifies how visually encoded information is read, interpreted, and transformed into a structured form. -
scanOS Templates
A collection of templates for recurring inputs.
Templates describe which information is extracted, what it is called, and how it is structured in the output. -
scanOS Image Description
A specialized definition for non-textual images.
It generates both a human-readable description and a structured, machine-readable representation. -
scanOS Start Guide
A short introduction explaining how scanOS is intended to work, how templates function, and how to get started. -
legal/
License, terms, and legal notices.
All components are explicitly readable, adaptable, and extensible. scanOS is not a black box system, but a collection of clearly defined descriptions.
Context: scanOS compared to existing solutions
Many tools process visual information today. You photograph a document, scan a page, or upload a screenshot — and get text, highlights, or recognized content back. Practical! You complete a task, get a result, and move on.
What’s missing, though, is continuity.
The recognized information often exists only for that one moment. As a text block, a table, a chat response, or a file you have to manually process further. When you capture the same kind of information again tomorrow, you start over — slightly different, slightly shifted, barely comparable.
scanOS takes a different approach.
It uses the same basic capabilities — text recognition, visual analysis, layout detection — but shifts the focus away from recognition toward retention and standardization. Information isn’t just read, but transformed into an explicit, stable form. A form that can be repeated, stays comparable, and can be reused.
In practice, that means:
- You can use scanOS very simply, just for OCR or text recognition.
- If you regularly process the same kind of information, you can define how it should look.
- Results don’t disappear in chat windows, but exist as files.
- Perception becomes something you can store, verify, and continue.
scanOS does not replace existing tools. It complements them where one-off recognition should become durably usable information.
Or put differently: many systems can see. scanOS makes sure that what is seen stays.
And the best part: for private, non-commercial use, scanOS is free 🙂
Download
License (short version)
- Private use: free
- Commercial / institutional use: license required
Details: → View license terms
Support
scanOS is a public good.
If scanOS helps you and you’d like to support its continued development:
Information doesn’t become valuable because it is read — but because it stays.