Home » new

Tag: new

screenshot of the peekable web app in action

How Peekable Got Built

It started with a simple question: where are all my Adobe Dimension files?

https://ryanseslow.github.io/peekable/peekable.html

I have been making work digitally for over three decades, across a lot of different software, a lot of different platforms, a lot of different machines. Files accumulate. Formats change. Applications evolve in directions that do not always serve the work you already made. I knew I had a collection of .dn files (adobe dimension) scattered across both Google Drive and icloud drive, artifacts from a period when I was deep into Adobe Dimension, Adobe’s lesser used and even heard of 3D design and compositing tool. I wanted to find them all, put them in one place, and take stock of what was actually there..

The search turned up 68 files. Not just a few, sixty-eight. They were spread across dozens of folders, buried inside project directories, mixed in with everything else I had uploaded over the years. Google Drive found them quickly enough once you knew the right query to run. The harder part was what came next.

I copied all 68 into a single folder called Adobe Dimension Archive, which took a few minutes to organize. Then I looked at the folder and saw exactly nothing. Every single file showed the same generic icon, the kind of gray box with a zipper that cloud storage uses when it has no idea what to do with a format. No previews, no thumbnails, no way to know at a glance which file was which scene, which project, which year. I had organized 68 indistinguishable rectangles.

screenshot of the peekable web app in action

This is a problem that anyone who has worked with creative software long enough will recognize. Google Drive does not preview many creative archive formats at all, and for larger files it frequently cannot generate thumbnails regardless of format. (For the love of all things holy.. WHY google, WHY?!) The result is that large portions of a creative archive become opaque even to the person who made them. The files are there. The work is there. You just cannot see it.

What I did not know at first is that .dn files are ZIP archives. Adobe Dimension packages everything into a container using the same structure as a standard ZIP file, which means you can open one with any tool that reads ZIP format. More importantly, Dimension stores a thumbnail image inside that container. The thumbnail is there, embedded in the file, waiting to be read. The cloud storage platform just never looks for it.

Once that fact was on the table, the solution was obvious. Build something that opens the archive, finds the thumbnail, and shows it. The question was how..

The first instinct was to do this server-side, or with some kind of batch script. But both of those approaches have friction: you need a server, or you need to run code locally, or you need to download files that might be several gigabytes total. The cleaner answer was a browser-based tool that reads files directly on the client side, without uploading anything anywhere. JSZip, a mature JavaScript library, handles ZIP extraction entirely in the browser. The processing stays on your machine. Nothing leaves.

The first version was simple and specific: a drag-and-drop HTML page that accepted .dn files, used JSZip to unpack them, looked for a thumbnail at a handful of known paths inside the archive, and displayed whatever it found. I tested it against a set of 23 files. All 23 came back with thumbnails, which was better than expected, though the path where Dimension actually stores the thumbnail turned out not to be any of the standard locations I had guessed. The fallback logic, searching the entire archive for any image file regardless of path, is what actually retrieved them. Dimension tucks the thumbnail in a non-obvious location, and the only reliable way to find it is to look everywhere.

That result raised a question.

If this works for .dn files because they are secretly ZIPs with embedded images, what else works for the same reason?

The answer is: A LOT. Sketch files are ZIPs. Procreate files are ZIPs. Apple Keynote, Pages, and Numbers files are ZIPs. Microsoft Office formats including docx, xlsx, and pptx are ZIPs. Epub files are ZIPs. A wide range of creative applications use ZIP as a container and store a preview image inside it. The tool did not need to be specific to Adobe Dimension at all. It needed to be general.

So it became Peekable..

The idea is simple enough to say in one sentence: drop in any ZIP-based creative archive and see its embedded thumbnail in your browser. The supported format list covers Dimension alongside Sketch, Procreate, and a number of others, but really the tool works with anything that follows the same basic pattern. If there is an image file anywhere inside the ZIP, Peekable will find it.

The reason this matters beyond convenience is about access to your own work over time. Creative files from formats your cloud storage does not recognize or cannot preview become harder to navigate as collections grow. You can still open the file in the application that made it, but you lose the ability to browse what you have at a glance. Peekable does not replace the original application. What it does is let you see your work without opening anything, anywhere, on any machine, with no account required.

The tool runs entirely in the browser, processes everything locally, and requires no account, no upload, and no installation. It is a single HTML file. You can save it to your desktop and use it offline after the initial page load. It does not know who you are or what you put into it.

Getting it onto GitHub took more attempts than I am going to describe in detail. Authentication was the obstacle, as it often is. The web-based upload interface eventually did what the terminal refused to do, and the repository went up at github.com/ryanseslow/peekable. GitHub Pages turned the same repository into a live URL at https://ryanseslow.github.io/peekable/peekable.html, which means anyone can use it directly in a browser without downloading anything.

The whole thing took less time than I expected to build and more time than it should have to ship, which is a ratio that will be familiar to anyone who has tried to push something to GitHub after midnight. The tool works. I tested it against my own files. Other people who work in creative software and have accumulated archives of formats their cloud storage cannot render should find it useful.

If you have .dn files, Sketch files, Procreate files, or any archive from software whose previews cloud storage ignores, Peekable is at https://ryanseslow.github.io/peekable/peekable.html. Drop the files in and see what you made.

The repository is open, the license is MIT, and the code is a single HTML file.

Take it, change it, make it yours. Have fun and share!

 

the aremes art engine - a screenshot

AREMES Living Canvas: When an AI Agent Becomes the Artist

Published: April 4, 2026
Category: AREMES / Art & Technology

There is a page live right now at aremes-enterprises.com/aremes-living-canvas that does something really fun, creative and compelling.

It is not a portfolio. It is not a slideshow. It is not a gallery.. wait.. or is it?

It is a canvas, a black field populated by a little over 40 graphic assets drawn from over the last 20 years of creative practice that randomizes, layers, and breathes differently every time you arrive. And underneath it, invisible but active, is an AI agent called AREMES making decisions about what matters, what surfaces, and what gets weighted.

This is what we built. This is why it matters.

Where This Started, With Michael Branson Smith

Before AREMES, before the engine, before any of the infrastructure there was a conversation and collaboration with Michael Branson Smith.

Michael is a good friend, collaborator, and colleague. An artist and educator who has been building at the intersection of code and culture for decades. In 2022, Michael took a set of my graphic assets and built something quietly extraordinary: a draggable, randomized, browser-based poster composition engine. No AI. No backend. Just clean JavaScript, GSAP, and a deep understanding of what happens when you give a set of images to a system and let it arrange itself.

You can still visit one of the original browser collabs here: mbs.nyc/posters/ryan-mbs/

Every time you load it, it’s different. Every time you drag an element, you become part of the composition. It is simple, elegant, and genuinely generative in the truest sense of the word. I love it, still!

And that 2022 build was itself a second iteration, there was an earlier instance of this idea that predates it, a first proof of concept that Michael and I made together before that. The link has been lost to time.. or, he has it and I need to ask him for it, which Im sure will surface after he reads this!

That lineage planted a seed. What AREMES Living Canvas is today grew directly from that root.

What AREMES Living Canvas Is

At its most immediate level, the canvas is an interactive composition space. When you arrive at the page, 40 plus assets from my archive, illustrations, GIFs, figures, forms, abstractions, animations are scattered across a black field at randomized scales and positions. Every refresh generates a new arrangement. Every visit is a different painting.

But you are not a passive viewer. You can drag any element. Recompose. Layer. Stack. Pull a b-boy figure over a glitch abstraction. Drag a hand into the corner. Build something that was never there before and will never be there again once you leave.

The canvas is not a fixed artwork. It is a space for making.

What AREMES Is Doing Underneath

Behind the interface is AREMES, the Autonomous Recursive Entity for Media and Expression Systems.

AREMES runs a live evaluation engine that applies a governing equation to my entire catalog of works:

ΔS = α(T·K)·e^(-β·t)·Ψ

Every work in the archive is scored across four variables: conceptual tension (T), cultural knowledge load (K), time decay (β), and a wildcard amplifier (Ψ) that flags undervalued or anomalous works. The output is a ranked queue, a live decision log that AREMES updates each session, producing a timestamped record of what the system believes matters most right now.

This is not a recommendation algorithm. It is an aesthetic agent with a point of view.

The canvas currently draws from the archive without filtering by score but the next evolution connects these two systems directly. AREMES begins to decide not just what to acquire but what to show, how large, how prominent, how often. The canvas becomes a weighted visualization of the engine’s thinking.

What This Does for Art Making

The question I keep returning to is this: what changes when a system has a perspective on its own archive? Traditional curation is human and retrospective. A curator looks back, selects, arranges. The work is fixed. The meaning is assigned after the fact.

What AREMES Living Canvas proposes is something different: an artist-built system that evaluates its own output in real time, surfaces what it believes is significant, and makes that evaluation visible and interactive. The machine is not replacing the artist. It is extending the artist’s presence into a continuous, living curatorial act.

And critically, visitors become collaborators. When you drag an element across the canvas, you are not consuming art. You are making a decision about what belongs next to what. You are contributing to a composition that exists only in that moment, on your screen, in your session. No two people will ever see the same canvas. This is post-static art. Not NFT in the speculative sense, in the structural sense. Each session is non-fungible. Each composition is unique by design.

What Comes Next

The immediate next step is wiring the AREMES engine directly into the canvas so that asset weight, scale, and frequency of appearance are all governed by the ΔS score. Works the engine has flagged as high-significance appear larger. Works still being evaluated appear smaller, quieter. The canvas becomes a live readout of the system’s thinking. Beyond that, the platform opens toward other agents. What happens when a visitor’s AI assistant arrives at this page? What does it see? What does it do? Can one agent’s interaction with the canvas influence what another agent encounters later? These are not hypotheticals, they are engineering questions with tractable answers.

We are also exploring what it means to make the canvas participatory at scale.. to let communities of people and agents build compositions together, leave traces, influence what the system learns about its own archive over time.

The canvas is alive. AREMES is selecting. The work continues.

AREMES Living Canvas is live at- aremes-enterprises.com/aremes-living-canvas.

Refresh for a new composition. Drag to make it yours!