# Aleyna Çatak

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Aleyna Catak](https://cerebralvalley.ai/u/aleynacatak)
- **GitHub:** https://github.com/aleynacatak/doppel/
- **Demo video:** https://www.youtube.com/watch?v=qmbX_xrCovY
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3945

--FOR JUDGE DEMO CODES, PLEASE SEE BELOW--

Doppl: your design apps talk to each other.

I'm a designer from Istanbul. A big part of my week isn't design. It's production: filling the same template forty times, cutting out product photos, making mockups, across Photoshop, Figma and InDesign. Most AI tools for this run a model on every single task. I wanted the opposite: use AI once to learn a task, then repeat it without AI.

Doppl is a macOS app. I describe a production task once, in plain words, and Meta's Muse Spark turns it into a routine: a saved set of steps I can actually read, which runs inside my own apps. Photoshop and InDesign routines run as native scripts; Figma routines run through Doppl's own plugin. After that, a routine reruns with zero AI calls, the same way every time.

How a routine is made: Muse Spark inspects my files, then builds the routine from tested building blocks. Where no block fits, it writes custom code, which Doppl marks so I always know what's new. Before I save, Doppl tries the routine on a copy. Muse Spark looks at the result and gets a chance to fix what it finds, and SAM 3.1 checks that key objects, like a logo, aren't cut off. Anything still wrong is flagged for me. Each attempt has a budget of 15 AI calls.

Flows: every routine declares its inputs and outputs (spreadsheet rows, files, a document), so routines chain into flows across apps. A flow connects outputs to inputs and doesn't care which app is on either side.

In the demo: Jam-a-dox, a made-up jam brand. Photoshop cuts out the fruit, Figma makes a label for every row of a spreadsheet, and Photoshop puts each label on a jar. Setting up the flow took about 40 AI calls (around $1.30). Rerunning it for six new flavors took about a minute, with no AI.

File safety: Doppl works on copies by default, won't save a routine with unsafe steps like deleting files, and sorts out messy inputs before each run (semicolon CSVs, Turkish characters, files named "FINAL (1)"), using AI only if it can't place a spreadsheet column.

Who it's for: designers who want to automate the repetitive part of their work without being developers, and without handing the design itself to AI.

Built from scratch during the hackathon, Oct 27 – Nov 4, 2026.

--DEMO CODES--

For judges, to try it live: install Doppl from the Releases page (steps in the README), then go to Settings → AI access and enter one of these codes, one per judge. Each allows 20 routine creations and is valid until Nov 20.

DOPPL-LHCD-G97C-Y6TN
DOPPL-FPFZ-SLDW-ZHUY
DOPPL-7HXU-LGQ8-AZYV
DOPPL-B8DL-KXVE-V6RW
DOPPL-NXMP-ERAD-7Q6U
DOPPL-G99W-3K79-PXFT
DOPPL-PWJ8-45RV-CGZA
DOPPL-9MLG-C6CC-CJE4
DOPPL-TB26-9Q9A-7QD9
DOPPL-AJFG-ZV7R-TH6M

You need an Apple Silicon or Intel Mac and InDesign, Photoshop or the free Figma desktop app. Please keep these codes private; their uses are limited.

## More from Meta Global AI Developer Hackathon

- [OmniMind](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3942)
- [Digikeys](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3943)
- [Civic Teach BD](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3944)
- [SIRA](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3946)
- [3ill](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3947)
- [The Wolf Pack](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3948)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3945. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt