Kortshut

Designing a context-first operating system.

An ongoing exploration into AI-native workflows — and what computing looks like when context matters more than applications. This documents how the thinking evolves, not a finished product.

Role
Co-Founder & Product Designer
Timeline
2025 – Present
Team
3 founders · Product · Eng · AI
Platform
macOS

Focus areas

  • Product Strategy
  • AI Workflow Design
  • macOS Product Design
  • Information Architecture
  • Human–AI Interaction
  • Design Systems
  • AI-Assisted Development
Kortshut running on macOS

The shift

LLMs didn't eliminate work — they shifted where work happens.

The bottleneck was no longer writing. It became assembling enough context for AI to produce meaningful results.

Every design review, engineering task, research session, or strategy doc followed the same pattern: capture information, screenshot, copy text, switch apps, reconstruct context, prompt an AI, return to work — and repeat.

The models improved rapidly. The workflow did not. Kortshut began as an exploration into reducing that friction.

The question

What happens when computers understand context instead of applications?

Observation

Computers organize work around files and applications. People organize work around context.

A designer isn't thinking about Figma — they're thinking about redesigning onboarding. A developer isn't thinking about Finder — they're thinking about solving a bug using notes, screenshots, logs, and previous discussions.

Current operating systems understand where files live. They don't understand why they belong together. That observation became the foundation for every subsequent product decision.

Fragmented context across Figma, browser, Claude, Cursor, screenshots, clipboard, and notes

Early hypothesis

“If we make AI easier to access, people will work faster.”

Early concepts focused on faster prompting, keyboard shortcuts, AI launchers, and clipboard improvements. They improved speed — but failed to address the larger issue.

The real bottleneck wasn't opening AI. It was rebuilding context before every conversation.

Research & continuous discovery

Kortshut evolved through continuous observation and dogfooding — not a fixed philosophy.

The same patterns kept surfacing: screenshots became temporary memory; clipboard contents disappeared despite remaining valuable; users repeatedly rebuilt the same prompt context; AI quality depended more on context than prompt wording; and keyboard shortcuts only mattered when attached to repeatable workflows.

That shifted the product away from “AI utilities” toward workflow orchestration.

~1yr
Of product exploration
~15
Workflow & onboarding iterations
20
Active beta users
3
Person founding team
100s
Design explorations across Figma, Claude Code, Cursor, Codex & Stitch
Emerging principle

Context is more valuable than prompts.

Why context, not prompts

Early versions emphasized prompts. Research suggested users cared far less about prompts than outcomes. The better question became: how quickly can someone gather everything AI needs without breaking their flow? This principle continues to shape the roadmap.

Exploration 01 — Persistent context

What if copied information wasn't disposable, but persistent working memory?

Instead of viewing the clipboard as transient, we explored it as persistent working memory. Historical clipboard items, screenshots, and captured references became reusable pieces of context rather than temporary artifacts.

The goal wasn't to remember what had been copied. It was to recover thinking.

Clipboard architecture — copied content as persistent, searchable working memory

Exploration 02 — Workflow shortcuts

Keyboard shortcuts were never the destination — they became vehicles for repeatable workflows.

The design challenge shifted from reducing clicks to preserving cognitive flow.

The workflow, compressed

Traditional workflowExplored workflow
Capture, copy, screenshotOne shortcut
Switch apps, open AIContext assembles automatically
Paste, rebuild the promptAI responds in place
Return, repeatKeep working
Keyboard-driven workflow — a shortcut that assembles context and returns AI in place
Decision log

We stopped designing around prompts.

Initial belief
Prompts were the primary unit of interaction.
Observation
Users repeated workflows more consistently than prompts.
Current direction
Design around reusable workflows that naturally contain prompting.
Decision log

The clipboard became working memory.

The shift
Copied content stopped being transient and became searchable, referenceable, and context-aware.
Status
This remains an active area of product exploration.

What we chose not to build

Kortshut intentionally avoids becoming another chatbot, another launcher, another note-taking app, or another prompt marketplace. Those categories already exist. The opportunity lies in connecting them through context.

Designed while adopting AI

Kortshut was designed while actively adopting AI throughout the process. Claude Code, Cursor, Codex, Google Stitch, and Figma AI accelerated prototyping, implementation, and experimentation — shortening the feedback loop between hypothesis and validation, rather than replacing design thinking.

AI-assisted product development — hypothesis → prototype → validate → iterate

How my thinking changed

Earlier thinkingCurrent thinking
AI needs faster access.AI needs better context.
Prompts are the product.Workflows are the product.
Clipboard history adds utility.Persistent context enables better decisions.
Keyboard shortcuts save time.Embedded workflows preserve focus.

Open questions

  • Should context be assembled manually, or inferred automatically?
  • Is the clipboard the right primitive — or is context itself the primitive?
  • When does automation become invisible enough to feel natural?
  • How should AI balance initiative with user control?
Reflection

Designing AI products is rarely about designing better AI. It's about designing better systems for people to capture, preserve, retrieve, and reuse context.

An active exploration

Kortshut remains an active exploration into that problem. Rather than documenting a finished product, this case study documents an evolving way of thinking about human–computer interaction in an AI-native world.

The product remains alive. The thinking continues.