#032 • 5 August 2025

Contents of this Newsletter

👩🏻‍💻

Design Context For QA

⏰ : 4 min

📓

Blog Picks

⏰ : 11 min

💹

Figma's IPO Debut

⏰ : 3 min

🗞️

Weekly AI News

⏰ : 6 min

MAIN STORY

QA Tools Have Ignored Design Context For Too Long

When we started building Quash, our first assumption was that code is the most reliable source of context. Code was the starting point because it reflects what is actually being built. It’s the raw material, the bricks that build the application from ground up. Our approach was simple: you have an existing application with a master branch and a feature branch. You compare the two, identify the impacted workflows and the changes, and write test cases based on those changes. Code gave us two important things: test logic and functionality.

But as people started using it, we realized that code only gives you limited context. It tells you what was done and not what was meant to be done. There’s a difference between how someone intends to perform a task and how they actually do it. Looking only at code is like inspecting a finished house without ever seeing the blueprint. You might check whether the window works, but you can’t know whether it was built in the right spot to begin with.

Code helps test logic and functionality, but it doesn’t capture intent. And to truly test a product, you need to understand both what’s built and what was intended.

Understanding How QAs Actually Work

That shift in thinking happened when we started building Quash as a tool for QAs. We asked: how do QAs actually understand what needs to be tested? Where do they get their context?

Two things stood out: PRDs and design files. That’s where intent lives. PRDs give you high-level information: the goals of a feature, why it’s being built, the business impact, the user impact, and so on. But Figma files are the closest thing to the actual product.

Design files when prototyped show you exactly how something is supposed to look and work. In Figma, flows are laid out, screen connections are marked, and designers simulate interactions to match the real product. That makes it an incredibly rich and detailed source of information.

Figma doesn’t just say, “Go to the payments screen.” It says, “Click on this specific button labeled ‘Make Payment,’ then choose between Pay Now or Pay Later.” It shows which UI elements are involved at each step. It defines the interactions.

Why Does This Matter?

Testing interactions requires detailed knowledge of how the product should behave and in automation, every step is a command: tap, scroll, swipe, long press etc. Figma gives us the action and the context to execute it correctly. That improves the accuracy of test cases significantly.

We came to this idea through two lenses. First: how do QAs actually work today? And second: what assets do teams already create to define a feature before it’s built? Figma was a clear answer to both.

QAs already refer to Figma when matching the UI. Designers run design acceptance testing to make sure the built product matches what was designed by checking styles, colors, padding, and layout. So it made sense to bring Figma into our testing pipeline. Even though we don’t do visual/UI testing yet, this integration lays the foundation for it.

From a tech standpoint, the process looks like this:

Identify flows within the Figma file. A single file might have 20 or more flows, and designers often follow consistent layout conventions. We use those to locate them.
Use vision models to map out which screens belong to which flows.
Look at individual screens to identify components and infer how users move through them.

Figma As a Design Context

Ultimately, effective testing starts with a clear understanding of what the product is supposed to do. That reference point is essential. One side is the reference; the other is the actual product. And when it comes to capturing intent, Figma is the most detailed and reliable source we’ve found.

It brings back the missing reference QA has needed for a long time.

Check out our Recent Blogs

industry spotlight

Design Tools Are Hot For Public Markets

Figma's recent IPO highlighted a truth software teams have always known: design is now a core part of the infrastructure.

The initial public offering was a massive success. Priced at $33, shares opened at $85 and closed at $115, a 250% jump that was one of the largest for a billion-dollar IPO in decades. Headlines celebrated the company's dominance, the return of the IPO market, and the long-overdue recognition of design tools.

Days later, the stock settled around the $90 range, a predictable correction after the initial surge. This rebalancing wasn't due to a scandal or poor performance, but rather the natural result of initial investor enthusiasm and market adjustments.

The real takeaway isn't about the stock's typical volatility. The product remains strong, with a massive user base and impressive growth. The true signal lies beneath the price swings: tools that shape how teams think about products are now significant enough to go public. This shift puts a spotlight on a long-standing tension: while the entire software stack is evolving, QA remains stuck on manual checklists, struggling to keep up. Figma's IPO makes this gap impossible to ignore.

news

This Week in AI

OpenAI launches Stargate Norway: Europe’s first AI gigafactory
OpenAI, in partnership with Nscale and Aker, has unveiled Stargate Norway, its first European AI data center under the OpenAI for Countries initiative. Located near Narvik in Northern Norway, the facility will house 100,000 NVIDIA GPUs by the end of 2026 and deliver 230 MW of compute capacity, with plans to expand to 520 MW. Designed to support Norwegian startups, researchers, and European institutions, Stargate Norway marks a major step toward sovereign, scalable, and sustainable AI infrastructure in Europe backed by a $1 billion initial investment and shared ownership between Nscale and Aker.

Anthropic Overtakes OpenAI in Enterprise AI Usage
According to Menlo Ventures, enterprise spending on large language models has surged past $8.4B, and Anthropic now leads the pack with 32% usage market share, ahead of OpenAI at 25%. This marks a dramatic reversal from 2023, when OpenAI held a dominant 50% share and Anthropic just 12%. In coding use cases, Claude (Anthropic) commands 42% share, nearly doubling OpenAI’s 21%. Enterprises are favoring closed-source models heavily, with open-source models now accounting for only 13% of production workloads.

Meta CEO Marks a New Era: AI That Improves Itself
Meta’s Mark Zuckerberg says his company is seeing early signs of self-improving AI, which is a critical step toward superintelligence. The newly minted Meta Superintelligence Labs aims to build agents that evolve without human input, focusing on “personal superintelligence” over mass automation. Meta continues recruiting aggressively, by luring top AI talent with nine-figure offers and plans infrastructure expansion via massive data centers.

All Newsletter Issues | Security | Contact Us | Blogs | Test Case Generator

Unsubscribe · Preferences

The Design Gap in QA: How Figma Changes Testing Forever 🌉