#025 • 17 June 2025

Contents of this Newsletter

🌄

Big Break For QA

⏰ : 5 min

📓

Blog Picks

⏰ : 11 min

💵

Meta Bets Big

⏰ : 2 min

🗞️

Weekly AI News

⏰ : 6 min

MAIN STORY

Not Everything We’ve Built Has Been Shared Yet

If you’ve been watching Quash from the outside, it may have seemed quiet lately. But behind the scenes, it’s been anything but that.

We’ve been deep in it: building, questioning, unlearning. Not just improving features, but rethinking the foundation of QA itself. While the industry kept expecting QA teams to catch up, we decided to build something that would help them go further ahead.

And it wasn’t easy.

If you’d told us a year ago that we’d be building what we’re building today, we wouldn’t have believed you. The vision was always big. But the reality has been something else entirely. Every rewrite, every rethink, every late night—it led us here.

Because every now and then, there’s a shift. Not a noisy one. But a quiet rebuilding of how something works at the core.

The Old QA Model Is Broken

We’ve learned that much of what passes as “best practice” in QA is built on outdated assumptions. Some of these assumptions include:

Writing tests by hand is scalable.
It’s not. Too often, time is lost, releases are blocked, and the edge cases that matter gets missed. Teams spend hours writing and rewriting tests, but it still doesn’t keep pace with modern development.
Flaky, unreliable execution infrastructure is just something teams have to live with.
They shouldn’t. But for years, broken CI runs, inconsistent local setups, and test results no one could trust were treated as normal. It delayed feedback and drained time—every single day.

These assumptions are dead wrong and they were limiting what QA could be. We had to let go of the idea that QA is a separate phase or team. It’s not. It’s a layer across the stack of design, product, infrastructure, code—all woven together perfectly. That’s the work we’ve been doing. And that’s what’s starting to ship now.

A New Foundation for QA

Everything we’re building now comes from a different set of beliefs:

That engineers want to test—but only if the tools work with them, not against them.
That knowing what can break is more valuable than what’s covered.
That testing shouldn’t lag behind the product. It should evolve with it.

Some of what’s coming includes a fully-optimized version of our PRD-to-test case generator—faster, smarter, and tuned for scale. Real test coverage, in minutes. And test execution that is finally stable, whether you’re running it locally or on a device farm.

We’re not done yet. But what’s coming is the culmination of everything we set out to achieve, to redefine what QA can be.

We Refuse to Settle For What QA Has Become

Too many QA tools were built to check boxes, not to solve problems. They reward fake progress—like test coverage stats and rerun pass rates, while the real issues keep slipping through.

We don’t want to play that game. We want to change it. If you’ve ever felt like QA was slowing you down, working against you, or just not doing enough—you’re not alone. We’ve felt it too. And we’ve built this for all of us who are done settling.

So no, not everything we’ve built has been shared yet. But it’s coming. And when it does, we think it will change the way you work.

Stay close. You’ll want to see what’s next.

Check out our Recent Blogs

social

Meta Bets $14.5B on Scale AI and Everyone Backs Away

Meta just dropped $14.5 billion for a 49% stake in Scale AI—the startup quietly powering much of the AI boom by labeling the data models like GPT rely on. It’s Meta’s biggest external AI investment ever. As part of the deal, Scale’s CEO is heading up Meta’s new “superintelligence” lab, while still staying on Scale’s board. Translation: Meta now has a front-row seat to the training data pipelines used by its biggest competitors.

That’s where things get messy. Google, Scale’s top customer, is reportedly cutting ties. Microsoft and xAI are reconsidering their relationships. These big players have every right to be concerned as Meta might now have indirect access to everyone’s proprietary data workflows. OpenAI, still working with Scale, says it’s diversified—but it started dialing back its reliance months ago. With CEO Alexandr Wang heading to Meta to lead its new superintelligence lab, and a number of key Scale employees joining him, the boundary between Meta and Scale is starting to blur. Wang will remain on Scale’s board, but his dual role raises real questions about how independent the company can stay, especially as Meta now holds a 49% stake.

Rui Ma

@ruima

😂 😂 this FT quote on Scale AI cofounder Alexandr Wang

3:51 AM • Jun 11, 2025

Retweets

1683

Likes

Read 39 replies

This isn’t just about competition—it’s about control. Scale was supposed to be neutral infrastructure. Now it’s a strategic asset with divided loyalties. And just like that, the trust in shared infrastructure just changed.

news

This Week in AI

OpenAI Launches o3-pro, Its Most Capable Reasoning Model Yet
OpenAI has released o3-pro, a more advanced version of its o3 reasoning model, now live for ChatGPT Pro and Team users (with Enterprise and Edu next week). Built for step-by-step thinking in domains like math, physics, and code, o3-pro replaces o1-pro and is now also available via API. Pricing: $20M/input tokens, $80M/output—about 750K words per million tokens.

Mistral Unveils Magistral — Its First Reasoning Model
Mistral AI has launched Magistral, a new reasoning model built for domain-specific expertise, transparent logic, and multilingual fluency. Available in open and enterprise editions, Magistral tackles the gaps in current reasoning models by offering step-by-step analysis you can trace, trust, and translate.

Apple Debuts Custom Foundation Models for On-Device AI
Apple’s new language models—optimized for Apple silicon and Private Cloud Compute—power Apple Intelligence across devices. The lineup includes a 3B on-device model and a Mixture-of-Experts server model, supporting 15 languages, multimodal inputs, and better reasoning. Developers can now tap into these via the new Foundation Models framework, bringing powerful AI features to apps while keeping user data private.

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

QA Tools Are Failing + Meta's 💸 Billion Dollar Betrayal!