#045 • 04 November 2025

Contents of this Newsletter

🗿

A New Milestone: Backend Validations in Quash

⏰ : 5 min

✅

Quash Tests Your Apps

⏰ : 4 min

🔔

What We Shipped Last Week

⏰ : 2 min

📓

Blog Picks

⏰ : 10 min

🗞️

Weekly AI News

⏰ : 7 min

MAIN STORY

Our Agent Was Always Meant For More Than The UI

From the beginning, we knew a mobile-use agent that interacts with apps like a real user had potential far beyond simple automation. It was exciting to see a system that could tap, scroll, and type on real devices by following natural language instructions, because that already offered a new way to run apps end to end without maintaining brittle scripts. For QA teams, this is a refreshing alternative to the usual framework setup, since the agent could behave like a real person and reveal issues that only appear during genuine usage.

Even once Mahoraga started showing that power visually on screen, through complete flows and real execution results, a question remained: how should we help users understand that the value is deeper than what appears in the UI? We received a comment from a QA at Leap Scholar that captured this perfectly:

That line did not discourage us at all. In fact, we loved that comment. Because it said exactly what we were thinking. It affirmed that people who understand quality also understand that an app is only as reliable as its backend resilience. It signaled that users were ready for a smarter kind of testing. Internally, we smiled because this is exactly where we were heading.

Real Quality Is Not Screen Deep

Most of what determines whether an app works never appears visually. When a price updates correctly only after a discount rule is processed, when a ticket gets booked only after an order is written into a database, when a login succeeds only because the right token has been generated, the success of those flows is invisible to the eye and visible only in the backend.

Current AI testing tools are exciting because they take on the burden of covering flows through the interface, yet the community is very clear about one limitation: testers want systems that do not only observe screens but also verify that business logic executed correctly behind those screens. There is strong optimism that AI can accelerate this category of testing and reduce manual checks across APIs and services, although people expect reliability and transparency before they fully trust it. No one wants a “magic box” that claims something passed without proving why it should be trusted.

This is exactly the intersection where Quash is choosing to operate, because quality cannot be evaluated by visuals alone and workflows are not fully validated until backend consequences have also been confirmed.

Testing the Logic That Runs the App

We have been working quietly on a deeper capability and we are excited to share that backend validations are now part of executions in Quash. This means the agent can run a flow exactly like a user and also confirm that the backend handled the flow correctly. UI actions and backend checks stay within a single test rather than being stitched together across tools.

Here is what this brings into day to day testing: validations allow you to define the method, endpoint, and request data once and reuse them across tasks, which keeps tests consistent and reduces effort. You can configure authentication, base URLs, and retry logic for real environments without repetitive setup, which helps when endpoints are occasionally unreliable. You decide what counts as success by specifying status codes, JSONPath comparisons, or regex checks, so the correctness criteria are explicit and aligned with real business rules. Finally, every run provides results that can be inspected directly in the execution timeline and in reports, with pass or fail details, latency, and a redacted response preview that gives confidence in what was verified.

The experience ties directly back into what a user is actually doing. If they add an item, the backend total should increase. If they log in, the token should be valid. If they check out, the order and inventory should update. Because Quash looks at the truth that lives behind each action, bugs that would otherwise slip through become visible immediately.

This is the first time a mobile AI agent does not just interact with an app but also confirms that the complete system responded correctly. UI validation alone tells only half the story. Backend validation completes it. QA finally receives evidence from both sides of the interaction.

This is the direction AI testing needs to go, because confidence does not come from watching an app behave correctly on screen but from proving that the platform behind it delivered the intended results.

Your Apps, We Test

Each week, we put our execution engine to work on real apps.

This week, we put our execution engine to work on Leap Scholar, the EdTech platform helping students pursue study-abroad opportunities. Before running the test, we set up a backend validation for Leap Scholar’s course API so Quash could check data accuracy in real time during execution. Using a simple natural language prompt, we asked Quash to open the app, head to Explore, and search for MBA programs. It selected the first college from the results, added it to the shortlist, returned to the list, and did the same with the second college. Then it navigated to the Shortlist section to confirm both colleges were saved correctly. With the validation in place, Quash ran the @get_courses check and verified that the college data shown in the app matched the API response, ensuring the UI and backend remained perfectly aligned.

Want us to test your app next? Drop us a mail at hello@quashbugs.com

What Did We Ship?

Backend Validation

v1.0.4 | November 4, 2025

This week, we shipped Backend Validations, enabling Quash to verify the API and business logic behind every UI action. As your tests run, Quash now checks the responses from the endpoints that power your flows, ensuring the app is doing the right work under the hood.

You can define validations once with the method, endpoint, and expected result. Quash will reuse them across tasks, handle environment authentication, and even retry when things get flaky. Success criteria can include status codes, JSONPath targets, or regex patterns to match the exact response you expect.

Check out our Recent Blogs

News

This Week In AI

OpenAI Signs $38 B Cloud Deal With Amazon Web Services
OpenAI has signed a seven-year, $38 billion cloud computing deal with Amazon Web Services that gives the company access to large-scale infrastructure, including hundreds of thousands of Nvidia GPUs. The agreement ends OpenAI’s previous cloud exclusivity with Microsoft and is expected to provide full deployment capacity by late 2026, with options to expand further in the following years.

NVIDIA and Oracle to Build US Department of Energy’s Largest AI Supercomputer

NVIDIA and Oracle have announced a major collaboration with the Department of Energy to build the largest AI supercomputer in the U.S. at Argonne National Laboratory. The system, named “Solstice”, will include 100,000 NVIDIA Blackwell GPUs with another system, “Equinox”, featuring 10,000 Blackwell GPUs; together they’ll deliver around 2,200 exaflops of AI performance and are aimed at advancing scientific research across energy, materials, and security domains.

Anthropic is Making Waves

Tech giants Alphabet Inc. and Amazon.com, Inc. are reporting multi-billion-dollar paper gains thanks to their investments in the AI startup Anthropic, the maker of the Claude chatbot. Google’s net gains on equity securities reached $10.7 billion, while Amazon recorded a $9.5 billion pre-tax boost—both largely driven by Anthropic’s soaring valuation and the infrastructure deals underpinning it. Founded in 2021 by former OpenAI employees, Anthropic has rapidly become one of the most valuable AI companies globally, reportedly valued at around $183 billion.

All Newsletter Issues | Security | Contact Us | Blogs

How Did We Teach Our Agent to See Beyond the Screen 👩🏻‍🏫

Product

Solutions

By role

By use-case

By industry

Resources

Comparison

Company

Product

Solutions

By role

By use-case

By industry

Resources

Comparison

Company

Get Quash

Product

Solutions

By role

By use-case

Resources

Comparison

Company

Get Quash