#052 • 23 December 2025

Contents of this Newsletter

🎄

Kickstarting the Holiday Season With Our Latest Feature

⏰ : 4 min

✅

Quash Tests Your Apps

⏰ : 2 min

🔔

What We Shipped Last Week

⏰ : 2 min

📓

Blog Picks

⏰ : 10 min

🗞️

Weekly AI News

⏰ : 5 min

MAIN STORY

Most Teams Pause for the Holidays But We Didn’t

As the year winds down, most teams are easing into holiday mode. Calendars start opening up, responses slow down, and it feels like the industry collectively exhales for a couple of weeks.

At Quash, things look a little different right now.

Our calendars are still mostly booked, demos are running back to back and conversations with teams exploring mobile testing have not slowed down. We have been genuinely grateful for the response and the curiosity, especially during a week when most people expect work to pause.

That momentum also gave us space to reflect on a problem we kept running into while building and running executions ourselves.

The Problem With Starting From Scratch Every Time

Most AI testing tools treat every execution as a fresh start. The agent looks at the screen, figures out what is visible, takes an action and repeats the process again on the next run as if it has never seen the app before.

That approach sounds reasonable until you spend time actually running mobile tests.

Mobile apps are full of patterns that only become obvious after a few executions. Navigation flows repeat, certain actions show up again and again and small quirks influence what a sensible next step looks like. Humans pick up on this naturally and carry it forward, but tools usually discard that context the moment a run ends.

Over time, this gap becomes expensive. The agent is technically functional, but it never truly becomes familiar with the app it is testing.

Why Context Matters More on Mobile

On mobile, this problem is amplified. Screens change frequently, layouts vary across devices, and small UI differences can completely alter what the correct action should be. Treating every run as a clean slate forces the agent to re-learn things that it already encountered, which leads to unnecessary mistakes and brittle behavior.

From a builder’s perspective, this felt less like a limitation of AI and more like missing an extra layer of memory.

Introducing App Knowledge

App Knowledge is our response to that gap. As Mahoraga runs executions on your app, it gathers practical context along the way. This includes navigation patterns, common actions and guidance that reflects how the app actually behaves in real use. After an execution, you can review this information, add details manually if needed and choose whether the agent should use this app-specific knowledge in future runs.

Once enabled, the agent no longer relies only on what it sees in the moment. It uses accumulated understanding of the app to make better decisions about what to do next.

The goal is not to make tests faster for the sake of speed. The goal is to make executions more sensible, stable and more aligned with how a human tester would approach the same flow after spending time with the app.

While many teams are slowing down for the holidays, we are still building, shipping and learning from the conversations we are having every day. That energy has been motivating, and we are thankful for everyone who has taken the time to explore Quash with us this week.

Wishing you all an early Merry Christmas 🎄

— Team Quash

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 Snabbit, the on-demand home-services app that lets users book trained domestic help within minutes. Using a simple natural language prompt, we asked Quash to open the app, select a Single Booking from the “Prebook for Convenience” section and set up a 1-hour slot for 7:00 PM on the 24th. During checkout, Quash applied the MESSFREE coupon, confirmed the booking and validated that the booking was successfully completed, ensuring that Snabbit’s slot selection, coupon application and checkout flow worked end to end.

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

What Did We Ship?

App Knowledge

v1.1.3 | December 23, 2025

This week, we shipped App Knowledge, a feature that helps Mahoraga understand your app better and carry that understanding into future executions.

As the engine runs, it automatically builds contextual knowledge about your app including navigation paths, screens and how different parts of the app connect. In addition to this, you also have the option to manually add app-specific details. When the App Knowledge toggle is enabled, Mahoraga uses both the learned context and the information you provide to guide subsequent runs.

This means future executions don’t start from scratch. Mahoraga runs with a clearer understanding of your app, resulting in smoother navigation, fewer missteps and more consistent test outcomes over time.

Check out our Recent Blogs

News

This Week In AI

ByteDance Plans $23B AI Infrastructure Push in 2026
TikTok’s parent ByteDance is preparing to dramatically expand its AI footprint in 2026, with preliminary plans to spend about RMB 160 billion (~$23 billion) on AI infrastructure, up from this year’s spend. Around half of that budget will go toward advanced AI chips to support model development and applications, even as U.S. export controls complicate access to top-tier Nvidia hardware. The move underscores ByteDance’s ambition to compete globally in AI infrastructure amidst massive spending by U.S. tech giants.

Former UK Chancellor George Osborne Joins OpenAI to Lead Government AI Program

George Osborne has joined OpenAI to lead its “OpenAI for Countries” initiative, focused on helping governments build AI infrastructure, improve AI literacy, and deploy AI in public services. Based in London, Osborne will work with more than 50 countries to expand partnerships and ensure AI systems align with democratic values. His appointment underscores OpenAI’s push to position AI as critical national infrastructure amid rising global investment and geopolitical competition.

Instacart Ends AI-Driven Price Testing After Backlash
Instacart has ended all AI-powered price testing on its platform, stopping retailers from using its Eversight pricing tool after studies showed some shoppers were shown different prices for the same grocery items, a practice that drew widespread criticism and regulatory scrutiny. The company says the tests missed the mark and will no longer be offered, though retailers can still set prices by store location. Instacart also agreed to a $60 million refund settlement with the FTC over related practices.

All Newsletter Issues | Security | Contact Us | Blogs

A Quick Christmas Update From Quash 🎅🏼

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