#018 • 29 Apr 2025

Contents of this Newsletter

🏢

Frontier Firms

⏰ : 9 min

📓

Quash Blogs

⏰ : 8 min

🎤

PM’s Playbook

⏰ : 5 min

🗞️

Weekly AI News

⏰ : 4 min

MAIN STORY

Microsoft Predicted Frontier Firms But We Are Already Living It

In the recent 2025 Work Trend Index annual report by Microsoft, a new term has emerged that’s gaining significant attention: Frontier Firms. These are organizations that are fully embracing the collaboration between humans and AI. According to the report, frontier firms are predicted to become productivity powerhouses, using AI to enhance work efficiency while fostering a culture where humans and AI can thrive together. The report focuses on this shift, forecasting that organizations will increasingly adopt AI agents that can reason, plan, and act autonomously. We’re entering an era where business as usual is no longer an option. The future of work is one where knowledge workers and digital labor coexist seamlessly, scaling businesses and unlocking new value faster than ever.

But here’s the thing: Quash has been a Frontier Firm since day zero.

We didn’t wait for the AI revolution to catch up to us because we’ve been living and breathing AI-human collaboration in our company and QA long before it became mainstream. While the rest of the world is catching up, we're realizing something crucial: the relationship between AI and work isn’t as simple as making tasks more efficient. In fact, it might just be more complicated than anyone expected.

We All Thought AI Was Gonna Replace Us

We were all under the impression that with the rise of AI, people would lose jobs, that AI would automate away entire industries and leave workers scrambling for new roles. AI has already demonstrated its ability to accelerate productivity, but what happens when efficiency leads to even more work? This is the reality we’re facing, and it’s an economic paradox known as Jevons Paradox.

The theory, first proposed by British economist William Stanley Jevons in 1865, states that increased efficiency in resource use tends to increase demand for that resource, rather than decrease it. Think of it this way: When AI makes QA faster and cheaper, companies don’t just stop testing they test more. The more efficient we make the process, the more work we create. It’s a feedback loop that accelerates, rather than reduces, the need for human effort.

Satya Nadella

@satyanadella

Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of. https://en.m.wikipedia.org/wiki/Jevons_paradox

11:18 AM • Jan 27, 2025

2045

Retweets

13125

Likes

Read 821 replies

How We Built a Frontier Firm?

While others are just beginning to experiment with AI-human collaboration, we’ve been designing it from the start. Our philosophy is simple: AI isn’t just a tool for automating tasks, it’s a partner that scales testing to a level we couldn’t imagine in a world of traditional QA.

At Quash, we’ve reimagined what knowledge work in QA should look like. By integrating AI at the core of our process, we’ve moved beyond simple automation of repetitive tasks. AI doesn’t just handle the mundane; it acts as a reliable partner, working alongside human testers to deliver results faster and with greater accuracy.

Think of us as the exoskeleton for QA. We can take on the time-consuming, repetitive work like regression testing or running countless scenarios, which frees up human testers to focus on higher-level activities. This partnership between AI and humans allows us to solve more complex problems, identify issues earlier in the cycle, and provide strategic insights that drive quality across the entire development process.

Here’s where Jevons Paradox comes into play: efficiency breeds more work, not less. When you make something more efficient, the scope of what’s possible widens, and the demand grows. This is the new and interesting reality of working with AI.

Is Vibe Testing More Than Just a Tool?

The truth is, AI’s impact on productivity is about creating the space for more work to happen. The more efficient AI becomes, the more there is to do.

Take vibe testing, for example. Just like “vibe coding” allows non-coders to build apps or websites using simple AI prompts, vibe testing is pushing the boundaries of QA. We’ve moved beyond traditional testing tools. AI now takes on repetitive, data-heavy tasks, while humans focus on strategy, insight, and judgment. But the paradox is this: We’re not doing less, rather we’re doing more.

Testing cycles are faster, more tests are run, and the overall quality assurance process is scaled at a level that wasn’t possible before. And as efficiency increases, so does the demand for more testing, more coverage, and more iterations.

The Future Of Work Is Not What Any Of Us Expected

The future of work is already here and it’s not just automation. At Quash, AI has always been the default. We see it not as a shortcut, but as a way to scale smarter. And as Jevons Paradox reminds us, efficiency breeds more work, not less.

For decades, intelligence was one of the most valuable and limited assets in business, bound by human time, energy, and cost. That’s changing. AI is turning intelligence into a durable good: abundant, affordable, and available on demand. With agents that can reason, plan, and act as digital labor, companies are gaining capacity at a scale we’ve never seen before.

How do you stay employable as AI gets better?

Use the tools. The productivity gap between those who leverage AI and those who don’t is growing fast.
Understand the architecture. These models aren’t magic. Know what they can and can’t do. Artificial General Intelligence isn’t human intelligence and it never will be.

The real differentiator is knowing how to work with AI.

Check out our Recent Blogs

Mastering Performance Testing: Key Metrics, Methodologies, and Tools

Performance testing isn't just about speed - it's a strategy. From load...

Read article

The Complete Guide to Smoke Testing vs Sanity Testing in 2025

Confused between smoke and sanity testing? This guide breaks down their...

Read article

Event

The PM’s Playbook by Quash

It’s never been easier to build something. With AI unlocking new possibilities almost daily, product managers today have more tools and more opportunities, than ever before. That energy was unmistakable at PM’s Playbook, an event we recently hosted in partnership with Women in Product India, where a packed room of product leaders, builders, and curious minds came together to explore how AI is changing the craft of product management.

The day kicked off with a session on building AI features, led by Hritika Mehta from Truecaller. She showed just how quickly PMs can move from an idea to a working MVP using low-code and no-code tools, making it clear that AI innovation doesn't have to wait for a full-stack engineering team. From there, Pragya Saxena from Healthify challenged everyone to rethink product design entirely—starting not with screens, but with user outcomes. Her workshop on AI-driven personalization had the room buzzing, as people mapped user intents and imagined how AI could tailor experiences in ways traditional product thinking often misses.

Later, Dhruval Rana from Groww sparked a lot of curiosity by reframing the conversation around content. Instead of focusing on creating endless new material, he talked about using large language models to curate and orchestrate experiences that feel effortless for users. It was a simple but powerful shift that had many in the audience rethinking how they approach user engagement.

The panel discussion, featuring Rishabh Nayan (Jupiter), Daren Lobo (Ex-Jar, Flipkart), Swati Awasthi (Women in Product), and Ameer Hamza (Quash), brought honest reflections on what it really takes to ship AI-driven products—moving beyond the hype to focus on real challenges, early wins, and how mindsets need to evolve alongside technology. The audience kept the conversations sharp, asking practical, grounded questions about building real AI-driven products.

Huge thanks to everyone who made this event possible—from the team behind the scenes, to DocketAI for providing the space, and to every attendee who showed up curious, open-minded, and ready to dive into what’s next.

We’re just getting started. If you want to be the first to hear about upcoming events and stay part of the conversation, join our WhatsApp community here.

news

This Week in AI

Alibaba unveils Qwen3: Hybrid AI models that rival Google and OpenAI

Alibaba has released Qwen3, a new family of hybrid AI models, with sizes ranging from 0.6 billion to 235 billion parameters. The models are available for download under an open license via Hugging Face and GitHub. Alibaba claims Qwen3 matches or outperforms top models from Google and OpenAI, putting pressure on American labs while also facing geopolitical challenges over access to chips for training.

IndiaAI Mission shortlists Sarvam AI, Soket AI Labs, Gnani.ai to build indigenous LLMs

India’s AI push is picking up pace. As part of the ₹10,000 crore IndiaAI Mission, MeitY is set to back Sarvam AI, Soket AI Labs, Gnani.ai, and Gan.ai to develop foundational models for the country. Soket AI Labs aims to build a 120B parameter Indic LLM in 12 months, while Sarvam AI has already begun work on a 70B parameter multimodal model. Gnani.ai and Gan.ai are focusing on smaller models and speech tech.

Perplexity launches voice assistant on iOS

Perplexity’s AI-powered voice assistant is now live on iOS, expanding from its earlier Android debut. Announced by the Nvidia- and Bezos-backed startup, the assistant lets users hold fluid voice conversations, and even take actions like booking a ride, checking calendars, or sending emails—all without leaving the app. Read more on how to use.

OpenAI’s viral image model lands in the API

OpenAI has released gpt-image-1, the image model behind ChatGPT’s viral creations to its API. Devs can now plug high-quality image generation, editing, and text-in-image features directly into their apps. Pricing: $10/M input tokens and $40/M output tokens.

ByteDance drops Seedream 3.0 to challenge GPT-4o, Imagen 3

ByteDance has launched Seedream 3.0, its new bilingual image generation model said to outperform OpenAI’s GPT-4o in image quality. Trained on double the data of its predecessor and featuring smarter sampling techniques, Seedream 3.0 can generate images up to 2K resolution—and does so up to 8x faster without sacrificing quality.

Meta launches PerceptionLM: open 8B vision-language model for video understanding

Meta has released PerceptionLM, an open-source 8B vision-language model trained without distillation. It comes with 2.8M human-labeled video QA pairs and a new benchmark suite, PLM-VideoBench, to evaluate detailed video understanding.

Inside Microsoft’s Frontier Firms: AI × QA Paradox 🚀