#056 • 20 January 2026

Contents of this Newsletter

🤖

The age of product-led growth to the age of agent-led growth

⏰ : 4 min

✅

Quash Tests Your Apps

⏰ : 4 min

🔔

What We Shipped Last Week

⏰ : 2 min

📓

Blog Picks

⏰ : 10 min

🗞️

Weekly AI News

⏰ : 5 min

MAIN STORY

Are Humans Still the Customers?

It sounds like a strange question at first. People still use software. They still care about how it looks, how fast it feels and whether it does what it promises. None of that has suddenly disappeared. But there is a growing sense across the industry that humans may no longer be the first ones deciding which software even gets a chance.

That idea came into sharper focus recently after a conversation on TBPN, the daily live tech talk show and podcast, featuring Sequoia Capital partners Sonya Huang and Pat Grady. In the discussion, Sonya suggested that product-led growth, as we have known it for the last decade, is starting to lose its edge. Her point was not that humans no longer matter, but that AI agents are increasingly doing the early evaluation work that humans used to do. They scan documentation, compare APIs, weigh trade-offs and shortlist tools long before a person clicks a website or starts a free trial.

It is a provocative take and like most provocative takes, it is easy to oversimplify. Product-led growth is not dying overnight. Humans are not disappearing from the loop. But something meaningful is changing in how software gets evaluated and that change has implications for how products are explained, documented and positioned.

Why Product-Led Growth Worked So Well

Product-led growth thrived because it aligned closely with human behavior. Humans skim and jump between tabs. They make decisions with incomplete information and move on quickly if something feels confusing or slow. Most people do not read documentation end to end. They rarely explore edge cases unless something breaks. They respond to familiarity, polish and momentum.

As a result, products were designed to compress complexity. Onboarding flows were optimized to feel effortless. Rough edges were hidden wherever possible. Messaging focused on clarity and confidence rather than completeness. This was not a flaw in the model. It was a rational response to how humans actually evaluate tools.

What Changes When Agents Do the Evaluating

AI agents do not behave like humans. They do not get tired or impatient. They do not skim because they are busy or distracted. They do not stop at the landing page or take confident messaging at face value.

When an agent evaluates a product, it reads the documentation in full. It cross-checks claims made on the homepage against what the docs and APIs actually support. It compares behavior across versions and looks for inconsistencies over time. Edge cases that humans often tolerate or work around become first-class signals. Where humans are forgiving, agents are methodical.

This is where many assumptions baked into product-led growth begin to weaken. What worked because humans were willing to fill in the gaps starts to break down when evaluation becomes thorough and literal.

Messaging and Content Now Have a Second Audience

This shift is often described as a move from human-led growth to agent-led growth, but that framing misses something important. The deeper change is about who the product is being explained to during evaluation.

Landing pages, for example, are no longer the primary source of truth. They still matter for humans, but agents treat them as summaries rather than evidence. A claim like “reliable at scale” only carries weight if it is backed up by documentation that clearly explains limits, failure modes and real-world behavior. Any mismatch between surfaces becomes a negative signal.

Documentation also plays a different role. It is no longer just an onboarding tool for users who have already chosen you. It is part of the decision-making process itself. Clear explanations of constraints, retries and error handling often matter more to an agent than polished examples of the happy path.

Pricing is another area where this shift becomes visible. Ambiguous tiers or hidden limits are manageable in a human-led sales process, but they slow down or block automated comparison. Pricing pages that clearly define usage units, limits and overage behavior are easier for agents to evaluate and easier to recommend.

In all of these cases, messaging is no longer just about persuasion. It is about being legible under scrutiny.

What This Shift Quietly Rewards

None of this means that humans no longer matter. People still use software, and experience still counts. But persuasion is no longer the primary gatekeeper it once was. Evaluation is becoming more systematic and less forgiving.

AI does not replace products. It reveals how well they are explained and how closely reality matches what is promised.

As agents become part of how software gets chosen, the qualities that matter most are the boring qualities that hold up when systems are examined closely. Things like predictable behavior, clear contracts, transparent pricing, consistency across documentation and more.

The companies that benefit from this shift will be the ones that are easiest to understand, both for humans who eventually use the product and for machines that quietly decide whether it deserves attention at all.

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 Rapido, the ride-hailing and logistics app best known for its bike taxis and quick parcel deliveries. Using a natural language prompt, we asked Quash to open the app, switch to the Parcel flow, and select Home as the drop address. It then filled in the contact details, chose the parcel category for items up to 10 kg, and completed the booking. This run validated Rapido’s parcel booking flow end to end, from address selection and contact entry to final confirmation.

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

What Did We Ship?

Improved Onboarding and UX Fixes

v1.1.7 | January 20, 2026

This week, we focused on making the first experience with Quash smoother and easier to get started.

We’ve rolled out an improved onboarding flow that guides new users through setting up their initial devices with more clarity and fewer steps. The goal is to reduce early friction and help teams reach their first successful run faster, without needing extra setup or guesswork.

Alongside onboarding, we’ve shipped a set of UI and experience improvements across the app. Together, these changes make Quash easier to pick up, navigate and more comfortable to use from day one especially for teams getting started for the first time.

Check out our Recent Blogs

News

This Week In AI

OpenAI to Start Showing Targeted Ads in ChatGPT

OpenAI announced that ChatGPT will soon display targeted advertisements to users on its free and ChatGPT Go subscription tiers, marking a major shift in its monetization strategy. Ads will show up at the bottom of conversations and will be relevant to the topic being discussed, with users able to dismiss or turn off personalization. OpenAI has stressed that these ads will not influence ChatGPT’s answers, that user data won’t be sold to advertisers, and that ads won’t be shown to users under 18. This move aims to offset rising infrastructure costs and could also drive some users toward paid tiers to avoid ads.

Google Expands Its AI Video Generator Flow to Workspace Users

Google has expanded access to its AI-powered video creation tool, Flow, making it available to users on Google Workspace Business, Enterprise, and Education plans (in addition to earlier availability on AI Pro/Ultra). Flow lets users generate short video clips from text prompts or images using Google’s Veo 3.1 generative model and stitch them into longer sequences. It also includes tools for adjusting lighting, camera angles, object insertion/removal, vertical video and audio generation — turning AI video creation into a practical productivity tool for teams, educators and creators within the Workspace ecosystem.

xAI Activates World’s First Gigawatt-Scale AI Training Supercluster

Elon Musk announced that xAI’s Colossus 2 supercomputer is now operational, marking the first gigawatt-scale AI training cluster in the world. This infrastructure milestone gives xAI a massive computational platform to train large AI models, which is a strategic advantage in the race with rivals like OpenAI and Anthropic. xAI plans to expand Colossus 2 to 1.5 GW and beyond by April, signalling an ongoing push to scale AI compute faster than competitors.

All Newsletter Issues | Security | Contact Us | Blogs

Are Humans Still Choosing Software❓

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