Hey — It’s Nico.

Welcome to another Failory edition. This issue takes 5 minutes to read.

You might have noticed I didn’t send out Tuesday’s email. The reason is I’ve been rethinking the schedule a bit.

Running Failory while juggling my day-to-day at daydream has been a lot, and sticking to two emails per week means I’m not dedicating time to other, more important things I want to improve at Failory.

So for now, I’m making some temporary changes to the newsletter schedule and I will be sending only one weekly email, while I figure out what makes the most sense. Not sure what that will look like just yet, but I’ll keep you posted.

Anyway, let’s dive into today’s issue! These are the 5 most important things:

• Cushion, another FinTech startup, shut down after 8 years — learn why below.

• YC has shared their Spring 2025 “Requests for Startups”.

• Stripe acquired the stablecoin platform Bridge for $1.1B.

• OpenAI released a new ChatGPT agent for ‘deep research’.

• OpenAI’s Operator is powerful — learn how to use it below.

Let’s get into it.

This Week In Startups

🔗 Resources

The $1M ARR growth playbook.

What AI startups can learn from the Golden Age of Enterprise Sales.

YC’s Spring 2025 Request for Startups

How to differentiate your product.

📰 News

Stripe acquired the stablecoin platform Bridge for $1.1B.

Google wants Search to be more like an AI assistant in 2025.

OpenAI unveils a new ChatGPT agent for ‘deep research’.

AI systems with ‘unacceptable risk’ are now banned in the EU.

💸 Fundraising

Alice & Bob raises €100M to build Quantum Computers by 2030.

Smart grid startup Ionate raises $17M.

Qeen.ai, a startup making AI agents for ecommerce, has raised $10M.

StackAdapt raises $235m to advance programmatic advertising.

Fail(St)ory

Another Fintech Fail

This week, Cushion, a fintech startup that aimed to help consumers fight bank fees, has shut down.

Cushion launched in 2016 with a simple but valuable promise: identify hidden fees in users’ bank accounts and negotiate refunds on their behalf. The idea resonated with consumers, and in its early years, the company saw rapid growth.

But despite reaching $3M in ARR within 10 months and onboarding over 1 million users, Cushion couldn’t scale profitably. Like many fintech startups before it, the company struggled to sustain its model, ultimately shutting down after nearly eight years in business.

What Was Cushion: Cushion was built to solve a common but frustrating problem: bank fees.

By connecting to users’ bank accounts, the app would automatically detect fees, overdraft charges, and other penalties, then negotiate refunds on behalf of users—without them having to lift a finger.

The idea stemmed from founder Paul Kesserwani’s personal experience. While helping his parents manage their finances remotely, he saw firsthand how quickly banking fees could pile up. Digging deeper, he realized he had unknowingly paid $400 in fees himself. That insight led to the creation of Cushion, which sought to give consumers more control over their finances.

Over time, the company expanded its mission beyond fee negotiation, aiming to help consumers manage recurring bills and avoid penalties. But while the value proposition was clear, the challenge was making the business model work at scale.

The Numbers:

• 📅 Founded: 2016

• 💰 Total Funding: $21.6M

• 📈 Revenue Growth: Reached $3M ARR in 10 months

• 👥 User Base: 1M+ onboarded users, with 200k+ paying customers

Reasons for Failure: 

• Scale and Sustainability Issues: Cushion’s model worked, but not at the scale needed to sustain a venture-backed company. The startup needed a huge user base to cover operational costs and make the economics work. While they managed to onboard over a million users, converting them into long-term paying customers was another challenge. As the founder put it, “Despite bringing multiple new fintech products to market, we didn’t reach the scale needed to sustain the business.”

• The Challenge of Consumer-Facing Fintech: The last few years have been brutal for fintech startups, especially those targeting individual consumers. Companies like Synapse, CapWay, Tally, and Bench have also shut down, struggling with regulatory complexity, high acquisition costs, and shifting consumer behaviors.

• Trust Barriers in Fintech: Cushion wasn’t just competing with traditional banks—it also had to convince users to trust a third-party service with their financial data. That’s always a tough sell, and fintech adoption tends to be slower than in other industries. This trust gap made customer acquisition harder and likely limited growth.

• Business Model Constraints: Cushion operated in a space where margins were tight. The company made money by taking a percentage of refunded fees—which meant that in order to grow revenue, they needed to help users recover more money. But there’s a ceiling to that model, especially as banks tighten their policies and reduce refunds over time.

Why It Matters: 

• Consumer fintech is a tough space. Even a useful product isn’t enough—scaling profitably is the real challenge.

• Relying on refunds as a business model has limits. If the revenue stream depends on banks continuing to refund fees, it’s inherently unstable.

• The fintech landscape is shifting. More startups in this space are shutting down, showing just how difficult it is to build a lasting consumer fintech business.

Trend

Operator

Last week, OpenAI launched Operator, its first AI agent capable of interacting with the web on its own. Unlike ChatGPT, which primarily responds to text prompts, Operator can browse websites, fill out forms, and complete multi-step tasks—all without requiring a user to click a single button.

Why It Matters:

• AI is becoming more than just a chatbot. Operator represents a shift from AI as a passive tool to an active assistant that can take meaningful action on your behalf.

• Automating the mundane. Repetitive browser tasks—filling out forms, ordering groceries, booking flights—can now be handled by AI, potentially saving hours of manual work.

• A new battleground for AI companies. OpenAI isn’t the only player in this space. Google’s Project Mariner and Anthropic’s AI tools are also exploring AI-driven automation.

• Limited access for now. Operator is currently only available to GPT Pro users in the U.S. ($200/month plan), but OpenAI plans to expand access over time.

What Is Operator?

Operator is OpenAI’s first AI agent designed to go beyond answering questions. It can navigate the web, interact with sites, and complete tasks just like a human user would.

Unlike traditional AI assistants that rely on APIs to fetch data, Operator can function within a browser. This means it can work with any website, even if the site hasn’t built a custom integration for AI. This flexibility makes Operator far more adaptable than existing AI solutions, allowing it to complete a broad range of digital tasks.

The key behind Operator is a new model called the Computer-Using Agent (CUA). By combining GPT-4o’s vision capabilities with reinforcement learning, CUA enables Operator to see webpages (via screenshots), interpret what’s on the screen, and take actions like typing, clicking, and scrolling. It mimics how a human would interact with websites, making it effective in navigating dynamic and complex digital environments.

This ability means Operator doesn’t require websites to build custom AI-friendly APIs. Instead, it can function within existing website structures, significantly broadening its usability. 

Additionally, Operator incorporates reasoning and self-correction mechanisms. If it encounters an issue, it can analyze the problem and attempt to adjust its approach, reducing the need for user intervention. However, if it gets truly stuck or if it finds a CAPTCHA, it hands control back to the user.

What Can It Do?

The potential applications are broad, but here are a few ways people have already used Operator:

• Job Hunting on LinkedIn. Dan Mac instructed Operator to find him a job in Linkedin.

• Booking Flights and Scheduling Appointments. Instead of manually entering flight details and comparing options, Operator can navigate airline websites and make selections based on user preferences. It can also make an appointment in any website taking your availability into consideration.

• Testing: One user leveraged Operator to quickly verify their website’s login process—without writing a single test. This streamlined approach not only made testing simpler but also had the potential to save significant development time.

Help Me Improve Failory

That's all of this edition.

Cheers,

Nico