Hey — It’s Nico.

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

If you only have one, here are the 5 most important things:

• WhyHive, the Canva for Data Analytics, has shut down — learn why below.

• Build your MVP using Y Combinator's Proven Framework.

• Runway releases an impressive new video-generating AI model.

• Space solar startup Aetherflux raises $50M to launch first space demo.

• Amazon launched Nova Act, a developer toolkit for building autonomous AI agents — learn more below.

Let’s get into it.

This Week In Startups

🔗 Resources

10 startups to watch from Y Combinator’s W25 Demo Day.

Unlocking profitable B2B growth through gen AI.

Build your MVP using Y Combinator's Proven Framework.

How Lovable got to $40M ARR in just 4 months.

📰 News

Apple loses $250B market value as tariffs tank tech stocks.

ChatGPT users have generated over 700M images since last week.

Amazon reportedly submits last-minute bid to acquire TikTok.

Runway releases an impressive new video-generating AI model.

💸 Fundraising

Thatch raises $40M to give employees more control of their health care choices.

Runway, known for its video-generating AI models, raises $308M.

Fintech Plaid raises $575M at a $6.1B valuation.

Space solar startup Aetherflux raises $50M to launch first space demo.

Fail(St)ory

Canva for Analytics

A few days ago, WhyHive, a Melbourne-based startup that aimed to make data analysis as accessible as Canva made design, announced it’s shutting down.

Just 8 months after the pre-seed round, the team has decided to call it quits.

What Was WhyHive:

WhyHive set out to “democratize data analysis”. The idea was to give anyone, regardless of technical skill, the power to analyze complex data sets with a few clicks. Think surveys, sales spreadsheets, product reviews—all uploaded and processed using WhyHive’s custom-built chart engine combined with AI.

Their pitch was simple: what Canva did for graphic design, they wanted to do for data.

@marytheanalyst

How to use AI to analyze data 🤖 #sponsored #whyhive #dataanalyst #dataanalytics #dataanalysis #data #analyst #ai #artificialintelligence ... See more

WhyHive offered free plans to grassroots projects and steep discounts for nonprofits. The commercial plans ranged from $29 to $199/month, with custom deals for larger organizations.

But in a world already saturated with data analytics tools—and where giants like Salesforce and HubSpot are rapidly building out their own solutions—being useful wasn’t enough.

The Numbers:

• 📅 Founded as a social enterprise in 2018.

• 🔄 Pivoted into product development in early 2023.

• 💰 Raised $600,000 in pre-seed funding.

• ❤️ Discounted or free plans for nonprofits and impact-driven orgs.

• ❌ Shut down in 2025, eight months after funding round.

Reasons for Failure: 

• Crowded Market: WhyHive entered a space full of heavyweights. From traditional BI platforms to all-in-one CRM systems expanding their analytics capabilities, competition was fierce—and often bundled into tools businesses already paid for. Even with a strong product, cutting through the noise proved difficult.

• Founder Burnout: After years of pushing, testing, tweaking, and trying new strategies, the emotional and physical toll caught up with the founders. As cofounder T. Guthrie put it: “Over time, the toll of that effort built up, and both Matt and I ended up pretty burnt out. For me, that also came with some health issues.” Startups often talk about resilience, but there’s a point where even the most committed team needs to stop. 

• Not Enough Growth Post-Funding: WhyHive officially started product development with investors in early 2023 and announced their $600K raise publicly in August. But despite the excitement and clear user value, the platform couldn’t grow fast enough to justify long-term sustainability. As cofounder Matt Cohen wrote: “Ultimately, however, we needed more revenue and growth to sustain this project long-term”

• Hard to Nail a GTM Strategy: WhyHive’s founders mentioned they tested different go-to-market strategies, pricing models, and feature sets—but couldn’t find the combination that unlocked real traction. In crowded markets, distribution often matters more than the product itself. Without a repeatable, scalable way to acquire and retain users, even a strong product can struggle to grow. WhyHive had clear value, but not enough people were seeing it—and that’s a tough problem to solve under time pressure.

Why It Matters: 

• When bigger platforms bundle your core feature, it becomes incredibly hard to compete.

• Burnout doesn’t always come after failure, sometimes it causes it. The emotional and physical cost of pushing too long can end a startup early.

• WhyHive shows how hard it is to go from consultancy to product. Different skill sets, different rhythms, different risks.

Trend

Amazon’s Nova Act

Last week, Amazon quietly took a big step into the AI agent race: it launched Nova Act, a developer toolkit for building autonomous AI agents that can actually complete tasks inside a browser.

Why It Matters:

• Developer-first approach: This isn’t a product for consumers. It’s built specifically for developers who want to create real-world products using AI agents. That makes it a foundational tool rather than a finished solution.

• Serious improvements in reliability: Amazon claims Nova Act performs over 90% on internal benchmarks that typically trip up other models—especially browser-based tasks with multiple steps. Most agents struggle to even hit 60%.

• Open-source, sort of: The SDK is open-source under Apache 2.0, but it only works with Amazon’s own Nova model. So while the tooling is open, it’s still tightly tied to Amazon’s ecosystem.

• Free for now: Nova Act is currently available as a research preview, with no announced pricing for production use. That makes it an easy entry point to start experimenting with agent-based workflows.

Amazon Joins The AI Race

For a while, it looked like Amazon was playing catch-up in the AI arms race.

OpenAI, Google, and Anthropic were rolling out flashy models and integrations. Amazon, on the other hand, was relatively quiet—until late 2024, when it launched its own foundation model family, Amazon Nova, capable of generating text, images, and even video.

That was followed by a revamped Alexa, now partially powered by Anthropic’s Claude models, signaling Amazon’s growing interest in building AI products—not just infrastructure.

With Nova Act, Amazon is finally putting something on the table that feels like a strategic swing. They’re not just releasing another model. They’re building a stack: models, infrastructure, and now, tooling for autonomous agents.

And they’re going after a specific pain point—reliability.

AI agents have been around for a while. But ask anyone who’s tried to build something serious with them: they break. Especially when you move beyond toy tasks. They’re great at navigating a website in theory, but in practice? Dropdowns, pop-ups, dynamic content—they fail more often than not.

That’s what Amazon says Nova Act is designed to fix.

Instead of trying to build one all-powerful agent, Nova Act takes a building block approach—letting developers compose small, reliable steps into workflows that actually work. In internal tests, Amazon says it’s hitting over 90% success rates on tasks where other agents get stuck.

So, What Can You Build With This?

Anything that requires automating a digital task inside a browser.

Think:

• Tools that fill out online forms based on user input.

• Agents that monitor and act on dashboards or admin panels.

• AI operators that complete repeatable workflows—things your ops team does 50 times a day.

It’s not hard to imagine startups building lightweight “copilots” for B2B tools, automating internal workflows, or creating AI-powered integrations with third-party web apps—without relying on APIs that may or may not exist.

This Amazon employee used the SDK to build and Agent that automatically an agent that automatically gathers and organizes comprehensive information on top-rated PlayStation 5 games, including prices, descriptions, and user reviews:

Of course, there’s still a lot to prove.

Nova Act works with Amazon’s own Nova model. That limits flexibility. It’s also unclear how it’ll perform in the wild—where browser environments are messier and less predictable than in controlled tests. And Amazon hasn’t said anything about long-term pricing, which could make or break its adoption.

Nevertheless, this announcement shows that Amazon is going all in on the AI race—not just with models, but with infrastructure and tooling that others can build on.

Help Me Improve Failory

That's all of this edition.

Cheers,

Nico