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:

• Inseco, a startup that made protein from flies, shut down — learn more below

• 32 things we’ve learned about building a startup that scales.

• Uber will offer gig work like AI data labeling to drivers while not on the road.

• Sizable Energy raised $8M to store clean energy under the sea.

• OpenAI launched Atlas, it’s first AI Browser — learn more below

Losing focus slows startups down. Today’s sponsor, TRMNL, helps you stay on track with an open source e-ink display that cuts out the noise.

No subscription, no ads, no alerts, only focus^AD

Most startups fail. There are many reasons for this, but one of them is losing focus.

So how do you keep your team on track? Meetings are OK, but they take a lot of time. Dashboards are OK, but you might measure the wrong things.

Introducing TRMNL, an open source e-ink display that helps you stay focused, without getting distracted.

No touch screen. No colors. No back light. Nothing to “play with.” Just information that’s important to your startup, on your desk or wall.

Monitor server analytics, web traffic, sales leads, git commits, and even personal data like your calendar, the weather, or the occasional The Office quote.

Get informed at a glance, then get back to work.

Stay focused. Get TRMNL.

This Week In Startups

🔗 Resources

The Ultimate Guide to the perfect SaaS pricing page.

32 things we’ve learned about building a startup that scales.

Do you have a demand problem?

Your pricing is (probably) broken.

📰 News

Amazon unveils AI smart glasses for its delivery drivers.

Netflix goes ‘all in’ on generative AI.

Uber will offer gig work like AI data labeling to drivers while not on the road.

Walmart partners with OpenAI so shoppers can buy things directly in ChatGPT.

💸 Fundraising

Tensormesh raises $4.5M to squeeze more inference out of AI server loads.

OpenEvidence, the ChatGPT for doctors, raises $200M at $6B valuation.

AI operating system startup UnifyApps raises $50M.

Sizable Energy raised $8M to store clean energy under the sea.

Fail(St)ory

Food From Flies

Inseco was supposed to be Africa’s insect protein success story. Instead, after seven years and a $5.3 million seed round, it shut down and sold off its assets.

A reminder that scaling biology isn’t just about science: it’s about execution in messy, real-world conditions.

What Was Inseco:

Inseco was born in Cape Town in 2017 from an ambitious idea: turn food waste into high-value protein. Founder Simon Hazell saw how much organic material (from wineries to food processors) was ending up in landfills. He figured black soldier fly larvae could eat that waste and become food themselves.

Inseco would collect organic byproducts, feed them to larvae, and process the grown insects into protein meal, oil, and fertilizer. The products, branded EntoMeal and EntoOil, targeted aquaculture, poultry, and pet food markets hungry for alternatives to fishmeal and soy.

At its peak, Inseco ran a 10,000-square-meter facility producing over 100 metric tons of insect meal and oil a month. Investors took note. In 2022, it raised one of South Africa’s largest-ever seed rounds: $5.3 million.

But insect farming isn’t software. You can’t “move fast and break things” when your product is alive. The larvae needed constant heat, humidity, and airflow—conditions that didn’t always cooperate.

As the company scaled, cracks started to show. Power became unreliable, capital scarcer, and patience thinner. The elegant model began to strain under real-world complexity.

By 2024, the optimism that once surrounded Inseco had faded. Operations paused, assets were sold, and the founder signed off quietly: “We got close, but ultimately ran out of time.”

The Numbers:

• 🐛 Founded: 2017

• 💰Funding: $5.3 million seed (2022)

• 🏭 Facility: 10,000 m² tech-enabled plant

• ⚙️ Production: 100+ metric tons/month

•  📉 Shutdown: 2024
  

Reasons for Failure: 

• Loadshedding crippled operations: Inseco’s production depended on stable power—something South Africa couldn’t guarantee. Loadshedding, the country’s rolling blackouts, cut electricity for hours at a time, throwing off the temperature and humidity the larvae needed to survive. “The recurring 4-hour power outages devastated operations and increased our energy costs fourfold,” Hazell said. 

• Overaggressive scaling: Inseco ramped up production before proving it could make money on each ton of insect protein. The factory was big, the costs were bigger, and the margins never caught up. Hazell later admitted they “chose an intersection that proved too aggressive”, meaning they bet that scale would fix their unit economics. It didn’t. They built a commercial-scale plant on startup-level proof.

• Weak hiring and slow pivots: When growth came fast, discipline slipped. Hazell admitted, “We made some bad hires. We scaled too quickly and pivoted too slowly.” The team spread itself thin between running a large facility and fixing its tech. Instead of doubling down on the higher-margin process they were developing, they kept chasing scale. Focus drifted, and in manufacturing, that’s fatal.

• Investor sentiment turned: The end of zero-interest-rate money hit hard across deep-tech and climate sectors. Even stronger insect startups in Europe were struggling to raise. Inseco, operating from a less mature market, couldn’t weather the drought.

Why It Matters: 

• Scale is not validation. Inseco hit industrial scale but not industrial margins. Sometimes, scaling a broken model just multiplies the burn.

• Reliability is credibility. Every blackout didn’t just stop production, it also shook investor faith. In hard-tech, operational slip-ups echo louder than financial ones.

Trend

GPT Atlas

Two days ago, OpenAI launched Atlas, an AI-powered browser built around ChatGPT. You’ve seen versions of this before: Perplexity’s Comet, The Browser Company’s Dia, Microsoft’s Copilot tab in Edge. 

Atlas just happens to be the cleanest execution yet: less friction, better context, and an agent that actually does work where you are. That’s enough to shake up the browser status quo.

Why It Matters:

• The browser is the new workspace. Atlas treats the web like one giant app. Instead of bouncing between tools, you can research, summarize, plan, and book things in one flow.

• Context finally sticks. Atlas remembers what you’ve been doing online (sites, chats, projects) and picks up right where you left off. No more hunting for that tab or file from last week.

• Agents get a real interface. Atlas gives ChatGPT a native playground. It can read a page, click links, and take action, which makes it feel more like a coworker than a chatbot.

What is Atlas:

Atlas is OpenAI’s AI-native browser, designed to make ChatGPT feel integrated instead of bolted on.

You open a page, and ChatGPT already knows where you are. You can ask it to summarize an article, compare it to what you read last week, or draft an email response based on the content in front of you. All without switching tabs or copy-pasting.

Two features matter most: memory and agent mode.

• Memory lets ChatGPT recall your browsing history and past tasks, so you can ask, “Find all the job postings I looked at last week and summarize trends.” 

• Agent mode lets it take action inside the browser: researching, booking meetings, filling forms.

The Trend:

We’ve known for a while that the browser was the next battleground of the tech world. Now it’s official. With the launch of ChatGPT Atlas by OpenAI, the fight for who controls how you access and use the web has stepped into a new phase.

The trend is simple: the browser is turning into the home for AI agents.

Instead of adding AI as a feature, companies are rebuilding the browser around it: where your assistant doesn’t just answer questions, it reads, acts, and remembers directly on the page.

The Idea:

Big companies will fight over browsers. You don’t need to. The smarter move is to build around the layer Atlas creates.

A few simple ideas worth exploring:

• AI safety shell. A tool that monitors what the browser-agent sees and does. Think analytics + alerting rather than more chat UI. Given that Atlas already faces prompt-injection risk and security warnings, this is very much needed.

• Agent-aware extensions. Atlas is Chromium-based, which means you can already build normal Chrome extensions. They can’t talk directly to the ChatGPT agent yet, but that’s coming. The moment OpenAI exposes APIs for agent communication, extensions that can feed, filter, or visualize agent behavior will explode.

Help Me Improve Failory

That's all for today’s edition.

Cheers,

Nico