Hey - It’s Nico.

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

If you only have one, here are the 3 most important things:

• ClimateAi, a startup using IA to predict weather, has shut down — learn more below.

• This post will save you tokens

• A new category of startups is building layers that let SaaS users create their own features — learn more below

This Week In Startups

🔗 Resources

How to make people care about your startup

This post will save you tokens

How the GTM playbook has changed in 2026

Dear VCs, Just Give Up

📰 News

OpenAI launches a safer ChatGPT for teens

Perplexity’s free AI offer left it with millions more users in India

SpaceX officially closes its Cursor acquisition

OpenAI introduces ‘Ultrafast,’ a new mode that makes GPT-5.6 Sol work at 14x the speed

💸 Fundraising

Neros Technologies, a defense startup building autonomous and interceptor drones, raises ⁠$250M Series C funding

Skan AI, an enterprise AI platform for mapping and automating business workflows, raises ⁠$63M funding

Axle, an AI-native insurance data clearinghouse, raises ⁠$17.5M Series A funding

Alloy Robotics, an AI platform for debugging and analyzing robot fleets, raises ⁠$8M to help engineers debug robot fleets

Fail(St)ory

AI Meteorologist

ClimateAi sold weather and climate forecasts to companies whose crops, supply chains, and commodity costs depended on what happened months or years ahead.

It shut down on two weeks ago, after raising about $38 million. The challenge was turning better forecasts into recurring software spend and real operating decisions.

What Was ClimateAI:

ClimateAi sold weather forecasts built for business decisions. Its main product, ClimateLens, combined weather data, satellite observations, climate models, and machine learning, with most of its work focused on food and agriculture.

A normal weather forecast tells you what might happen. ClimateAi pushed further into what that could mean for crop yields, supply, inventory, and prices over periods ranging from a few weeks to several years.

That could change when a company harvested, how much inventory it carried, where it sourced from, or when it bought a commodity. 

Some of those decisions were worth a lot. ClimateAi said a wet-harvest warning helped one customer avoid losses worth hundreds of thousands to millions of dollars. Another customer used its crop outlook as part of a trading strategy that the company estimated produced $2 million to $6 million.

The product fit best when weather already sat close to an existing business decision. Procurement teams, for example, already managed supply, contracts, and inventory, so a forecast could feed directly into work they were doing anyway. 

By late 2024, ClimateAi said it had worked with more than 56 partners across 40-plus crops and more than 60 countries.

Customers used it in different ways. Some paid for recurring access, while others hired ClimateAi for a specific analysis and came back when another planning problem came up. The company wanted the economics of software, with the same forecasting system serving many customers at low added cost.

The forecast still had to land somewhere useful inside the customer’s workflow. ClimateLens 2.0, released in early 2026, pulled in planting records, contracts, and delivery schedules so ClimateAi could send alerts tied to decisions the customer already had to make.

That was the basic bet behind the company: give businesses enough warning to change what they buy, plant, move, or protect before weather changes the outcome. ClimateAi eventually said it served roughly a quarter of the world’s top 200 food and beverage companies.

The Numbers:

• 💰 Raised: $38M

• 🌍 Countries covered: 60+ by late 2024

• 🌾 Crops: 40+ across 56+ partners

• 📈 ARR growth: 5x before April 2023

• 👥 Staff: ~52 in August 2025

• 🛑 Wound down: August 7, 2026

Reasons for Failure: 

• No clear budget owner: ClimateAi said customers often lacked a standing budget for climate adaptation. Procurement could pay when forecasts affected sourcing or inventory, but broader resilience work crossed several departments, which made sales and renewals harder to standardize.

• The software required operating changes: A forecast sitting in a dashboard had limited value. Customers had to change contracts, inventory, planting dates, hedges, or capital plans early enough for the forecast to matter, often across teams with existing approval cycles.

• ROI arrived unevenly: ClimateAi could point to cases where one forecast saved or generated millions. Those wins depended on specific weather events, crops, and decisions, so a customer might see a large payoff one year and little visible financial impact the next.

• Repeatable scale was still unresolved: In late 2024, the company expected profitability in 2025 and discussed reaching roughly 200 partners before another funding round. It was also expanding across yield tools, APIs, demand forecasting, government work, and workflow products, suggesting it was still working out which buyer and use case could support the company at scale.

Why It Matters: 

• Budget ownership shapes adoption: products that affect several departments need a clear person who controls the money and the decision.

• Risk products need value between big events: one avoided loss can justify years of fees, but customers still review the bill every year.

• Forecasts become more useful when they sit inside the decision: linking weather data to contracts, schedules, and inventory reduces the work required to act on it.

Trend

Extensible SaaS

Lately, I’ve been noticing a new startup category with one pitch: add a generative layer to an existing SaaS product so every customer can build the features they are missing. Vendo calls them “user-generated features.”

Say your sales team uses a CRM but needs a page that groups renewals by month, shows the account owner and includes a button for sending a Slack reminder. You describe that page inside the CRM. The layer builds it using the records already there, then saves it to your account so the team can keep using it.

From the user’s point of view, the SaaS has learned a new feature for their workflow. The new page looks and behaves like part of the product, follows the same access rules and can be shared with colleagues. One product can now grow into a slightly different version for each customer.

Why it Matters

• A feature request can become something the customer uses. Teams can build the niche dashboards, forms and workflows they need without waiting for them to reach a product roadmap.

• The work stays connected to the software where the data already lives. Users can avoid exporting records to spreadsheets, buying another tool or asking a developer to rebuild the same context elsewhere.

• Customer-made features create better evidence of demand. A vendor can see which additions people keep using, promote the best ones to the main product and let other customers reuse them.

The Signals

Vendo offers the clearest version of this experience. A user can ask for a dashboard, workflow, new action or persistent micro-app inside an existing SaaS product. The result uses the host product’s data and visual components, and it can connect to tools such as Slack or Salesforce when the vendor allows it. Sensitive actions still require approval.

OnClaw uses the line “Every user gets their own version of your app.” A user opens a command bar, describes a missing feature and receives a new component inside an area the vendor has made customizable. The feature stays with the product. Developers can later inspect and edit what was generated.

Gigacatalyst has the strongest disclosed production example through UpKeep Studio. Maintenance teams can ask its Nova assistant to build inspection forms, shift-handoff tools, compliance workflows and specialized dashboards from their existing UpKeep records. Users can review the result, revise it and publish it to colleagues. Gigacatalyst claims more than 800 generated features in six weeks and 70% day-30 retention, though those figures are unaudited.

Several other startups are selling variations of the same outcome. Legato lets business users create apps, workflows, automations and agents inside B2B platforms. Vezel offers generated dashboards, reports and internal tools, while Fork gives each customer a separate branch of the original application. Fork also lets vendors bring popular customer-made features back into the main product.

The Trend

People have customized software through spreadsheets, scripts and no-code tools for decades. These startups bring that behavior inside the SaaS product itself. The language model turns a plain-English request into a feature, and the layer gives that feature access to the right data, a place in the interface and a permanent home in the customer’s account.

I think this is becoming a real software category. The startups are selling the machinery that turns one fixed product into many customer-shaped versions, while the original vendor continues to control the data, permissions and core services.

The evidence is still early. UpKeep is the only named production deployment in this group. The category will matter if users keep returning to the features they generate and trust them for everyday work.

That is where this new category could matter. SaaS companies would keep building the core product, and customers would use these layers to create the smaller dashboards, workflows and tools that make it fit the way they actually work.

Help Me Improve Failory

That's all for today’s edition.

Cheers,

Nico