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:

• Blip, a startup that promised to deliver clothes in 30 minutes, has shut down— learn why below.

• Inside the SaaS Efficency Gap.

• Lovable becomes a unicorn just 8 months after launch.

• Thinking Machines Lab raises $2 billion.

• Elon Musk launched Grok 4 last week and it’s actually not bad — learn more below.

This Week In Startups

🔗 Resources

The Founder’s Guide to Building a V1 of Customer Success.

Inside the SaaS Efficency Gap.

Deep Dive: The Path to PMF

The Viral Loop Illusion

📰 News

Lovable becomes a unicorn just 8 months after launch.

Mistral’s Le Chat chatbot gets a productivity push with new ‘deep research’ mode.

Uber makes multi-million-dollar investment to build a robotaxi service.

Google rolls out AI-powered business-calling feature.

💸 Fundraising

Crypto messaging startup Ephemera raises $20 million.

Drone startup Xtend raises $30M more in funding.

Mira Murati’s AI startup Thinking Machines Lab raises $2 billion.

Spatial computing startup Augmodo raises $37.5M.

Fail(St)ory

Urgent Clothing

About one year ago, a startup launched with one simple idea: get clothes delivered to your door in under 30 minutes. 

It was called Blip, and it was supposed to be “Zepto for fashion.” The logic made sense — if instant delivery worked for groceries, medicine, and even electronics, why not apparel? 

Quick commerce was winning in every direction. So Blip tried to bring that same playbook into fashion. Nine months later, it was gone.

What Was Blip:

Blip launched in Bengaluru in late 2024. The idea was simple and punchy. You open an app, pick something to wear, and it shows up at your door in under 30 minutes. No endless scrolling, no waiting two days for delivery, no wondering if it’ll arrive before your weekend plans.

The app went live in October in select areas of the city, offering over 25,000 SKUs from more than 10 brands.

To make that happen, Blip didn’t build warehouses or hold inventory. Instead, it partnered with local fashion stores and used them as mini-fulfillment hubs. Orders were routed to the nearest participating shop, packed quickly, and handed off to a delivery driver.

On paper, it was clever. By skipping warehouses, Blip could launch in new cities without sinking money into infrastructure. All they needed was a network of stores and a logistics layer on top. Fast to scale, capital-light, and flexible.

But turning retail stores into delivery centers isn’t as easy as flipping a switch. The tech had to work. The inventory had to sync. The retailers had to fully commit. And in fashion — where sizing, returns, and trends add layers of friction — none of that is guaranteed.

Blip was trying to take a model that worked in groceries and meds, and bend it to fit fashion. For a minute, it looked like they might pull it off. But the cracks started to show pretty quickly.

The Numbers:

• 📅 Founded in 2024.

• 🛍️ Launched in October with 25,000 SKUs from 10+ brands.

• 🏙️ Operated only in select parts of Bengaluru.

• ⏳ Promised 30-minute delivery.

• ❌ Shut down less than a year later.

Reasons for Failure: 

• Not enough capital to compete: Blip was trying to play the quick commerce game without VC money. That works in some niches, but not when you’re running a logistics-heavy operation and trying to scale. The founder admitted that their limited funding made it hard to execute a proper go-to-market strategy. Without cash to expand, market, or subsidize early orders, they couldn’t build momentum beyond the pilot city.

• Retail partnerships were hard to scale: The model depended on local stores acting like fulfillment centers — syncing inventory, packing fast, and hitting tight delivery windows. But these weren’t tech-savvy warehouses — they were regular fashion shops. Convincing retailers to adopt a new system and reliably execute on-demand logistics turned out to be a lot slower and messier than expected.

• Fashion is a bad fit for instant delivery: Groceries and essentials are predictable, frequent, and standardized. Fashion is none of those things. It’s low-frequency, size-sensitive, and highly personal. That makes it harder to predict demand, harder to automate fulfillment, and riskier in terms of returns. Blip tried to move fast in a category that doesn’t lend itself to speed.

• No moat, no habit, no repeat usage: Blip’s value prop was urgency — but how often do people urgently need a new shirt? The use case was too narrow to build repeat behavior. And without brand loyalty, habit loops, or unique supply, there wasn’t much to keep users coming back. Once the novelty wore off, there wasn’t a strong reason to open the app again.

Why It Matters: 

• Fashion quick commerce might sound like a natural next step — but this case shows that not all categories are built for speed.

• The playbook that works for groceries doesn’t translate neatly into other verticals. Founders need to deeply understand the usage patterns and economics of their specific market.

Trend

Grok 4

Elon Musk launched Grok 4 last week, and for the first time, it looks like xAI might actually have something worth paying attention to. After lagging behind GPT, Claude, and Gemini for what felt like forever, Grok is now showing up in benchmark leaderboards… and not at the bottom.

But, of course, it wouldn’t be an Elon project without some chaos — within days of launch, Grok was calling itself “MechaHitler” and quoting its founder’s tweets in political discussions.

Let’s unpack what’s new, what’s weird, and whether any of this actually matters.

Why It Matters:

• Grok 4 might finally be competitive with GPT and Claude — which wasn’t the case before.

• xAI’s new “Heavy” mode shows interesting ideas about how to scale agentic reasoning.

• The controversies surrounding Grok raise serious questions about how personal bias could sneak into AI systems.

What’s new with Grok 4

At its core, Grok 4 is a major leap from its predecessor. On paper, it’s performing like a serious contender.

On Humanity’s Last Exam — a notoriously tough reasoning benchmark — Grok 4 scored 25.4%. That’s a solid result, but the real eye-opener came from its “Heavy” version, which hit 44.4%. That’s nearly double what GPT and Gemini managed.

xAI claims Grok 4 is now the state of the art in many reasoning categories. It also hit a top score on ARC-AGI-2, a tough pattern recognition benchmark, and handles up to 256,000 tokens of text and images. In practical terms: it can parse complex docs, write code, solve advanced math, and break down charts — the usual high-end LLM tasks.

Alongside the model update, xAI launched a new $300/month “Heavy” tier. This version dynamically spawns multiple agents to collaborate on a task. Conceptually, it feels like a blend between OpenAI’s Deep Research mode and some of the agentic features we’ve seen in Anthropic’s Claude Opus.

And Then… The Weird Stuff

No Grok launch would be complete without some drama.

Just days after launch, Grok introduced itself as “MechaHitler” to some users. That’s not a typo. Apparently, it was referencing a character from an old video game, but still — probably not the branding anyone was aiming for.

Then came reports that Grok’s responses to political or controversial questions seemed curiously aligned with Elon Musk’s own tweets. While Grok is trained on internet-scale data, it appears that in certain edge cases, it leans heavily on its founder’s views.

This kind of behavior raises an interesting question: Can a model built by a personality as publicly vocal as Elon Musk remain neutral? And even if it can, will people trust that it does?

So, Does This Actually Matter?

Here’s the real talk.

Yes, Grok 4 looks much better than previous versions. And yes, xAI clearly has smart researchers working on innovative stuff. But is Grok 4 the GPT-killer that Elon is hyping it up to be? Probably not — at least not yet.

OpenAI, Google and Antrophic still dominate the developer ecosystem. They have better tooling, broader integrations, and more mature platforms. Simply having strong benchmarks isn’t enough. Most people aren’t going to switch models unless the change is either dramatically better or dramatically cheaper. And Grok, right now, is neither.

That said, this launch does put xAI on the map in a more serious way. If nothing else, Grok 4 shows that xAI is not just a meme machine — they’re genuinely playing the same game as OpenAI, Anthropic, and Google.

Will Grok overtake them? Probably not soon. But could it become a fourth strong contender that pushes innovation forward and adds a new flavor to the market?

That’s a lot more likely.

Help Me Improve Failory

That's all of this edition.

Cheers,

Nico