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:

• SWAP Food, a startup making plant-based chicken, shut down — learn more below

• Using context-engineering to 5x conversion and 2x activation

• Founders are using AI assets to market non-exist products and validate ideas — learn all about this strategy below

A huge thanks to today’s sponsor, HubSpot. Download their free AI guide to supercharge your workday.

100 ChatGPT Prompts to Revolutionize Your Day ^AD

Supercharge your productivity with HubSpot's comprehensive guide. This free resource is your fast track to AI mastery:

• Industry-Specific Use Cases: 15+ real-world applications across various sectors

• Productivity Guide: 21 best practices to 10x your efficiency with AI

• Prompt Powerhouse: 100+ ready-to-use prompts for immediate implementation

• Challenge Buster: Overcome common AI hurdles with expert strategies

Plus, in-depth sections on email composition, content creation, customer support, and data analysis.

Find Your Guide Here

This Week In Startups

🔗 Resources

How to pivot well

What AI-native GTM looks like at public company scale

Using context-engineering to 5x conversion and 2x activation

Your unfair advantage in the age of AI

📰 News

OpenAI unveils its first custom chip

Anthropic released Claude Tag

New data suggest engineering jobs are the most resilient

Anthropic’s Claude is winning over paid consumers, a market owned by ChatGPT

💸 Fundraising

Arcade.dev, a secure action layer for production AI agents, raises $60M⁠

Convey, an AI teammate platform for enterprise operations, raises $38M ⁠

Verse, an energy infrastructure platform for AI data centers, raises $54M ⁠

HyperLight, a photonics chip startup for AI infrastructure, raises $80M

Fail(St)ory

Plant Chicken

SWAP Food raised over €100M to make plant-based chicken look and cook like a real chicken breast.

By 2026, the Paris-based company was in judicial liquidation. Its factory in Alsace had been built for industrial scale, but the installed line ran at under 10% capacity. 

This is a story of what happens when you bet on industrial scale before the market is ready to absorb it.

What Was SWAP Food:

Most plant-based meat companies started with the easy shapes: burgers, nuggets, sausages, mince. SWAP went after the harder format. 

It wanted to make a plant-based chicken breast that looked like a real filet, cooked like one, and could sit in the middle of a plate without looking like a compromise.

SWAP’s technology was called Umisation. In simple terms, it was a way to turn plant protein into long, meat-like fibers without the usual heavy extrusion process. The company wanted whole-cut texture from soy and pea protein, with a shorter ingredient list than many alt-meat products.

Its main product, SWAP Chicken, was a thick plant-based filet with 19g of protein per 90g serving. It used eight ingredients and avoided GMOs, added gluten, cholesterol, artificial flavors, colorants, and texturizers.

The pitch to restaurants was simple: use it like chicken. Put it in a sandwich, bowl, salad, or plated dish without rebuilding the menu. For chefs, that was the appeal. They could add a plant-based option without making it feel like a separate vegan menu item.

The 2024 rebrand from Umiami to SWAP made the proposition clearer. The company was selling a swap from animal protein to plant protein. The target was the mainstream diner who still ate meat but wanted to cut back, not the tiny group already committed to vegan food.

That made SWAP different from many alt-meat startups. It was trying to solve the two complaints that kept hurting the category: weak texture and products that felt too processed. A chicken-style filet with a clean label gave foodservice buyers a better story to tell.

So, what do you do when you’re still proving a new food technology and the market has not fully shown up yet? You build a giant factory, apparently.

SWAP made a huge bet on a production site in Alsace. The plant was built for 7,500 tons of annual capacity, with plans to reach 20,000 tons. By the end, the installed line was reportedly running at under 10% capacity.

The Numbers:

• 💰 Funding: €100M+ 

• 🏭 Factory: 7,500 tons of annual capacity at launch, with plans to reach 20,000 tons.

• 📉 Utilization: Under 10% capacity.

• 🧾 Debt: Around €16.5M 

• 🔥 Burn: Around €1M per month after cost cuts.

• 👥 Team: 66–70 employees

Reasons for Failure: 

• The factory made small demand look ridiculous: A pilot line can live with messy early sales. SWAP had already built the expensive version: rent, debt, salaries, utilities, maintenance, and a lot of stainless steel waiting for orders. When utilization sits under 10%, every slow week burns money in the most physical way possible.

• The product was too premium for a cheap-protein fight: SWAP Chicken sold at around €20/kg. That can work in a nice restaurant where the menu needs one strong plant-based option. It is much harder when the benchmark is ordinary chicken, one of the most efficient proteins on the planet. 

• The balance sheet left no patience for a slow category: SWAP was carrying debt, heavy losses, and an underused factory while the plant-based meat market grew slower than expected. Management cut fixed costs by 50%, but losses were still around €1M per month. The company needed fresh financing to keep going. Investors had become much less willing to fund companies that needed industrial-scale capital before proving industrial-scale demand.

Why It Matters: 

• Industrial scale turns slow adoption into an expensive fixed-cost problem very quickly.

• “Drop-in replacement” still has to win on price, format, storage, prep, distribution, and menu economics.

• A category can be growing and still be too small for one company’s factory, debt, and burn.

Trend

Synthetic Validation

This week I saw a tweet that caught my attention because it was a perfect example of a marketing strategy I am seeing more and more of. The tweet was about BYLT, a Kickstarter campaign for miniature construction kits, and how an AI-generated promo video helped turn a stalled campaign into real pledge demand.

People have been calling this strategy a bunch of things: AI smoke tests, synthetic MVPs, testing the fantasy first. I’ll use Synthetic Validation, mostly because it gets closer to what is actually happening: founders are using AI-generated assets to test demand before the real product is fully built.

Why it Matters

• The launch asset is becoming the first product test. Founders can now test whether people want the story, use case, aesthetic, or emotional payoff before spending months on production. 

• Realism changes the trust equation. A rough mockup tells users they are looking at a concept. A polished AI video can make an unfinished product feel like a finished one, which pulls more attention and also raises the bar for disclosure.

• The asset can backfire once people realize it is AI. BYLT’s video helped drive attention and pledges, but also triggered backlash over what was real, generated, or actually finished.

The BYLT Story

BYLT: Real Construction, Tiny Scale was a Kickstarter campaign for miniature home-building kits. The promise was oddly specific and very internet-friendly: build miniature homes using real construction methods, including blueprints, a concrete foundation, framing, electrical, siding, and finishing.

After launch, the campaign became stagnant for around 15 days. Then Ethan, the founder, posted a sleek AI-generated launch video. It reportedly crossed 5M views and helped bring in more than $30K in pledges for a product that was still basically non-existent.

Of course, this was not all good. The video also got a lot of backlash once people realized it was AI-generated. Some called the campaign fake, others questioned what was real, what was rendered, and how much of the product had actually been built. Kickstarter later suspended the campaign.

But the key thing is that we now know there is potential in Ethan’s idea. Before the video, miniature construction kits could have sounded like an insane niche product. After the video, it looked like something that might actually have PMF.

And Ethan tested that with basically zero production budget. That is the power of synthetic validation: you can test the fantasy before building the product. Because the video looked like a finished product launch, rather than the classic Kickstarter video of a founder talking into a camera and pitching a concept, people got more excited and were more willing to pledge.

Other Signals

• Wonders of the First, a collectible card game, raised more than $1.4M on Kickstarter while using a lot of AI art. The campaign showed the same dynamic at a bigger scale: AI assets can make a project look polished fast, but they also make backers inspect the whole thing more aggressively.

• Maket used AI-generated influencers to test creative concepts and marketing hooks before spending more on campaigns. That is synthetic validation applied to messaging: cheap fake people, real customer reactions.

• Pieter Levels had already described the indie playbook. In 2023, he suggested validating ideas with a mockup, daily TikToks, Linktree, Stripe preorders, and repetition across many ideas. AI makes the mockup, video, product shots, copy, and variants cheaper to produce.

• A 2025 Alibaba-linked paper described a “sell it before you make it” system where merchants generate photorealistic fashion items from text, show them to customers, and produce after enough orders come in. The paper reported over 13% relative improvements in CTR and conversion versus human-designed counterparts.

• Kickstarter is now requiring creators to disclose when they use AI-generated images, text, or other outputs, and explain what was generated versus original. That is a signal by itself. Platforms do not write policies for edge cases unless enough projects are already doing it, and synthetic campaign assets are now common enough to need their own trust rules.

The Trend

Synthetic validation is becoming a normal part of startup and creator workflows. The founder no longer has to start with the product.

They can start with the promise, package it into a believable asset, distribute it, and measure whether anyone reacts strongly enough to pay, pledge, or preorder.

That makes the next question less about whether you can make something look real, and more about how clearly you tell people what they are actually buying.

Help Me Improve Failory

That's all for today’s edition.

Cheers,

Nico