- Failory
- Posts
- The Vibe Design OG
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:
Visily, a startup building an AI design tool, shut down — learn why below
AI is decompiling all games and mixing them together — learn what this means for the future of softare below
Also, this week’s sponsor is AcceleratorCON, a startup event happening in NYC on October 23 for founders, investors, and accelerators.
A startup event actually built for connections AD
Most startup events are long on noise and short on meaningful interactions.
AcceleratorCON, happening on October 23 in New York City, is built differently. It brings together founders, investors, accelerators, and ecosystem partners for a day centered around live pitches, curated meetings, speed networking, and real opportunities to connect face-to-face.
If you’re a founder, it’s a chance to get in the room with investors, accelerators, and operators who can help move your company forward. If you’re an investor or accelerator, it’s a more intentional way to discover companies and meet builders.
Failory readers can get 25% off tickets with code FAILORY25.
This Week In Startups
🔗 Resources
Was this Anthropic's best product decision?
What PostHog learned from building multiplayer AI
Forum Ventures invests $100K in your MVP, then helps you scale. You get a partner who works with you 1:1 every week, helps you close dozens of customers, and introduces you to investors.*
📰 News
Anthropic launches Claude Haiku 5.5
SAP agrees to acquire TechWolf, an AI startup that maps employees’ work and skills
Google launches Playground for building games with AI
Google opens SynthID Detector to everyone, letting users check images, video and audio for supported AI watermarks
💸 Fundraising
Vinci, a startup that uses AI to simulate hardware designs, raised $250M in Series B funding
Stuut, a startup that automates invoice collections with AI, raised $52.5M in Series B funding
Volantis, a startup that develops light-based connections between AI chips and memory, raised $88M in Series A funding
Ampersand, a startup that connects AI agents to enterprise software, raised $15M in Series A funding
Fail(St)ory
Early to AI Design
Visily launched in 2022 with an AI design tool for people who wanted to create software interfaces without becoming designers. It entered the market years before Figma Make, V0, and Google Stitch.
On October 3, it announced it would shut down on October 31, saying its market had changed substantially. Its position had narrowed as design platforms added AI and app builders made working software a starting point for prototyping.
What Was Visily:
Visily was an AI wireframing and prototyping tool built around a Figma-style canvas. It generated web and mobile interfaces from prompts, screenshots, or sketches, then let users edit the output as individual design objects and connect screens into interactive prototypes.
At its April 2022 launch, screenshot conversion was a central feature. An existing interface could become an editable layout, combined with ready-made components and templates. It began rolling out text-to-design in December 2023.
Visily had a head start on several products that would later compete for the same work. v0 arrived in 2023, followed by Figma Make and Google Stitch in 2025. The opportunity was to turn that early start into a design workflow customers would keep using as their alternatives improved.
The pitch was UI design for non-designers: product managers, founders, and engineers who needed to explore an interface before handing it to a design or development team. Its differentiation came from combining AI generation with simple visual editing and collaboration in one place.
It was useful for early wireframes, where getting a concept onto the screen mattered more than perfect spacing. More demanding designs still needed manual refinement. Visily's strongest use case was the early design stage, with enough control to keep changing the idea.
The company spent its first couple of years free, deliberately studying usage before introducing paywalls. In 2024, it began selling subscriptions per editor. Pro cost $14 a month, with more AI credits, unlimited boards, and export features.
During 2025, Visily expanded the work users could do after generating the first screen. Its 4.0 release combined screenshots, text, and selected objects in a prompt, while keeping context across a conversation. Users could ask for changes to an existing design and continue editing on the canvas.
That effort continued into 2026. Design Instructions let teams preserve logos, themes, and other details across generated designs. Deep Design used Gemini 3 to produce more detailed interfaces. These additions addressed the work of turning generated screens into a consistent product design.
Then Visily added Figma asset imports and, in March, React, Vue, and HTML code generation. Teams could bring their existing designs in and take front-end code out. Visily was expanding beyond initial mockups as competitors connected more of the process from idea to application.
The Numbers:
🚀 Public Launch: April 2022
💰 Reported Seed Round: $3M in 2024
💳 Paid Plans Introduced: 2024
🧾 Pro Price: $14/editor/month
🛑 Scheduled Shutdown: October 31, 2026
Reasons for Failure:
Design platforms reached Visily's target users: Figma Make brought generated interfaces into editable design layers, while Google Stitch offered an AI design canvas. Figma said nearly 60% of Make files created in 2025 came from non-designers. For teams already keeping their design assets in Figma, those capabilities reduced the need to move early work into Visily. Its easier entry point became harder to sell as a separate subscription.
App builders made the mockup stage optional: Vercel's v0 and Lovable combined interface generation with working code, backend connections, and publishing. For simpler projects, the application itself could serve as the prototype used to test an idea. That likely removed some demand for a dedicated wireframing tool.
AI generation became easier to add: Visily described screenshot-to-design as its most distinctive AI feature at launch. By 2026, its Deep Design mode used Gemini 3, whose Pro model was available to other developers through an API. Google had already built Stitch around Gemini 2.5 Pro. That suggests a weakening technical advantage: improving general models lowered the custom AI work needed to build a competing interface generator, leaving Visily's editor and collaboration features to justify its separate subscription.
Why It Matters:
Check whether customers still need your tool once their existing platform becomes easier to use.
Make a separate design step earn its place through less rework or better collaboration.
Give customers a reason to stay when competitors can access the same AI models.
Trend
Will Everything Be Open Source?
Over the past two weeks, Twitter has been full of videos of people mixing games together. You can play Minecraft inside Skyrim, skateboard through Call of Duty maps, or bring Minecraft into GTA V.
I’ve spent the last few days going down a rabbit hole trying to understand how this works. How can you combine the systems of two games without access to their source code? And why are we suddenly seeing so many people do it?
The answers go well beyond gaming. They made me rethink how much control software companies can keep over the programs we install on our computers.
Why it Matters
Customers could customize software without waiting for its vendor, including products that have stopped receiving updates.
A competitor could reproduce the features customers actually use, putting pressure on businesses that charge for access to a local application.
If any binary is easy to decompile, then what happens to proprietary software?
How This Works
When you install a program, you usually receive compiled code: instructions your computer can execute. The readable source code its developers wrote usually stays with the company.
Decompilation works backward, turning those instructions into an approximation of readable code. Reverse engineering uses that code, observations of the running program, and other clues to work out how the software behaves.
Specialists have been doing this for decades. The problem was the amount of skilled labor involved: understanding unfamiliar code, recovering its logic, and checking that a change actually worked. For a small customization, that work could cost more than the result was worth.
Nowadays, AI can get the job done in a couple hours. This is why we are seeing this trend all over twtter. As more of the decompilation labor can be delegated to an agent, projects that were too expensive to attempt become worth trying. Access to the compiled program becomes a starting point for building something its original developers never offered.
Beyond Games
A game is a large program running on your computer. An image editor or a design tool also arrives as compiled files that can be inspected. The jump from experimenting with games to changing software people pay for is fairly small.
This week, the conversation has already reached Adobe. ArtCraft is building a family of seven free alternatives, including PhotoCraft for image editing, VectorCraft for illustration, and FilmCraft for video editing. Its AI-assisted projects aim to recreate the tools and workflows people already know.
The launch prompted claims that Photoshop was decompiled with AI. PhotoCraft describes its method as rebuilding from public specifications and observed behavior, and it's still an early alpha. The appeal is straightforward: keep familiar tools and workflows while replacing a paid application with a free one.
Here's another example: developers have recreated Yamaha's FS1R synthesizer, a physical musical instrument released in 1998, as a free application you can run on your computer.
They started with the software stored inside the original instrument. Earlier reverse-engineering work had turned it into readable code, which AI agents helped them understand and translate into a desktop version. The result is FSVR: you can hear it in action, make music with it, or change its code yourself.
The Trend
Are we heading toward a future where everything we install is effectively open source? If I can take apart a program, change it, and rebuild it, how different is that from forking a project?
One possible future is that local applications become routinely forkable. Users could add the features they want, and developers could maintain free alternatives to commercial products. A replacement would only need to do the jobs a particular customer cares about.
Companies could respond by running more of their valuable software on their own servers. A customer would receive the interface, while the underlying computation or private data remained remote. Decompiling the downloaded application wouldn't reveal the implementation of the service behind it. That gives businesses another reason to move parts of their products to the cloud.
We could also end up with free local applications attached to paid services. ArtCraft already sells AI generation credits alongside its free creative tools. You might edit an image in a free program and pay for the hosted model that generates the next one. That would make the installed app a way to bring in customers, with recurring revenue coming from something the app can't provide by itself.
Help Me Improve Failory
How useful did you find today’s newsletter?Your feedback helps me make future issues more relevant and valuable. |
That's all for today’s edition.
Cheers,
Nico