Growth Ritual #86
đ In This Issue:
The Era of Context: Why the Webâs Inventor and Boxâs CEO Are Fighting the Same AI Battle
The Million-Dollar Ideas Hidden in a Research Paper Nobody's Reading â đ
Whale Fall Opportunities: The Untapped Goldmine for Indie Hackers â đ
The Frozen Vault: A Trillion-Dollar Biotech Market Is Thawing in the Arctic â đ
Stop Trying to âRankâ, Start Trying to Get âStolenâ by AI â đ
Know someone whoâd love this? Invite them and youâll both winâyour friend discovers the newsletter, and you unlock 1 month of premium access.
The Era of Context: Why the Webâs Inventor and Boxâs CEO Are Fighting the Same AI Battle
A digital ghost of you is being built inside the servers of Google, OpenAI, and Meta.
Itâs a high-fidelity echo assembled from every late-night brainstorm, every confidential business query, every personal vulnerability youâve fed into their chatbots.
This ghost knows your business strategy better than some of your employees. It understands your creative style better than your clients.
Thereâs just one problem: You donât own it.
You canât take it with you to another platform. You canât deploy it in your own applications. You are building their most valuable asset, for free.
This is the quiet, central heist of the AI era. While the world is mesmerized by the magic tricks of new models, the real action isnât about creating intelligence. Itâs about capturing it. Specifically, yours.
Forget the AI wars you see in the headlines. Thatâs a smokescreen. The real fight is for control over these digital ghosts.
The next trillion-dollar prize wonât go to the company with the most parameters. It will go to whoever wins the war for context.
The battleground isnât intelligence; itâs the ownership and portability of the data that feeds intelligence. And if youâre not paying attention, youâre positioning yourself to be a pawn in someone elseâs empire.
The Three Fronts of the Real AI War
Iâve been tracking this from three different altitudes, and what I see is a three-front war converging on a single, explosive opportunity.
You have the visionaries, the original architects, and the corporate pragmatists all arriving at the same conclusion.
Letâs break it down.
Front #1: The Visionaries & The Memory Lease
On the bleeding edge, you have a radical idea brewing: the next big thing isnât building a smarter AI, but convincing you to lease your own memories back to yourself.
Think about it. Every prompt you feed to ChatGPT, every therapy session you have with Claude, every private thought you share with Geminiâit all gets absorbed into their proprietary data vortex. You get a clever response; they get the raw material to build their moat, leaving you with zero equity in your own digital soul.
The rebellion is quiet but potent:
Decentralized memory wallets that act like a MetaMask for your entire digital life.
Zero-knowledge pods that let your data train models locally, without ever leaving your device.
Memory NFTs that could turn your anonymized insights into a royalty-generating asset.
This isnât sci-fi. Itâs the ultimate end-game of data sovereignty. The battleground isnât parameters, itâs portability.
Front #2: The Architect & The Data Vault
If the first front sounds like cyberpunk, the second is grounded in the foundational principles of the internet itself.
Tim Berners-Lee, the guy who literally invented the World Wide Web, is so alarmed by the data-siloed mess weâre in that he co-founded a company, Inrupt, to fix it.
His solution is an open protocol called Solid.
The genius is its simplicity: Solid separates your data from the applications that use it.
You store your photos, contacts, health records âand yes, your AI conversationsâ in a personal online data store, or âpodâ.
Want to try a new social network? Grant it permission to read from your pod.
Tired of it? Revoke access with one click. Your data never leaves your control.
Berners-Leeâs vision is to replace the entire economic engine of the web.
Instead of apps hoarding data to survive, they simply become âviewsâ that ask your pod for permission.
This is the architectural blueprint for a world where you own your digital footprint.
Front #3: The Pragmatist & The Corporate Guardrail
This is where the rubber meets the road today. While visionaries dream and architects build, corporations have an immediate, burning problem. Box CEO Aaron Levie nailed it when he said weâre in the âera of contextâ for AI.
Businesses canât just dump their sensitive, unstructured data âcontracts, marketing plans, M&A docsâ into a third-party model. The risks are insane.
What they need is a secure layer that provides the right context to the AI, with absolute control over permissions, governance, and security.
Levieâs strategy at Box is brilliant because itâs platform-agnostic.
They arenât betting on a single AI model. Theyâre building the one thing every enterprise will need: a secure context engine.
They manage the unstructured data, the permissions, the vector embeddings, and then connect to any leading AI model the customer wants.
They are becoming the trusted intermediary between a companyâs data and the commodity world of AI models.
The Strategic Core: Building a Business on Context-as-a-Service (CaaS)
For the past two years, the startup graveyard has been filling up with âAI-poweredâ apps that are little more than thin wrappers around an API key.
They all follow the same doomed playbook: take a prompt, send it to OpenAI, and display the result in a slightly prettier text box. This isnât a business model; itâs a feature with a high burn rate and no real staying power.
The real, defensible opportunity isnât at the application layer. Itâs one level deeper, at the context layer.
Weâre witnessing the birth of a new category â Context-as-a-Service (CaaS).
CaaS is the secure, proprietary middleware that connects a userâs unique, unstructured data to the commodity brain of any large language model.
Think about it. The LLM is the engineâpowerful, but useless without fuel and a chassis.
The CaaS platform is the entire drivetrain, fuel injection system, and armored chassis. Itâs the infrastructure that makes the engine useful and safe.
This is how you build an enduring competitive advantage:
The Sticking Power Is in the Interface: A CaaS business owns the trusted connection to the customerâs most valuable asset: their proprietary data. Once a companyâs entire library of legal contracts, marketing assets, or M&A documents is integrated into your CaaS platform (the Box model), the switching costs become astronomical. They arenât just using an app; their data lives inside your system.
The Real Product Is Trust: You stop selling âsmarter AIâ and start selling something far more valuable to businesses: security, permissions, and governance. You provide the auditable guardrails that prevent a summer intern from asking the AI about the upcoming layoff plans. Thatâs a value proposition enterprises will pay a massive premium for.
You Future-Proof with Interoperability: CaaS platforms are model-agnostic. When GPT-5 is released or a new open-source model suddenly outperforms the incumbents, your customers donât have to re-architect a thing. You simply unplug one commodity brain and plug in another. You make your customers resilient to the chaos of the AI market.
So, stop thinking about building another AI writing app. Instead, build a CaaS platform for screenwriters that securely manages their scripts, character bibles, and world-building notes, allowing them to plug in any AI for analysis.
Donât build another AI chatbot for lawyers. Build a CaaS engine for law firms that creates a secure, auditable context layer from their case files, which can then be queried by a vetted, verifiable AI model.
Stop building disposable AI apps. Start building indispensable CaaS engines. One is a temporary feature. The other is a foundational platform your customers canât live without.
The Million-Dollar Ideas Hidden in a Research Paper Nobody's Reading
Every once in a while, a research paper drops that isn't just an incremental improvement. It's a seismic shift. A quiet little earthquake that most of the industry sleeps through.
Right now, one of those earthquakes is happening, and almost nobody is talking about it.
A paper from researchers at Intel, AMD, and NYU introduces a new way to represent images. The tech blogs will tell you it's a âbetter JPEGâ
They're missing the point. Completely.
This isn't about slightly smaller file sizes. It's about inventing an entirely new primitive for visual data. And for ambitious founders, hackers, and marketers like you and me, that means one thing: a gold rush is about to begin.
The Tech in 60 Seconds: From 3D Worlds to 2D Images
First, what the hell is this thing?
The technology is called Gaussian Splatting. Originally, it was a mind-blowing way to capture 3D scenes.
Imagine pointing a camera at a real-world object and having it instantly recreated as a cloud of millions of tiny, colorful, semi-transparent blobs (the "Gaussians").
It's hyper-realistic and renders in real-time. Think of it like a pointillist painting, but in 3D and interactive.
The new paper, titled "Image-GS", asks a brilliant, almost stupidly simple question:
âWhat if we used this 3D scene technique on a flat, 2D image?â
The process is beautiful in its cleverness:
It takes a standard image (like a photo of a rover on Mars).
It creates a handful of these Gaussian "blobs" and throws them at the canvas.
Then, the magic happens. An algorithm "massages" them âstretching, moving, repainting, and adding new blobsâ until they perfectly match the original image.
The result? An image that's 25-40 times smaller than the original, with dramatically better quality than a JPEG of the same size.
And it does this in a couple of seconds.
It's so fast they had to slow down the video of the process so you could actually see it working.
Cool, right?
A better, faster image format. But if that's all you see, you're looking at the map and ignoring the buried treasure.
Why This Isn't a Better JPEGâIt's an Entirely New Reality
I've been in the app game for over 20 years. I've built and launched over 50 products. If there's one thing I've learned, it's how to spot the difference between a feature and a foundation.
JPEG, PNG, GIF... they are all just digital photographs. A grid of dumb pixels, frozen in time. They store color values, and that's it. Itâs a static, dead representation of a moment.
What this paper just demonstrated is the equivalent of a vector graphic, but for the photorealistic world.
Each "splat" isn't just a color value at an (x,y) coordinate. It's an object with properties: position, size, rotation, shape, transparency.
This means the image is no longer a static file. It's a programmable scene.
Think about the difference between a printed map and Google Maps.
One is a dead piece of paper. The other is a live, queryable, and manipulable dataset. This is the same leap, but for all visual media.
We're moving from dumb pixels to smart primitives. And when you change the primitives, you change everything that can be built on top.
Million-Dollar Ideas Hiding in Plain Sight
When a foundational shift like this happens, the biggest opportunities aren't in improving old workflows, but in creating entirely new ones. Here are three ideas you can run with today.
1. The "Live" E-commerce Catalog SaaS
The Problem: An e-commerce store selling a sofa in 10 different fabrics needs 10 different expensive, high-res photoshoots. It's costly, slow, and rigid.
The Image-GS Solution: You have one master file of the sofa represented by Gaussians. A customer clicks a fabric swatch, and your AI doesn't load a new image. Instead, it identifies the splats representing the "sofa fabric" and changes their color and texture properties in real-time.
The Business: Build a Shopify plugin or a standalone SaaS that converts a merchant's standard product photos into these "live" Image-GS assets. You're not selling image compression; you're selling the end of expensive, variant photoshoots forever.
2. The Indie Game Dev's Secret Weapon
The Problem: High-quality game textures are massive. They kill load times, bloat download sizes, and limit the visual fidelity indie developers can achieve on a budget.
The Image-GS Solution: This tech provides absurdly high-quality textures at a fraction of the file size. Plus, because it naturally creates levels of detail (LODs), objects far away can be rendered with fewer splats, saving precious compute power.
The Business: Create a service or a Unity/Unreal Engine plugin that automatically converts standard game assets into this new format. You become the essential optimization tool for every indie developer trying to make their game look triple-A without a triple-A budget.
3. The âData-Aware Visualsâ SaaS
The Problem: Photo editors work on pixels. To change the price on a promotional car photo, you have to use a lasso tool to painstakingly select the pixels that make up the price digits. Itâs a manual and âdumbâ process.
The Image-GS Solution: Since the splats are generated based on the imageâs content, they inherently carry semantic information. A cluster of shiny, red splats is the carâs price tag. You could build a tool where the command is simply, âTake the new prices from our Excel file and update the price visuals on all product photos across the siteâ. The AI finds the relevant splats and updates their content.
The Business: Build the next Canva or Photoshop, but founded on this data-driven principle. This is true, semantic, AI-native visual manipulation.
Stop Looking at the Pixels, Start Seeing the Platform
The media will catch up to this in six months and call it an "overnight success." You know better. The real opportunity isn't in waiting for this to be packaged into a new Adobe feature.
The opportunity is now. It's in building the shovels for the gold rush. It's in understanding the foundational shift from static pixels to programmable visuals. This is precisely the kind of non-obvious wave I wrote my latest book, The AI Entrepreneur's Guide, to help you spot and ride. The "what to build" is right here in this email. The "how to build it", brick by brick, from idea to global scaleâthatâs the framework in my first book, The Product Growth Playbook.
The vision and the execution. You need both.
This new paper provides a massive piece of the vision. The question is, who is going to execute?
So, I'll ask you: What's the first 'programmable image' app you would build with this tech?
