Listen to the Great AI Filter.
0:00
0:00
Google just released the future of their Chrome browser. To put it simply, it's AI everything. Meta also released their new smart glasses, complete with a "neural" wristband for input. It too is AI everything. The more I watched these product launches, with their proclamations about the future, the more I was reminded of this observation from Catch-22:
"While none of the work we do is very important, it is important that we do a great deal of it."
Everyone wants to use AI to do things for me that I actually enjoy doing. But they also want to automate tasks I don't like doing. Things that I wouldn't bother with in the first place.
Chrome's new Agentic browsing promises to, well, browse the web for me. But wait. I want to browse the web. That's the point. I want to read and discover the web. I want to stumble upon rabbit holes, I want to understand and connect ideas. When I search for something, I'm not just seeking an answer; I'm seeking the journey to that answer.
What Chrome is really offering isn't convenience. Chrome will wrap the web into a Google product and serve it to me algorithmically. It's a TikTokification of the web. On TikTok, you don't choose what to watch; the algorithm serves you what it believes you should consume next. When Chrome browses for me, it will surface what it wants me to see, filtered through corporate priorities, advertising relationships, and engagement metrics. Over time, through a kind of digital Stockholm syndrome, I'll start believing these curated choices reflect my authentic preferences. I'll mistake algorithmic manipulation for personal agency.
The Wrapper Economy
Most AI startups aren't much different. They're building what I call "wrapper products". Sleek interfaces layered over existing AI models like ChatGPT or Claude, promising specialized functionality that I could probably achieve on my own with a well-crafted prompt. A $50/month "AI writing assistant" that essentially runs your text through GPT with a custom system prompt. An "AI research tool" that performs Google searches and summarizes results. An "AI productivity app" that schedules your tasks using the same reasoning capabilities you could access directly.
These products often solve problems their creators assume you have, rather than problems you actually experience. They're solutions in search of problems, built by teams who've mistaken the impressive capabilities of foundation models for proof that any application built on top of them will be equally impressive.
Side Note: I know every demo shows how AI can help you cook or follow a recipe. But to tell you the truth, following a recipe is not a challenge most people have. The solution is in the name: follow the recipe!
The Necessary Filter
As much as I hate participating in the hype cycle, I've come to believe that these wrappers, vaporware, or unrealistic promises, serve an important function. Not because their individual features matter, but because collectively, they help us filter out what AI is not supposed to be.
We were promised the most transformative technology since electricity. Instead, we've been presented with an avalanche of shoddy AI tools that automate the wrong things, complicate simple processes, and solve problems that were better left unsolved. I'm looking at you Copilot.
Yet they're useful as a collective learning experience. They illuminate the vast chasm between hyped-up novelty acts and genuine, transformative applications. Every failed AI gadget teaches us something about the difference between technological capability and actual utility.
When the MIT Study came out, the one that found that 95% of AI initiatives failed to be useful in 300 companies. All I could think about was what were the successful use cases in the remaining 5%? Looking through the study, it seems like what they did was automate repetitive and tedious tasks. They used a bottom up approach, where front line employees drove the adoption of tools to make their job easier. They partnered with vendors who integrated AI with their existing tools, like call summarization and routing for customer service. In other words:
- They solve clear, measurable problems.
- They work behind the scenes, augmenting rather than replacing human judgment.
- They've been iteratively refined over years, not rushed to market in months.
- They integrate seamlessly into existing workflows.
- They make mundane tasks more efficient rather than trying to eliminate meaningful work.
When the AI bubble eventually deflates (and it will) these are the applications that will remain standing. Useful tools won't always look impressive in a product demo or generate viral social media posts. Their success stories will be small percentage improvements in efficiency that add up over time.
The Great Filter in Action
The concept of filters in technology isn't new. The dot-com boom gave us Pets.com alongside Amazon. The mobile app gold rush produced millions of forgotten apps alongside Uber and WhatsApp. Each cycle of technological hype includes a massive filtering process where the market eventually separates genuine innovation from opportunistic noise.
AI is undergoing the same process, but at unprecedented scale and speed. The current proliferation of AI products represents our collective attempt to understand what this technology is actually good for.
Every overengineered AI feature that gets quietly deprecated teaches the entire ecosystem something valuable. Every startup that discovers their AI wrapper provides no real value helps establish boundaries around what constitutes genuine utility. Every user who grows frustrated with intrusive AI assistance helps calibrate our understanding of when algorithmic help becomes algorithmic hindrance.
So in a way, none of the work being done by the 95% may be individually important, but it's important that a great deal of it is being attempted. It's a necessary waste. It's noise that allows genuine signals to emerge.
The real breakthroughs won't come from the companies promising to revolutionize everything. They'll come from the teams quietly solving specific problems with appropriate tools, building on solid foundations rather than chasing the latest trend.
The great AI filter is working exactly as it should. Most of what's being built will fail, fade, or pivot into irrelevance. But in that process of collective experimentation and failure, the signal will emerge from the noise. It always does. And when it does, it will be worth the wait.
the study: State of AI in Business 2025
