Death by a thousand agents

Nobody is keeping up with AI. The ones who thrive will be the ones who stop trying to.

Apr 14, 2026

👋 Hey, I’m George Chasiotis. Welcome to GrowthWaves, your weekly dose of B2B growth insights—featuring powerful case studies, emerging trends, and unconventional strategies you won’t find anywhere else.

This note is brought to you by Ahrefs.

I’ve been using Ahrefs for years. That hasn’t changed.

What has changed is how we track visibility.

Now, I know there are a lot of tools popping up for AI search. (We counted 200+ in a report we published.)

But, I’m only using two: one we’ve built internally (and know its limits), and Brand Radar by Ahrefs.

It’s the only external tool I trust for AI search tracking.

You can track your brand’s visibility across AI platforms like ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews, and compare it against competitors with clean, useful data visualizations.

Plus, you can track citations, your AI share of voice, and much more.

If AI search matters to you (it should), go check it out:

Check it out here

In 2024, Tyler Denk wrote a piece called “Death by a Thousand Substacks.”

His argument was that Substack, despite marketing itself as a haven for independent writers, was quietly eroding creator ownership by turning itself into a social network.

Platform incentives slowly diverging from publisher incentives. A calculated drift that most creators wouldn’t notice until they were locked in.

I’ve been thinking about that framing a lot lately. Not in the context of newsletters. In the context of AI.

Because something similar is happening to every knowledge worker, marketer, founder, and content professional I know.

We’re not being killed by one big disruption. We’re being slowly overwhelmed by a thousand small ones.

A new model every few weeks. A new agent every week. A new startup that will replace you every other week. A paradigm shift every quarter.

Each one individually exciting. Collectively, they are frying our brains.

Buckle up and let’s see how we got here, and how we can come out stronger on the other end.

How We Got Here: 40 Months That Changed Everything

Before I get into the why, I want to lay out the what.

Because when you compress the timeline of AI development from April 2022 to March 2026 into a single view, the sheer density of what happened becomes visceral.

Read that table slowly, and try to really absorb what happened.

In 40 months, we went from “wow, this chatbot can write a poem” to autonomous agents that browse the web, operate software interfaces, write and deploy code, and connect to every enterprise tool through standardized protocols.

Forty months. Not a lifetime, right? And yet, we’ve experienced more changes than previous generations experienced in the span of their lifetimes.

The Cognitive Cost of Keeping Up

Here’s the thing:

The pace documented in that timeline isn’t just professionally challenging. It’s cognitively unsustainable.

A 2025 study published in PMC found a near-perfect correlation (r = 0.905) between prolonged AI tool usage and mental exhaustion, attention strain, and information overload.

The same study found an inverse relationship (r = −0.360) between AI use and decision-making self-confidence. In plain English: the more people use AI tools, the less confident they become in their own judgment.

That finding alone should stop us in our tracks. But the data gets worse.

According to an Upwork study of 2,500 workers, 77% of employees using AI said these tools actually increased their workload. Not decreased it. Increased.

Thirty-nine percent spent more time reviewing and moderating AI-generated content. Twenty-three percent invested more time learning how to use new tools. Twenty-one percent were simply asked to do more work because management assumed AI would handle the overflow.

And here’s where it gets truly contradictory:

Microsoft’s Work Trend Index found that 90% of AI users say it saves them time, 85% say it helps them focus on important work, and 84% report increased creativity.

Yet The Wall Street Journal reported in January 2026 that over 40% of executives claim AI saves them 8+ hours per week, while two-thirds of rank-and-file employees report saving zero to two hours. Forty percent of workers said they’d be fine never using AI again.

Productivity, apparently, depends on where you sit in the org chart.

Apollo’s chief economist captured the macro version of this paradox perfectly:

AI is everywhere except in the incoming macroeconomic data.

I think Robert Solow said something similar about computers in 1987. I guess history indeed rhymes…

The Burnout Nobody Talks About

The cognitive strain isn’t abstract. It’s showing up in burnout data that most people in our industry (myself included) are choosing to ignore:

A Quantum Workplace study found that 45% of frequent AI users reported burnout, compared to 35% among non-users.
An EY survey from 2025 revealed that more than half of senior leaders feel like they’re not keeping up with AI’s rapid evolution. A similar proportion reported company-wide enthusiasm for AI is actively declining.
42% of companies in 2025 scrapped the majority of their AI initiatives, up from 17% the year before. That’s not a slowdown in innovation. That’s organizational exhaustion (and pushback).
73% of tech founders report a hidden mental health crisis, with over half pointing specifically to AI-related industry disruption as a major stressor. And 78% of knowledge workers are now bringing their own AI tools to work, creating a shadow-IT situation where organizations have zero visibility into how AI is actually being used. Only 39% have received any formal training.

Needless to say, all this is not sustainable.

Death by a Thousand Agents

Tyler Denk’s original “Death by a Thousand Substacks“ argument was about platform incentives diverging from creator incentives.

Substack slowly extracting ownership, identity, and distribution from writers who thought they were building something independent. Each individual change was small. Collectively, writers woke up one day and realized the platform owned more of their relationship with readers than they did.

(Unfortunately, as my newsletter is on Substack, I can experience that change firsthand.)

The AI version of this story works in a similar way.

No single model release, no individual agent framework, no standalone product launch is overwhelming on its own. Each one is manageable. Interesting, even.

But the compound effect? That’s where it breaks down.

Consider what a marketing leader at a mid-market B2B company had to process in 2025 alone:

Claude 4 launching with genuinely new capabilities
GPT-5 arriving three months later
DeepSeek disrupting pricing assumptions
Autonomous agents from OpenAI (Operator) and Anthropic (computer use)
MCP becoming an industry standard
Cursor transforming how developers work
Vibe coding entering the lexicon
AI search reshaping marketing strategy
Grok 3 and Grok 4 entering the arena
Reasoning models becoming a distinct product category

That’s one year. For one person. On top of their actual job.

And every next year will always be faster.

What Actually Helps (A Practical Framework)

I promised this would be practical, not just philosophical. So here’s how I’ve been thinking about surviving this without losing my mind or falling behind.

Stop trying to track everything: The landscape has 10,000+ AI tools and 200+ open-source LLMs. Nobody can evaluate even a fraction of these. Pick two or three tools that solve real problems in your workflow and go deep on them. Ignore the rest until they become impossible to ignore.
Batch your learning: Instead of reacting to every announcement in real time, set a monthly review cadence. What shipped in the last 30 days that’s actually relevant to your work? Most of what seems urgent on announcement day is (very often) irrelevant by the following week.
Distinguish between capability shifts and noise: ChatGPT’s launch was a capability shift. GPT-4o was a capability shift. Computer use agents were a capability shift. MCP was a capability shift. Most incremental model updates are noise. Learn to tell the difference and only invest attention in the former.
Protect your judgment: Remember that inverse correlation between AI use and decision-making confidence. If you find yourself deferring to AI outputs without critical evaluation, step back. The tools should augment your thinking, not replace your conviction.
Accept that falling behind is the new normal. Nobody is keeping up. Not the VCs tweeting about every launch. Not the AI influencers writing daily threads. Not the founders building in this space. Everyone is selectively informed and selectively ignorant. The professionals who thrive will be the ones who are intentional about what they choose to learn and what they choose to skip.

Let’s wrap this one up.

Final Thoughts

Forty months ago, most people had never interacted with a large language model.

Today, 75% of knowledge workers use AI tools daily, autonomous agents operate software on our behalf, and the average professional encounters a significant AI development roughly every two weeks.

Progress is genuinely extraordinary. What we’re building and what we’re capable of now would have been inconceivable just three years ago. I don’t want to diminish that.

But progress without pause is just acceleration. And acceleration without reflection produces burnout, shallow adoption, and a collective inability to think clearly about what any of this actually means.

The companies and individuals who come out of this era strongest won’t be the ones who adopted every tool first.

They’ll be the ones who maintained the cognitive clarity to understand which tools actually mattered, and had the discipline to let everything else pass.

Don’t die by a thousand agents. Choose your battles.

Thank you for reading today’s note, and see you again next week.

Research Disclaimers and Limitations

GrowthWaves and its author are not sponsored by or compensated by any company mentioned in this note. This is independent editorial analysis and does not constitute investment, financial, or legal advice. The author may have relationships with, work with, or hold equity in companies referenced; however, no content in this piece was influenced, commissioned, or incentivized by any such relationship. AI tools were used as a research assistant in the preparation of this piece. All claims are sourced and linked throughout.

Sources

Intro

Death by a Thousand Substacks — Tyler Denk, Big Desk Energy

AI Timeline

Cognitive Overload & AI Fatigue

Productivity Paradox

Burnout & Mental Health

GrowthWaves by George Chasiotis

Discussion about this post

Ready for more?