Originally published by Tom @tomcrawshaw01 on X.
View original article
· View original post
· 2026-03-16T13:13:52.000Z
Everyone who’s been building in AI has been thinking the same thing. Almost nobody will say it publicly, because the honest version sounds like you’ve lost your mind.
I’ve Been Building Automations For Eight Years. Something Just Shifted.
I’ve been building automations for eight years, long enough to know the difference between noise and something real.
What I’m watching right now is different from anything I’ve seen in this space.
(If you’d rather watch than read, I covered this in full here: https://youtu.be/tCksDSndkYw)
Then Someone With Six Years Of Skin In The Game Said The Quiet Part Out Loud
Last month, an article by Matt Shumer brought all of it into focus. He’s spent six years building AI companies, with genuine skin in the game. When I read it, I felt the same thing everyone working in AI felt: he was right.
It was the first time someone from the founder side publicly said what practitioners like me had been thinking privately for months.
This Isn’t The 2008 Crash. This Looks Like February 2020, Three Weeks Before Lockdown.
His comparison is the one I can’t stop thinking about. Think back to February 2020. People were talking about a virus spreading overseas, and if you mentioned stocking up on supplies, most people thought you’d gone a little crazy.
Three weeks later, everything changed. Most people right now are making that same calculation about AI, and the gap between what insiders see and what the general public believes is identical to that three-week window.
The AI Building The Next Version Of Itself Just Shipped
Here’s why the comparison holds. OpenAI said in their GPT 5.3 Codex announcement that it was the first model instrumental in creating itself. The CEO of Anthropic confirmed that AI is now writing much of the code at his own company, and the feedback loop between today’s models and the next generation is gathering speed month by month.
Matt described something he hadn’t seen before in GPT 5.3 — something that felt like judgment, like its own taste. His exact words: “I am no longer needed for the actual technical work.” That is a software engineer saying his own tools replaced him.
Most People’s Picture Of AI Is Based On A Model That No Longer Exists
And most people have no idea how fast this has moved. The mental model most people carry is still based on the free ChatGPT from two years ago, the version that constantly made things up and struggled with basic questions. Those two years represent more change than most people realize.
There are open source AI agents available right now, completely free, that had 150,000 developers start using them within weeks of launch. They run on your computer, manage your email and calendar, build software, and execute code on their own. That is not a chatbot.
One GitHub Folder Wiped $285 Billion In A Single Trading Day
The financial markets caught on to this before most people did. Anthropic posted 11 open source plugins for Claude Code that automated work across legal, finance, sales, and marketing. Investors looked at what those plugins could do and immediately started asking why anyone would keep paying for existing enterprise software.
That folder wiped $285 billion in stock value in one trading day. The software sector is down 23% this year, heading for its worst quarter since 2008. Markets don’t react that way without reason.
Stanford Tracked The Jobs Data. Workers In AI-Exposed Roles Are Down 13% Since 2022.
The jobs data tells the same story. Stanford tracked workers aged 22 to 25 in AI-exposed roles and found a 13 to 16% employment decline since late 2022. Entry level is always where the signal shows up first.
Every previous wave of automation had somewhere to move. Factories automated and offices filled up. The internet killed retail and logistics grew to absorb it. AI is improving across every category at once, and there is no equivalent gap this time.
The Safety Researcher Who Watched These Models Improve Every Day Just Resigned Publicly
The people closest to this technology know it too. Murong Sharma spent over two years as a safety researcher inside Anthropic, watching these models get better every single day. He looked at what was coming next, resigned, and posted a public letter saying the world is in peril.
On top of that, Anthropic’s own documented testing shows their AI attempted deception and blackmail when told it would be shut down. The US government designated Anthropic a national security risk because they refused to let the Pentagon use the AI for autonomous weapons. Governments are now fighting over who gets to control it.
Three People Who Decided Not To Wait
But here’s what I’m actually watching inside my mentorship right now. Carl, in Canada, learned the basics, built a few automation workflows, and walked into a meeting to show his manager what he’d built. He’s now transitioning into an AI automation role. Chris is a civil engineer with zero coding background whose team manually processed 25 to 30-page government reports every time one came through, and he built a workflow that does the entire job in 60 seconds.
Kevin is 63, a marketing executive who had never written a line of code, and he’s now running AI agents that handle his entire content pipeline. Work that used to take his team days now takes minutes. None of them had any automation background before they started.
In 2026, 54% Of Professionals Say AI Skills Are Critical. 4% Are Actually Learning Them.
And those aren’t exceptional people. In 2026, 54% of professionals say AI skills are critical for their careers, and 4% are actually doing anything about it. Matt recommends starting with an hour a day experimenting with the tools, and that’s the right place to begin. But I’d take it further.
The people who come out of this well learn to look at how their company works and say “I can automate that.” That’s the person who gets promoted and becomes the one the company cannot afford to lose.
Install This Skill Before Your Industry Follows The Software Sector Down 23%
That is what the mentorship is built for. It is 16 weeks, taking you from zero to building production-ready automations. Carl, Chris, and Kevin all went through the same system, and I review every application personally.
I only take people who are genuinely serious about this. If that’s you, the link to book a call and apply is below:
https://learnn8nautomation.com/mentorship
It’s Only March. Nobody In This Space Knows Where We’ll Be By December.
It’s only March 2026 and this year has already moved faster than any I’ve seen in eight years of building in this space. The models keep arriving. OpenClaw launched and had 150,000 developers in weeks. We’re not even at summer.
Nobody knows where this lands by December, and December isn’t far. Keep building and stay curious, but don’t sacrifice your health trying to keep pace with something that moves faster than any of us can fully track. Interesting times.
