The February 2020 of AI: Navigating the Intelligence Explosion

Do you remember February 2020?

There were murmurs about a distant virus. Most dismissed it as media hysteria, booking flights and planning weddings, blind to the fact that their world would implode in weeks.

We're in that moment again. But this time, the contagion isn't biological; it's cognitive.

Last week, Matt Shumer, CEO of HyperWriteAI, published an essay ("Something Big Is Happening") that 80 million people read. It crystallized a reality that has been unfolding behind the closed doors of elite AI labs.

Matt Shumer, a tech industry thought leader, confessed something startling: he no longer does his job. Not the technical parts, anyway. He doesn't write code nor does he debug. He describes what he wants in plain English to AI, walks away, and returns to production-ready software.

This isn't a productivity hack.

It's an intelligence explosion, and the math is staggering.

METR, an organization tracking AI capabilities, reports that frontier models went from handling 10-minute tasks in early 2025 to executing five-hour expert-level projects autonomously today. By 2027, these systems will manage full workdays, then work weeks, without human intervention.

McKinsey projects AI will add $13 trillion to global GDP by 2030 while displacing 800 million jobs. Dario Amodei, CEO of Anthropic, predicts half of all entry-level white-collar work vanishes in less than five years.

Then there's this: OpenAI's GPT-5.3 Codex helped build itself. It debugged its own training, managed its own deployment, and evaluated its own outputs.

We have crossed a threshold.

We are entering a recursive loop where the thing we are building is building better versions of itself.

The economic implications are violent and the stock market reaction to the most vulnerable sectors reflects these fears.

SaaS stocks have cratered 40% since January. Companies built on the premise that humans need software to do their jobs are hemorrhaging value as investors realize AI builds its own tools.

Commercial real estate dropped 15% last week. Why lease 50,000 square feet when AI doesn't need desks or parking spots? Why pay premium service fees when AI replaces labor-intensive back office functions?

Logistics stocks are in freefall. Amazon eliminated 15,000 warehouse positions last quarter, replacing them with AI-orchestrated robotics. FedEx is closing distribution centers that AI-driven route optimization has rendered obsolete.

No industry is immune. Legal research. Financial analysis. Customer service. Supply chain management. Marketing. HR. Anywhere pattern recognition and execution can be codified, AI is already better, faster, and cheaper.

So what do you do?

You prepare to thrive in the turbulence instead of denying it exists.

Every major technological shift creates as much as it destroys. The industrial revolution displaced farmers but built cities, universities, and entirely new categories of human flourishing.

AI will be no different.

Somewhere right now, someone is building a company that couldn't have existed six months ago. They are solving problems we didn't know we could solve.

In a world where execution is free, vision becomes priceless. Where information is infinite, curation becomes the skill. Where AI handles the known, the unknown becomes your territory.

What comes next isn't predetermined. It will be built by people who see this moment as an invitation to discover what humans are capable of when we are finally free to focus on problems that actually matter.

This isn't the end of human work.

It's the end of humans doing work that machines should have been doing all along. And the beginning of something we've barely begun to imagine.

THE SHIFT

The generic advice of telling someone to learn how to use AI right now is like telling someone in 1995 to learn how to use a fax machine. It completely misses the magnitude of the moment.

We are experiencing a fundamental rewiring of how thinking happens. To lead through this, you have to internalize three specific shifts:

From Static Tools to Self-Improving Agents

In the past, technology was just a tool. A calculator only does math when you push the buttons. Traditional software only gets better when a human developer pushes an update. But the new frontier models are not tools; they are autonomous entities.

You aren’t buying a better hammer; you are leasing a digital carpenter. This carpenter works in the dark, checks its own measurements, fixes its own mistakes, and gets smarter without you having to train it.

You must stop managing your AI like a software subscription and start managing it like a digital employee in a self-improving workforce.

From Narrow Automation to Broad Substitution

When factory robots arrived, they replaced muscle. When the internet arrived, it disrupted industries like retail and media. If your industry was hit, you could just move to a safer one.

AI is entirely different. It isn’t learning a specific skill; it is learning how to reason. It is cognitive substitution across every discipline simultaneously.

In law, AI is drafting briefs faster than associates. In medicine, it is analyzing and diagnosing medical scans. In finance, it is building financial models and optimizing portfolios. There are no "safe" departments left to hide in.

Stress test your business model against a zero-labor-cost AI substitution. Plan for a smaller, more elite workforce.

From Human Taste to Strategic Direction

For the last two years, we held onto a safety blanket: AI can do the grunt work, but humans will always be needed for the final review because only we have taste and judgment.

That blanket has been pulled away.

The newest models don't just write code; they open the app, write the code, test the interface, then critique the design and user experience, and redesign the app if it feels clunky.

AI now has taste and judgment!

Human capital is no longer needed to manage the quality of the task. They are needed to manage the direction of the business.

Invest in developing leaders who can set direction in radical uncertainty, not managers who optimize known processes.

THE STRATEGY

The time for theoretical discussions is over. You need to structurally prepare your organization for cognitive substitution.

Do not wait for a company-wide AI rollout. It will be too slow. You need to stress-test the future today.

The Action Plan:

1. Isolate a Workflow: Identify one highly manual, cognitive-heavy process in your organization (e.g., Q2 budget reconciliation, generating initial legal contracts, or tier-1 customer support escalation).
2. Build a Skeleton Crew: Carve out a team of just one or two high-agency architects. Give them a budget for the most advanced frontier models available.
3. The Mandate: Their only job is to redesign this entire workflow so that an AI agent executes 90% of it, with the human only acting as the final approver.
4. The Post-Mortem: Compare the speed, cost, and error rate of the Skeleton Crew against your traditional team.

Use this data to audit your organization with brutal honesty. Which roles are doing tasks AI already does better? Stop hiring for positions that won't exist in 18 months.

THE STACK

Here are the three distinct tools driving the intelligence explosion and how your team must use them:

GPT-5.3 Codex (OpenAI)

• The Power: It possesses the ability to assist in its own creation and deployment. It doesn't just write a script; it builds, tests, and refactors entire software architectures.
• The Use Case: Empower your non-technical operations leaders to build their own internal tools. Instead of waiting 6 months for IT to build a custom dashboard, an ops manager can prompt Codex to build a secure, functional app over the weekend.

Opus 4.6 (Anthropic)

• The Power: Unmatched reasoning and taste. Anthropic has focused heavily on reducing hallucinations and increasing deep, multi-step logical deduction.
• The Use Case: Upload a competitor's 200-page SEC filing, your own internal product roadmap, and a year's worth of customer support tickets. Ask Opus 4.6 to identify the three biggest vulnerabilities in your current market positioning.

GitHub Copilot Enterprise

• The Power: Copilot is already boosting programmer productivity by 55%.
• The Use Case: If you manage a development team, mandate Copilot integration. You do not need to double your engineering headcount this year; you need to double the output of the engineers you already have.

Which roles in your organization could be entirely absorbed by a frontier AI model and a redesigned workflow?

Until next time...stay curious!

Cheers,
Nikki