TL;DR: AI just hit an inflection point. On February 5, 2026, both OpenAI and Anthropic dropped models that changed the game — GPT-5.3 Codex can help build itself, and the data shows AI capability is doubling every 4-7 months. Here's what the numbers actually say, what it means for your career, and what you should do about it right now.
The February 5th wake-up call
Matt Shumer, CEO of OthersideAI/HyperWrite, published an essay last week that's been making the rounds. It's called "Something Big Is Happening" and it compares where we are with AI to February 2020 — that window before COVID hit when a few people saw it coming and most didn't.
Strong comparison. Maybe even underselling it.
Here's what triggered it: on February 5th, OpenAI released GPT-5.3 Codex and Anthropic released Claude Opus 4.6 on the same day. Two frontier models, same day, both big leaps forward. Shumer describes telling AI what to build, walking away for four hours, and coming back to finished software that needed zero corrections.
I've been watching this space closely, and I can tell you — the reaction from people actually using these tools daily isn't hype. It's closer to shock.
The numbers behind the hype
Shumer's essay is compelling, but it's largely anecdotal. Let me fill in the data.
METR (Model Evaluation & Threat Research) has been tracking something specific: how long of a task can AI complete autonomously, end-to-end, without human help? They measure this as the "time horizon" — the length of a human-expert task where AI succeeds 50% of the time.
Here's the trajectory:
- 2022: AI couldn't reliably do basic math. 7 × 8 = 54? Sure, why not.
- 2023: Tasks taking a human a few minutes.
- 2024: Tasks taking about an hour.
- Late 2025: Tasks that take a human expert nearly 5 hours (Claude Opus 4.5, November 2025).
The doubling time? About 7 months on average from 2019-2025. But the January 2026 update shows it may be accelerating — closer to 89 days for recent models. That's a doubling roughly every 3 months on software tasks.
If you extend that trend (and it has held for six years with no sign of flattening), we're looking at AI that can work independently for days within the next year. Weeks within two. Month-long projects within three.
GPT-5.3 Codex just told us something important
OpenAI buried a remarkable line in their technical documentation for GPT-5.3 Codex:
"GPT-5.3-Codex is our first model that was instrumental in creating itself. The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations."
Read that again. The AI helped build itself.
This is what AI researchers call the beginning of a "recursive improvement loop." The model is smart enough to contribute to its own development. Each generation helps build the next one, which is smarter, which builds the next one faster. Dario Amodei, Anthropic's CEO, says AI is now writing "much of the code" at his company and that we may be "only 1-2 years away from a point where the current generation of AI autonomously builds the next."
GPT-5.3 Codex is 25% faster than its predecessor on coding tasks. It tops SWE-Bench Pro (the industry benchmark for real-world software engineering). And beyond raw performance, Shumer describes something more unsettling — the model shows what feels like judgment. Not just technically correct answers, but good taste about what the right call is.
Meanwhile, Opus 4.6 brings 1 million tokens of context
Anthropic's release deserves its own section. Claude Opus 4.6 features a 1-million-token context window (in beta), 128K max output tokens, and something called "adaptive thinking" where the model dynamically adjusts how deeply it reasons about a problem.
Translation: you can feed it an entire codebase, an entire legal case file, or months of financial data — and it can hold all of it in working memory at once. That wasn't possible six months ago. If you're wondering what that means for AI agents that can work for hours without supervision — yeah, it changes everything.
The job numbers are getting real
This is the part where it stops being a tech story and starts being everybody's story.
Amodei publicly predicted that AI will eliminate 50% of entry-level white-collar jobs within one to five years. He said this could push U.S. unemployment to 10-20%. And many in the industry think he's being conservative.
Here's what the research firms are saying:
- Goldman Sachs: AI could displace 6-7% of the U.S. workforce as a baseline — that's 10-12 million jobs. Under aggressive adoption, up to 14%.
- McKinsey: Up to 30% of U.S. work hours could be automated by 2030, with generative AI responsible for more than 10% of that reduction.
- IMF: 40% of global jobs are exposed to AI disruption.
- 2025 tracking data: 55,000 AI-related job losses in the U.S. in the first 11 months of 2025 alone — 75% of all AI-related layoffs since 2023.
And it's already showing up in specific industries. Google cut 6,000 jobs. Walmart eliminated 1,500 corporate roles. CrowdStrike cut 5% of its workforce. These aren't companies struggling financially — they're companies that figured out AI can do the work.
The $650 billion bet
If you're wondering whether the big companies actually believe this, look at the money.
Microsoft, Google, Amazon, and Meta are spending a combined $650-700 billion on AI infrastructure in 2026. That's a 50-60% increase from 2025. Amazon alone is spending $200 billion. Google is doubling its AI spending for the second year running.
These aren't speculative bets. These companies have seen their internal productivity numbers. They know what's coming because they're already living it. And they're building the infrastructure to deliver it to everyone else.
What this actually means for you
I talk to business owners and professionals every week who are still on the fence about AI. Some tried ChatGPT in 2023, got a mediocre answer, and moved on. Others think their field is somehow immune.
Here's what I'd say to both groups: the AI you tried in 2023 has almost nothing in common with what exists today. Judging current AI by your 2023 experience is like evaluating smartphones by using a 2007 flip phone. The free-tier versions are over a year behind what paying users have access to.
Shumer makes a point that stuck with me: the people who are ahead in their industries aren't dismissing this. They're spending hours a day with these tools and positioning themselves accordingly.
A few things worth doing right now:
Get hands-on with current models
Sign up for the paid version of Claude or ChatGPT. It's $20/month. Make sure you're using the best model available, not the default — dig into the model picker. Right now that's GPT-5.2 on ChatGPT or Claude Opus 4.6 on Claude.
Push it into your actual work
Don't just ask it trivia questions. Feed it a real contract. Give it a messy spreadsheet. Paste in your team's quarterly data. Try the thing you spend the most time on and see what happens. If it even partially works today, in six months it'll probably nail it.
Build the adapting habit
The specific tools matter less than getting comfortable with the pace of change. Make experimenting a daily practice. One hour a day, trying something new. Do that for six months and you'll understand what's coming better than 99% of the people around you.
Look at your security posture
As AI gets more capable, the attack surface grows. AI-powered phishing, deepfakes, and automated vulnerability scanning are already here. Make sure your defenses are keeping up with the offense.
Rethink the career advice you give
The standard playbook — good grades, good college, stable professional job — points directly at the roles most exposed to AI disruption. The people most likely to thrive are the ones who are deeply curious, adaptable, and effective at using AI to build things they care about.
The other side of this
It's not all disruption. If you've wanted to build an app, start a business, write a book, or learn something new — the barriers are basically gone. You can describe an app to AI and have a working version in an hour. I'm not making that up. I've done it myself.
The best tutor in the world is available for $20/month. Infinite patience, available 24/7, explains anything at whatever level you need. Knowledge is essentially free now. The tools to build things are nearly free. Whatever you've been putting off because it felt too hard or too expensive — try it.
The people who will come out of this well aren't the ones who mastered one tool. They're the ones who got comfortable with constant change. Who leaned in instead of pulling back.
Where we go from here
Shumer compares this moment to February 2020. I think he's right about the timing but underselling the magnitude. COVID disrupted our routines for a few years. What's happening with AI is going to reshape how work fundamentally operates — permanently.
The gap between what insiders know and what the general public understands is the widest it's ever been. And that gap is dangerous, because it's preventing people from preparing.
If this is your first time seriously thinking about what AI means for your career, you're not too late. But you're later than you think. Start today. The window where being early gives you an advantage won't stay open much longer.
Want help figuring out how AI fits into your business or workflow? Let's talk. We help teams and individuals get ahead of this curve — not with theory, but with practical, hands-on strategy. Check out our AI consulting services or join the community.