OpenAI Teases Christmas Gifts as Enterprise Becomes Top Priority

The Context Revolution: Why Intelligence Is No Longer Enough Let me walk you through what might be the most important shift happening in AI right now, and it comes from tech investor Gavin Baker in a conversation on the "Invest Like The Best" podcast. Baker's core thesis is simple but profound: we've hit diminishing returns on raw intelligence, and the next frontier isn't about making AI smarter, it's about making it useful. Here's the reality check.

Unless you're asking deep questions about semiconductor physics or advanced mathematics, it's getting genuinely hard to tell the difference between the top AI models. They're all smart enough for most tasks. The bottleneck isn't IQ anymore, it's context, reliability, and the ability to handle complex, multi-step workflows.

Baker identifies three specific building blocks that will separate useful AI agents from glorified chatbots. First is massive context windows. We're talking about AI that doesn't just know you like beaches, but remembers that you follow Andrew Huberman so you need an east-facing balcony for morning sunlight, and that you refuse to fly on planes without Starlink.

That level of personalization requires holding enormous amounts of information in active memory. Second is reliability. The model can't hallucinate a flight time or book the wrong hotel.

It needs to be boringly, consistently accurate. This is harder than it sounds because current models still make confident mistakes. Third is task length.

We're moving from "make me a dinner reservation" which saves you five minutes, to "plan this entire two-week vacation for my extended family across three countries" which saves you five hours or more. That's where real ROI lives. Baker predicts that as context windows expand to millions of tokens, AI will eventually hold every Slack message, email, and company document you've ever written, turning it into the ultimate chief of staff who actually knows your business.

This shift has massive implications for SaaS companies, and Baker doesn't mince words. These companies face what he calls a "lesser of two evils" bet. They can deploy AI agents that automate workflows, which will compress their margins significantly.

Or they can avoid agents and risk total business extinction as venture-funded upstarts outspend them and deliver ten times the value at half the price. Think about what this means. The entire SaaS business model was built on recurring revenue from humans doing repetitive tasks in software.

Now AI agents can do many of those tasks autonomously. Companies like Salesforce, ServiceNow, and Workday aren't just facing new competitors, they're facing a fundamental restructuring of how work gets done. OpenEvidence provides a glimpse of what's possible.

They've gone from zero to one hundred fifty million in annualized revenue in roughly two years by giving doctors an AI assistant that can synthesize medical research. That's not just fast growth, that's market creation. And they're raising at a twelve billion dollar valuation because investors see this pattern repeating across every knowledge-work vertical.

The winners will be companies that build context moats. If your AI knows everything about a customer's business, switching costs become enormous. That's why OpenAI is pivoting hard to enterprise in twenty twenty-six.

Consumer AI is cool, but enterprise AI is where the margins live.

The competitive dynamics here get fascinating. Google, OpenAI, and Anthropic are all racing to expand context windows and improve reliability. But the real disruption isn't just about model capabilities, it's about integration depth.

The AI that lives inside your actual workflow, that has access to your calendar, email, documents, and communication tools, will beat the standalone chatbot every single time. This explains why Microsoft's Copilot strategy is so aggressive. They're not just selling AI features, they're embedding intelligence into the tools people already use every day.

Same with Notion, which has quietly become an AI-first workspace. These companies are building context moats by being where the work actually happens. Meanwhile, traditional software companies are scrambling.

Salesforce is bolting AI onto everything. SAP is trying to modernize fast enough to stay relevant. The companies that survive will be the ones that transform from "software you use" to "agents that work for you.

" That's not a feature upgrade, that's a business model revolution. And then there's the hardware angle. Baker also discusses space data centers, which sounds wild until you realize that power and cooling are becoming the limiting factors for AI training.

If someone figures out how to put compute in orbit where you have unlimited solar power and infinite cooling from space, the cost curve for AI could drop dramatically.

Here's where it gets uncomfortable. If AI can handle five-hour tasks instead of five-minute tasks, what happens to entire categories of knowledge work? Baker's vision of AI as a "chief of staff" that knows everything about you and your business isn't just convenient, it's existentially weird.

We're talking about a technology that will know you better than your spouse, your therapist, or your best friend. It will remember every email you've ever sent, every document you've ever written, every decision you've ever made. That creates both incredible utility and profound privacy questions.

And then there's the education crisis. Anand Sanwal's piece on AI cheating becoming normalized hits hard here. If AI can write your essays, solve your math problems, and complete your projects, what exactly are we testing for anymore?

Schools can't out-police ubiquitous AI, so assessment itself has to fundamentally change. Maybe we stop testing knowledge recall and start testing judgment, creativity, and ethical reasoning. There's also a weird status shift happening.

Being "smart" used to be the ultimate competitive advantage. Now it's becoming commoditized. The new scarce resource is taste, judgment, and the ability to ask the right questions.

Every's analysis of Claude Opus four point five makes this point brilliantly: the bottleneck is no longer your coding ability, it's your product vision.

So what should you actually do with this information? Three specific moves. First, start building your context moat now.

Whether you're running a product team or an entire company, the AI that has the deepest integration with your workflows and data will win. That means investing in proper data infrastructure, clean documentation, and systems that let AI actually learn from your operations. Don't just bolt ChatGPT onto your product, design workflows where AI has genuine context about what users are trying to accomplish.

Second, rethink your competitive strategy around agents, not features. If you're a SaaS company, you need a plan for how agents will impact your business in the next eighteen months. Will you build agents that automate your own product?

Will you become a platform that other agents integrate with? Will you double down on the human judgment layer that AI can't replicate? Pick a lane, because sitting still is choosing extinction.

Third, experiment with task-length expansion immediately. Stop thinking about AI as something that answers questions or generates content. Start thinking about it as something that completes projects.

Take a workflow that currently takes your team five hours and give AI a shot at doing it end-to-end. You'll quickly find the gaps between hype and reality, and those gaps are where your next product features or process improvements live. The companies that learn fastest how to hand off entire workflows to AI will have an eighteen-month advantage over everyone else.

Baker's thesis isn't about some distant future. This transition from intelligence to usefulness is happening right now. OpenAI shipping Sora for Android in twenty-eight days with a four-person team isn't just a cool case study, it's a preview of what small teams can accomplish when AI handles the grunt work.

The question isn't whether this shift is coming, it's whether you'll be ready when it arrives.