- AI with Alec Newsletter
- Posts
- AIWA "The One Thing" #25: Karpathy Named the 3rd AI Paradigm. Your AI Employee Just Showed Up For Work. They're Ready to Go. Are You?
AIWA "The One Thing" #25: Karpathy Named the 3rd AI Paradigm. Your AI Employee Just Showed Up For Work. They're Ready to Go. Are You?
You remember the first time you used ChatGPT? Paradigm 1, the LLM as a website you visit.
You remember the first time you used an AI enabled application? Paradigm 2, the LLM inside your favorite application.
You remember the first time you worked with an AI employee? You’re about to…
Paradigm 3, a self-contained, persistent, asynchronous entity with org-wide context working alongside humans. Not a tool you use. A colleague who shows up and gets after it. With and for you.
Last week Claude Tag launched and my man Andrej Karpathy absolutely nailed it when he referred to it as the “3rd major redesign of LLM UIUX” but the most interesting thing he said had nothing to do with the interface.
When Karpathy calls something a paradigm shift, the enterprise AI world listens.
He called it an ‘org-level harness.’ That’s the phrase worth sitting with.
Because what makes the harness work has nothing to do with the interface.
It has to do with the contextual architecture aka the below the waterline work that determines whether Claude Tag, your new AI employee, is an intern or a digital twin of one of your biggest rockstars.
Claude Tag in Slack without a context graph is a super smart AI employee with all the horsepower in the world but no onboarding, no institutional knowledge around how things actually get done.
Marc Benioff called his shot in April. Salesforce going headless.
AI agents as primary users. Humans secondary.
At the time, AI Twitter filed it under vision and a bold strategy.
You know who owns Slack? Salesforce does. That’s not irony. That’s the market confirming the thesis.
The enterprises pulling ahead aren’t just welcoming a new AI employee. They’re doing the below the waterline work that gives it something real to work with.
Institutional knowledge. Decision traces. Context graphs.
It’s not the stuff that necessarily gets you on CNBC but it’s most definitely the stuff that your CFO is looking for.
It’s the difference between a smart intern and a force multiplier.
That’s the difference between Augment and Rewire.
Humans + Machines. Never Humans vs. Machines.