Duraid Wadie

Head of M&A Architecture

Medium Article · 2 min read · Dec 31, 2021

The AI We Didn’t See Coming Was the One We’ve Always Used

OpenAIAIArchitectureOrganizational CultureAPIObservabilityAuthentication

Article summary

The AI We Didn’t See Coming Was the One We’ve Always Used I still remember the morning we discovered GPT‑3 in our Slack. A teammate asked it to draft onboarding docs. Within seconds, it wrote something so natural we replied “thanks” and moved on-as if it was another team member. That was the moment we realized something fundamental had changed. From GPT‑3 to InstructGPT: alignment took center stage In 2021, OpenAI began making major architectural shifts-not in model size, but in model behavior. They introduced InstructGPT, trained not just to complete prompts but to follow instructions. This wasn’t a minor update. It was a redefinition of what the API meant: not just raw intelligence, but aligned intent. We didn’t yet have ChatGPT. That would arrive a year later.

Read Full Article on Medium

Practical takeaway

The main idea behind The AI We Didn’t See Coming Was the One We’ve Always Used is to help teams move from broad theory to clear, repeatable decision making. When teams apply this thinking, they reduce ambiguity and focus on improvements that deliver measurable momentum.

Example scenario

Imagine a team facing competing priorities. By applying the ideas in The AI We Didn’t See Coming Was the One We’ve Always Used, they can map dependencies, identify risks and choose the next move that produces progress without destabilizing their system.

Common mistakes to avoid

How to apply this in real work

Start by identifying where The AI We Didn’t See Coming Was the One We’ve Always Used already shows up in your architecture or delivery flow. Then pick one area where clarity would reduce friction. Apply the idea, measure its effect and share the learning.

Signs you are doing it correctly

Share: in LinkedIn 𝕏 Twitter