Have you spotted GPT‑6.2 Astra in ChatGPT’s Work mode yet? OpenAI released it yesterday, 17 November. help.openai.com

From 23 April to 17 November, OpenAI introduced five GPT model families in 208 days: GPT‑5.5, 5.6, 6, 6.1 and 6.2. In September, just seven days separated the releases of GPT‑6 Sol and GPT‑6.1 Sol. openai.com

If you’re thinking, “I’m still getting used to the last one. Do I need to learn this one too?” I understand the feeling.

As an AI strategist and corporate trainer, what I wanted to know was how much of the work behind that pace was already in place before launch day.

I looked through OpenAI’s public reports. On 26 June, when the company previewed GPT‑5.6, it explained who would get access first: a small group of partners who could try it while testing continued. On 9 July, GPT‑5.6 Sol began rolling out to eligible paid ChatGPT users. Yesterday’s GPT‑6.2 report offers another clue. The team built on evaluation and monitoring practices it had already developed for GPT‑6 Astra. deploymentsafety.openai.com

That’s where I see foundation and strategy at work. A foundation gives the next team something it can build on. Strategy guides decisions about who goes first, what needs testing and when to expand. The dates alone can’t tell us how far ahead OpenAI planned all five releases. But they prompted a question I’d ask any business owner: Before we spend time and money on AI, which part of our business do we want to improve?

Say you receive 20 enquiries. You follow up on WhatsApp and hardly anyone replies. A newer model could help you write all 20 messages faster. But look at the first enquiry. The customer asked, “How much does it cost?” Your reply opens with three paragraphs about your company. Even if AI writes a beautiful introduction, the customer still has to ask, “So, how much?”

I’d start with those 20 conversations. Where did people stop replying? Try using AI to draft an opening line that answers what the customer actually asked. Then see how many people reply or agree to a five-minute call. When the next model comes along, try it on the same task. Now you have a way to judge whether it earns its place in your business.

That’s why I tell my participants: Build the foundation and get clear on your strategy before you go deeper into AI.

It’s easy to stay busy trying every tool, lose sight of which ones help, then get back to your desk unsure where to start. By the next launch, you already feel behind.

My way of staying grounded is to ask myself one question whenever a new model arrives: “What real problem on my desk today is worth testing this on?”

掌握规律,锁定胜局。
Master the patterns. Secure the win.

Vince Tay 郑珺午
AI Strategist & Corporate Trainer