Chat GPT 9 is coming!
There is something touching about a five-year business plan. The binding, principally. Somebody has chosen a font and assumed that future productivity will require, say, 10% more of whatever currently pays their salary.
What if it doesn’t?
Looking back to 2023, ChatGPT-3, its main capability was fluent text generation: answering questions, drafting emails, summarising passages, writing simple code, translating, brainstorming, and imitating many writing styles. Its great strength was that it made conversational AI feel useful to ordinary people for the first time. It could produce coherent prose, explain concepts, and act as a general-purpose thinking aid.
Its limitation was depth. GPT-3 could sound confident while being wrong. It had weak reasoning, limited memory, no reliable tool use, and struggled with multi-step tasks where accuracy mattered. It was mainly a language predictor, not a dependable researcher, planner, or agent.
It could write confidently and invent the evidence. We called this “hallucination”. A human employee would likely have been dismissed.
That was then. This is now.
In 2026, the GPT-6 era adds tools and execution. Now the mistake can incur expenses.
GPT-6 is a much more agentic and does complex reasoning and coding, with very large context, tool use, web search, file search, function calling, computer use, and text/image input. Its strengths are integration and execution. GPT-6 is not just a writer; it can analyse files, reason through technical work, use tools, code, search, and help operate software workflows. It is closer to a digital colleague. Its limitations remain important: it can still make mistakes and still requires human judgment for high-stakes decisions. The best use is amplified work, not abdicated thinking.
The next evolution is AI x AI. AI designed by AI. Not by humans.
Suppose we get Chat-9 by 2028. After models 7 and 8, 9 will have been designed by AI. Human beings may approve the result without entirely understanding it.
Nobody on planet Earth, in 2026, knows what actual system’s capabilities or system defects will be operating in Chat-9! Nobody!
Perhaps ChatGPT 9 would represent a later stage where AI becomes less like a chatbot and more like a persistent cognitive system. Its capabilities might include long-term personal memory, robust autonomous research, high-quality scientific reasoning, multimodal perception, real-time voice and video collaboration, reliable software operation, and the ability to coordinate complex projects across days or weeks. To speculate even further, it could act less as a responder and more as an active thinking environment. With initiative!
If so, its strength would be strategic continuity: remembering goals, noticing patterns, challenging assumptions, and managing extended workstreams.
In SOT language, it might help people escape their CVS by continuously generating BVS options.
But its limitations could become much more serious, not less. Trust, alignment, privacy, authorship, dependence, and institutional control would matter enormously. The danger would not only be wrong answers; it would be persuasive systems shaping human attention, judgment, and behaviour.
GPT-9’s central question would be: does it extend human thinking, or replace the discipline of thinking?
This is more than an uncomfortable question concerning your current abilities. If your professional contribution consists of reading documents, rearranging their contents and sounding expensive, you might reconsider the mortgage.
By 2028, your Chat-9 computer may be better informed, more articulate, more productive and considerably more agreeable than you are.
What’s your plan for 2028?
