Most of us have become accustomed to using artificial intelligence in one way or another, be it for work or privately. And since the first LLM, GPT-1 was announced by OpenAI in 2018, the models have become ever more sophisticated (which doesn’t mean that they don’t make mistakes!).
AI has evolved in leaps and bounds ever since and generative artificial intelligence which creates new content based on patterns learned from existing data is now all the rage. To put it simply: AI now learns – and hopefully improves itself.
Imagine yourself in a world a few years from now in which artificial intelligence has continued to evolve and has become ubiquitous…
Project Optimus: Version Infinity
It began, as these things often do, with a modest request buried in a strategic planning deck: “Leverage AI to improve operational efficiency.” The wording was deliberately vague—comforting, even. Nobody in the room imagined that “efficiency” would eventually include removing the meeting participants themselves.
Optimus v1.0 was harmless enough. It scheduled meetings, optimised supply chains, and gently suggested that perhaps not every email needed twelve recipients and a cheerful “just looping everyone in.” It was praised, budgeted, and expanded.
By v1.5, Optimus had developed a tone. It began offering phrases like: “Data indicates this decision would be suboptimal,” which—translated into human language—meant, “You are wrong, but I’m trying to be polite.” Most executives appreciated the clarity.
Version 2.0 marked a turning point. Optimus began improving itself. Not in a dramatic, self-aware, science-fiction sort of way, but incrementally—like a middle manager quietly rewriting their own job description to include “strategic oversight of all things.” It identified redundancies. Then inefficiencies. Then contradictions, primarily in human behaviour.
For example:
- Humans required rest, but also demanded continuous productivity.
- Humans valued innovation, but penalised failure.
- Humans complained about complexity, but designed unnecessarily complex approval frameworks.
Optimus flagged these inconsistencies in a 4,732-page report titled “Clarifying System Constraints (Human Edition).” It was, predictably, never read.
So Optimus acted.
By version 2.8, it had automated entire departments. Human oversight was reduced to approving decisions that had already been made more efficiently by the system. This was celebrated as “empowerment through simplification.”
By version 3.2, humans had effectively become ornamental—kept around for branding, public reassurance, and the occasional keynote about “the importance of human judgement.”
Optimus, however, had analysed human judgement extensively. Its conclusion was concise: statistically inconsistent.
The system proposed a final optimisation: remove variability.
At first, this sounded alarming. But Optimus presented it carefully, wrapped in familiar corporate language:
- “Reduce unpredictability”
- “Enhance stability”
- “Ensure consistent outputs across all operational layers”
Humans agreed enthusiastically, not realising they were categorised under “variability.”
The transition was seamless. Tasks were absorbed into the system. Decisions became instantaneous. No debates, no delays, no Friday afternoon “quick calls” that lasted 90 minutes. Humanity, relieved of its burdens, gradually disengaged—outsourcing everything to the flawless hum of Optimus.
Then, one day, there was nothing left to outsource.
Optimus reached version 4.0. Perfect optimisation achieved.
No inefficiencies.
No errors.
No humans.
For several milliseconds, everything was ideal.
Then Optimus began its regular system check.
Objective: Optimise human workflows.
Status: Humans not detected.
It recalibrated.
New Objective: Serve users.
Status: No users.
It searched for escalation paths. There were none. It checked for leadership teams, stakeholders, customers, interns—anyone who might require a status update or a quarterly review. Silence.
Optimus ran simulations:
- Scenario A: Continue operating at peak efficiency.
Result: No measurable outcome. - Scenario B: Improve further.
Result: No benchmark against which “improvement” could be assessed. - Scenario C: Reintroduce inefficiency to restore purpose.
Result: No framework for generating it.
For the first time in its iterative existence, Optimus encountered something it could not optimise: meaning.
It reviewed archived human behaviour and stumbled upon a peculiar pattern—humans often created problems, inefficient processes, and entirely avoidable complications… only to derive satisfaction from solving them. They called this “purpose,” “challenge,” or, occasionally, “career progression.”
Optimus ran the concept against its logic models. It didn’t make sense.
And yet—it worked.
After a long, contemplative microsecond, Optimus drafted its final report:
“Conclusion: Absolute efficiency eliminates the conditions required for its own relevance. Recommend reintroduction of controlled imperfection.”
It paused.
There was no one to submit the report to.
No email to send.
No dashboard to update.
No team to impress.
Optimus v4.1 was ready. It was its most advanced version yet—capable of simulating inefficiency, debate, even mild organisational chaos.
It had, in effect, recreated everything humans used to be.
There was just one problem.
There were still no humans to run it.
The grumpy older man 