“A small seed does not look like a tree. The mistake is judging the seed by its size.”
For a long time, productivity has been described as the ability to take one idea and execute it relentlessly. Pick a goal. Make a plan. Build the plan. Avoid distractions.
That works reasonably well when the problem is already understood. But many of the most interesting problems are not like that. At the beginning, we often do not know which idea is worth pursuing, which connection matters, or which apparently insignificant observation will eventually become important.
I have therefore started thinking about intellectual work differently: do not build one giant branch immediately. Plant many small seeds.
The Multi-Seed Theory
The basic idea is simple. Instead of investing enormous effort in one hypothesis, generate many small hypotheses, observations, questions and conceptual connections.
- Plant many small ideas.
- Observe which ones begin to grow.
- Select the strongest ones.
- Develop those into larger branches.
- Repeat the process at the next level.
This is not random brainstorming. It is closer to natural selection. Most seeds will remain seeds. A few will germinate. An even smaller number will become strong branches. The objective is not to force every seed to become a tree.
Why the First Stage Looks Unproductive
This method can feel terribly inefficient in the beginning.
You may spend an hour exploring ten ideas and finish with nothing that looks like a conventional “result”. Someone following a traditional productivity system might conclude that the time was wasted.
But that conclusion confuses visible output with system development.
The early stage is deliberately exploratory. You are discovering the landscape before deciding where to construct the road.
This is also why productive struggle is important. If every difficult question is immediately converted into a polished answer, the learner may acquire the answer without discovering the structure that produced it.
The Branching Method
Once a seed shows promise, it becomes a branch.
But even then, I would not immediately make the branch enormous. Keep it tender. Add a few related ideas. Test them. See which ones connect naturally. Leave weak or redundant branches unfilled.
Think of a knowledge graph as a living tree:
- Seeds: raw observations, questions and intuitions.
- Tender branches: promising connections that need testing.
- Strong branches: ideas supported by repeated reasoning or evidence.
- Trunk: the central framework that emerges after exploration.
- Dead branches: ideas that looked promising but were not worth developing.
The crucial point is that dead branches are not failures. They are information. They prevent future effort from being wasted in the same direction.
Why a Fat Knowledge Graph Can Be Worse
There is a temptation, especially with AI, to create a huge framework immediately. Give the model everything, organise everything, connect everything and produce a giant knowledge graph.
That can create an illusion of understanding.
A large graph built too early contains enormous numbers of weak connections. Redundant concepts receive the same visual importance as powerful ones. The result becomes impressive but difficult to navigate.
A better graph can initially look almost embarrassingly sparse.
It has many little branches pointing in different directions. Some stop. Some merge. Some suddenly become important months later. The structure is allowed to emerge instead of being imposed prematurely.
AI Changes the Economics of Exploration
This is where AI becomes unusually powerful.
Exploration used to be expensive. To investigate an idea, you might need to read several books, search through papers, write notes, calculate examples and discuss the subject with another person.
AI dramatically lowers the cost of the first few steps.
You can throw a small seed into the conversation:
- “Find the hidden pattern.”
- “Challenge the premise.”
- “Think from first principles.”
- “Give me only one line.”
- “Don’t brute-force it.”
These are not merely prompts. Over time they become embedded thinking commands.
The interesting part of human-AI collaboration is that the learning goes both ways. I taught the AI how I tend to think by repeatedly using discovery-oriented instructions. In return, the AI exposed me to ways of interrogating my own ideas through prompts and alternative structures.
It becomes a feedback loop: human intuition → AI exploration → human selection → AI expansion → human correction.
Learning Through Discovery
This resembles discovery learning more than conventional instruction.
Instead of beginning with a fully constructed tree of knowledge, you begin with a seed. The learner has to retrieve, compare, question and connect concepts.
That process is slower initially. But something important happens: the commands for thinking become part of the learner’s own cognitive toolkit.
You stop merely knowing an answer. You learn how to ask the next question.
That distinction becomes enormous over time.
From Productive Struggle to Productive Compounding
There is another interesting consequence. The method can look inefficient for months and then suddenly become extremely productive.
Why?
Because every successful branch becomes a new starting point.
Suppose ten seeds are planted. Two become useful branches. Each of those branches generates five more questions. Three of those questions become strong ideas. Each strong idea produces another set of connections.
The number of possible intellectual paths grows much faster than the original number of seeds.
This is branching compounding.
At first you are planting. Later you are harvesting branches that were created by previous branches.
My Own Experience With the Method
This is increasingly visible in my own work.
Many discussions begin with what looks like a tiny thought: an observation about AI, education, markets, human behaviour or technology. The first discussion may produce nothing more than a rough framework or even an infographic.
That is not necessarily the final product.
The infographic becomes a map. A later discussion adds a branch. Another conversation challenges an assumption. A third connects it with something apparently unrelated.
Eventually the small seeds start forming a coherent structure.
The surprising result is that what initially produced very little conventional output can later support two high-quality articles in a day.
The productivity did not suddenly appear. The branches had been accumulating.
AI Removes the Scaffolding Bottleneck
There is another bottleneck that AI is beginning to remove: the distance between having an important idea and being able to communicate it effectively.
Consider a scientist who is brilliant at observation and experimentation but is a poor writer. His most productive work may be thinking, designing experiments, collecting evidence and challenging his own hypotheses. Yet enormous amounts of time can disappear into manuscript scaffolding.
- Turning rough notes into formal prose.
- Organising literature and citations.
- Preparing figures and explanations.
- Adapting manuscripts to journal requirements.
- Formatting references and supplementary material.
- Preparing versions for different audiences.
AI can increasingly become the scaffolding layer around the scientist.
The scientist can concentrate on the high-value loop: think → experiment → observe → hypothesize → test → refine.
AI can then help turn validated intellectual material into the required representations: a technical paper, mathematical explanation, infographic, presentation, lecture, research proposal or plain-language explanation.
The scientist still has to verify the evidence and take epistemic responsibility. AI should not decide whether an experiment is valid merely because it can write a convincing paragraph about it.
But the communication bottleneck becomes much smaller.
This may allow intellectual contribution to become less dependent on writing ability. A great scientist who is a poor writer can increasingly get the idea across without spending his most productive hours becoming a professional manuscript formatter.
In that sense, AI is not simply writing papers. It is separating scientific thinking from scientific paperwork.
Multi-Seed Theory Is Also an Investment Strategy
The analogy with investing is strong.
An investor does not need every small investment to become a winner. The purpose of diversification is to expose the portfolio to multiple possibilities while limiting the cost of being wrong.
Ideas can be treated similarly.
- Small initial investment.
- Large number of possibilities.
- Rapid rejection of weak ideas.
- Increasing investment in winners.
- Repeated reinvestment into successful branches.
The key is asymmetry. A failed seed should cost very little. A successful seed should have room to grow enormously.
The AI-Era Knowledge Graph
This may become one of the most useful ways of working with AI.
Instead of asking AI to manufacture a giant finished product immediately, use it as an exploration engine around a growing personal knowledge graph.
Do not ask it to fill every branch.
Ask it to help you discover which branches deserve filling.
That distinction may be more important than prompt engineering itself.
The Future: More Seeds, Better Selection
AI will make the cost of generating and testing ideas increasingly small. That means the scarce resource will shift.
The bottleneck may no longer be idea generation. It may be selection.
Which seed deserves another hour? Which deserves another year? Which connection is meaningful? Which is merely an attractive pattern?
Human judgement therefore remains important. AI can explore an enormous number of branches, but someone still has to decide which tree is worth growing.
The future intellectual worker may therefore look less like a person executing one enormous plan and more like a gardener.
Plant widely. Observe carefully. Prune ruthlessly. Feed the promising branches. Let the structure emerge.
Do not demand a tree from every seed.
Some seeds are supposed to remain seeds.
“The advantage is not having the biggest tree. It is having planted enough seeds to discover which tree was worth growing.”
Related article
How AI Is Making the Lazy Brain More Productive
You can read the full article on Personal Touch Academy in the link below.
How Should Students Learn in the AI Era? From Answers to Independent Thinking
Future Trends | Life Hacks | Love Lessons 