Before Working Harder, Ask Whether You Are Solving the Right Problem
Tomohiro Iida ·
First-principles thinking has value for everyone. But that does not mean anyone can master it easily.
Work is piling up, so we try to work faster. Sales have stalled, so we increase outreach. Learning is difficult, so we tell ourselves to be more disciplined. An AI makes a mistake, so we add even more instructions. Improving an existing approach often makes sense. But what happens when we are trying to solve the wrong problem?
Making unnecessary work more efficient still leaves us doing unnecessary work. Doubling an ineffective activity may merely double the effort. Sometimes changing the direction of our effort matters more than increasing its amount. First-principles thinking helps us ask what actually needs to be solved.
At Netsujo, this approach has helped us reconsider decisions in website improvement, AI-agent authority, and our own engineering workflow. It requires knowledge, observation, a willingness to revisit past choices, and evidence from the real world.
What is first-principles thinking?
In philosophy, a first principle is a fundamental starting point. In practical problem-solving, we use the idea to identify the desired outcome, separate facts from assumptions, understand real constraints, and reconstruct possible solutions without treating prevailing habits as immutable rules.
Consider a meeting that takes too long. The obvious response is to shorten it from an hour to half an hour. A more fundamental question is: What is the meeting for? If it only distributes information, a written update might work. If it makes a decision, perhaps only the decision-maker and relevant facts are needed. If it resolves a disagreement, adequate conversation time may be essential. The aim becomes choosing the right form of coordination, not simply cutting minutes.
Three reasons this way of thinking is useful
1. It reduces effort spent on the wrong problem
Suppose a business wants more enquiries and decides to buy more advertising. The team starts optimizing copy and targeting. But weak enquiries might originate in an unclear value proposition or a confusing contact form. Before spending more, we need evidence about where prospective customers get stuck. Asking why the proposed remedy is necessary helps direct scarce time and money toward the actual constraint.
2. It reveals options hidden by habit
We may assume a meeting, a specialized subscription, or a manual process is required simply because that is how the work has always been done. Some conditions are genuine obligations: law, safety, consent, budget or contracts. Others reflect a previous technology or organizational habit. Separating the two opens options such as stopping a task, simplifying a workflow, reusing existing infrastructure, or buying a service instead of building it.
3. It develops independent judgment and learning
Rather than copying a supposed best practice, we define a purpose, check the evidence, form a hypothesis, test it and update our view. Unexpected results are informative. They reveal which assumption, constraint or predicted outcome needs revision. The goal is not to be right on the first attempt; it is to make better decisions as evidence accumulates.
Where it helps in ordinary work and learning
If a team works late every day, telling everyone to work faster may miss duplicate data entry or slow approvals. Observing where time is spent and removing the bottleneck may produce a larger improvement than asking individuals to accelerate.
Language learning offers a similar example. If the goal is to discuss business with overseas clients, doubling study hours may be less useful than practicing how to understand questions, clarify uncertainty and express a proposal. Vocabulary and grammar still matter. The learning method should follow the ability we actually want to acquire.
In both cases, the practical benefit is the same: before adding effort, check whether the effort is pointed at the right outcome.
Four examples from our own development work
We did not call every decision first-principles thinking at the time. For some of these examples, that interpretation emerged retrospectively as we reviewed what changed the direction of the work.
1. From SEO improvement to the role of a business website
When improving our own sites, it was easy to focus on rankings, traffic and page speed. We instead asked what a B2B website must accomplish. A potential customer needs to find a company, understand its services, see trustworthy evidence and take an appropriate next step. Search visibility is only one part of that journey.
That led us to eight diagnostic areas: presence, discovery, retrieval, trust, understanding, action, measurement and improvement. They became part of how we designed Netsujo SIGNAL as a website sales foundation. We changed the question from how to raise a technical metric to what must improve for a customer to make an informed decision. Product implementation is real; attributing a particular increase in client sales to this model still requires outcome evidence.
How Netsujo SIGNAL grew out of improving our own websites
2. Do we need the external service, or the value it provides?
We evaluated GSC Wizard, a service for analyzing Google search data. Before continuing the subscription, we asked whether our essential need was that specific product or accurate Google data and a reliable improvement decision. We found that existing Google authentication, official APIs and our own storage and analysis capabilities could support an initial internal replacement.
On October 17, 2026, we accepted a bounded technical migration for our two sites, netsujo.jp and miyakodeit.com: direct data collection and storage worked, and reports remained readable with the external service disconnected. This establishes the initial technical result. It does not by itself prove cancellation of a paid contract or a particular amount of savings. Internal maintenance also has a cost, so build-versus-buy should be judged on total cost and quality.
Source: technical acceptance for the GSC Wizard migration
3. Does assigning an AI role grant permission to perform an action?
Giving an AI agent the title "report writer" does not authorize it to write into every directory. Role descriptions and operational authority are different things. That distinction motivated Agent Role Contracts (ARC), our open-source work on machine-checkable role, scope and handoff declarations.
In a CrewAI configuration adapter experiment, a permitted output path passed validation. Changing only the path to an unauthorized target caused validation to fail. Restoring the original path produced another pass. This was a configuration-level test, not a claim of complete runtime enforcement. Reframing the problem moved us from writing increasingly detailed instructions toward validating explicit boundaries.
Source: CrewAI adapter experiment and PASS–FAIL–PASS evidence
4. If an AI keeps fixing the same class of defect, should it simply keep trying?
One development effort accumulated 50 commits while independent reviews continued to find serious problems in the same family of boundaries. Each local fix exposed another related flaw. We asked whether the problem lay in the quality of each patch or in the underlying abstraction itself.
That led to a design for recognizing repeated strategy failure and reassessing the architectural boundary instead of generating another sibling patch. Parts of this mechanism still require implementation and operational acceptance. The broader lesson is relevant well beyond code: when failures repeat, examine the structure producing them.
Source: redesigning the repeated-review recovery strategy
Why first-principles thinking is difficult
Our own assumptions are hard to notice
Experience makes decisions faster, but it also turns previously useful solutions into habits. Re-examining our own successes can be harder than questioning someone else’s practices.
Distinguishing facts from assumptions requires expertise
A medical procedure or building safety standard may embody specialist knowledge and lessons from past failures. We need to understand why a constraint exists before challenging it. Knowing what remains unknown is a strength, not an inconvenience.
It takes courage to reconsider sunk costs
An expensive system may not fit the actual goal. A project we have defended for years may no longer be necessary. Resources already spent and the next decision to make must be evaluated separately.
A plausible idea still needs a real-world test
An elegant explanation may fail in an actual organization or with real users. The value of a new approach must be tested against alternatives, measured outcomes and the burden of maintaining it. Novelty is not evidence of benefit.
Five questions to put it into practice
- What outcome do I actually want? Separate an end goal, such as faster decisions, from a means, such as shorter meetings.
- What do I know for certain? Separate observations, second-hand reports, hypotheses and unknowns.
- Which constraints are real? Identify obligations involving safety, law, permissions, budget and commitments; distinguish them from convention.
- What would I try without the current method as a starting assumption? Compare improving, reducing, stopping, replacing and retaining the existing method.
- How will I test the result? Run a bounded, reversible experiment and decide in advance what evidence would support or reject the idea.
For a weekly meeting, try one week with a written update and a shorter decision-focused conversation. Compare decision time, misunderstandings and participants’ workload. Keep the change if it helps; revise or reverse it if the evidence suggests otherwise. Small experiments make a demanding way of thinking more practical.
You do not need to start from first principles every time
Reconstructing every routine from scratch would waste time. This approach is especially useful when effort repeatedly fails to produce results, the same problem returns, circumstances change, or a decision carries substantial cost and risk. When a reliable existing method already works well, keeping it may be the best option. Even the decision to use first-principles thinking deserves a cost-benefit check.
Questioning the familiar can reveal new possibilities
At Netsujo, our purpose is to open up human potential through technology. That requires us to ask what a person is trying to accomplish, what has kept the problem unresolved, what constraints are genuine, and what a different combination of existing knowledge or technology could make possible.
First-principles thinking is hard. It takes humility, investigation and experiments that may fail. Yet it gives each of us another option beyond following a method simply because it is familiar: we can reconsider the problem and choose deliberately.
If you are working hard without seeing progress, consider asking one question before you increase your effort: What am I really trying to achieve?
That may be where a previously unseen option begins.
Further reading: first-principles engineering and value discovery at Netsujo