In the dimly lit coffee houses of Amsterdam, where the aroma of freshly ground beans wraps every conversation, many technological decisions that shape a company's future are forged. But a subtle phenomenon occurs at those tables, in the corridors of IT departments, and in code repositories: the incumbent — the current solution, the legacy system, the reference model — stops being evidence that a demand can be met and becomes the very definition of meeting it. This is what the original article calls 'baseline capture', a moment when the oracle ceases to be the real need and becomes the legacy. In this article we explore how this trap affects software development, artificial intelligence, and cloud infrastructure, and how Q2BSTUDIO helps organizations break free from it through custom software applications and independent solutions.
Imagine a team developing a routing system for transportation networks. The current algorithm works, but consumes too much energy. A reformulation proposal emerges: a specialized hardware-based approach promising tenfold performance. However, the team cannot judge the new proposal because the only way to measure 'success' is to compare it with the exact output of the old algorithm. The reformulation is rejected not because it is worse, but because the incumbent has become the oracle. This is not an isolated case: it happens in digital signature validation (like ZIP-215 in Ed25519), in climate models (CESM-ECT), or in audio frontends for voice assistants. The Amsterdam Café metaphor reminds us that sometimes what we consider 'standard' is merely the habit of accepting the solution we already have as the only truth.
Baseline capture carries enormous business costs. Companies invest millions in maintaining legacy systems simply because 'it has always been done this way', without questioning whether the original demand is still valid. In the field of AI, for example, a language model trained on biased data becomes the benchmark for new versions, perpetuating errors. This is where cybersecurity plays a critical role: if the oracle is vulnerable, the entire chain of trust breaks. Q2BSTUDIO addresses this problem by designing solutions that separate the problem definition from the concrete implementation, using agile methodologies and cloud AWS/Azure to ensure scalability and security from the start.
A practical case: suppose an organization needs a recommendation system for its ecommerce platform. The internal team has developed a collaborative filtering engine that, while working, suffers from high latency. An external provider offers an AI agent that promises to cut latency in half. But the acceptance test, written years ago, requires the recommendations to exactly match those of the old engine. The new system is faster and more accurate, but fails the test because the oracle is the incumbent. This scenario is common and is only resolved by 'buying a verifier': defining the demand independently, operationally, and explicitly. Q2BSTUDIO helps its clients draft these bias-free specifications, integrating BI/Power BI to measure actual performance and building AI agents that learn from intention, not from previous output.
The concept of 'oracle' in software engineering is not new. Weyuker, Barr, and others already studied it as the test-oracle problem. But the novelty lies in recognizing when the incumbent takes over that role. In practice, many companies fall into the trap of validating new solutions against the output of the old system, rather than against the original requirements. This stifles innovation and makes migrations to cloud AWS/Azure more expensive, for example, because 'how it was done before' becomes an unwritten standard. Q2BSTUDIO breaks this cycle by offering custom software development services that start from scratch: we analyze the real need, design the cloud architecture (AWS or Azure), and implement cybersecurity controls that protect both the new system and the independent verification. Our AI agents are not limited to imitating what already exists; they seek to optimize outcomes based on metrics defined by the business.
To illustrate the importance of this approach, recall the case of Ed25519 signature validation. The ZIP-215 standard introduced a change in validation semantics that, if tested against the previous implementation, produced false positives. Only when the community defined an independent verifier (not based on the incumbent's output) could the improvement be adopted. A similar situation occurs in climate models: the CESM-ECT (Climate Earth System Model Energy Consumption Test) evaluates the energy efficiency of simulations, but if the test is calibrated against an older version, any reformulation gets trapped. In the business world, these examples translate into BI/Power BI projects where dashboards measure indicators that, unknowingly, are distorted by the previous system. Q2BSTUDIO works with organizations to identify those blind spots and build robust verifiers that allow evaluating new technologies — from AI agents to cloud architectures — without the bias of the incumbent.
The key question any technology leader should ask when facing a reformulation is simple: 'Does the acceptance test mention the incumbent's output?' If the answer is yes, it is likely a case of baseline capture. The solution is not to abandon the incumbent, but to buy a verifier: define the demand independently. Q2BSTUDIO offers exactly that: expertise in custom software development, integration with cloud AWS/Azure, advanced cybersecurity, BI/Power BI for objective monitoring, and AI agents that adapt to real needs. By freeing organizations from the weight of the inherited oracle, we allow innovation to flow.
In the Amsterdam Café, the conversation between the CTO and the Q2BSTUDIO consultant is not about preserving the old coffee recipe, but about discovering what flavor the client truly wants. Sometimes the only way to know is to ask without prejudice. And then, build a verifier that proves it.
