Most consultancies that sell knowledge work — and most software companies that try to encode it — operate on vibes.
We mean that precisely. A vibe-driven engagement looks confident, decisive, expert. It produces deliverables. The deliverables sound right. Six months later nobody can explain how the deliverable was reached, why a different approach was rejected, or what would have to change for the recommendation to flip.
The vibe was the deliverable. The reasoning was a side-effect.
This is the dominant operating mode in professional services and in most "AI-augmented" software. It is fast, it is fluent, and it is structurally fragile. The fragility does not show up until something forces a re-examination — a compliance audit, a senior departure, a regulator's question about how a recommendation was actually derived. Then everyone tries to reconstruct the reasoning from notes that were never written down. The vibe does not survive scrutiny.
What systems do differently
A system, in our usage, is the smallest set of explicit components that makes a piece of work auditable end-to-end. It is not a "process" in the corporate sense (steps and approvals); it is the structural commitment to write down what would otherwise be assumed.
An idea is a vibe. An idea written down with its sources, its reviewers, its rejected alternatives, and its rollback plan is a system.
A recommendation is a vibe. A recommendation that traces backwards to the institutional knowledge that produced it and forwards to the metric that will tell you if it was right is a system.
A piece of code is a vibe. A piece of code linked to a work item linked to a source document linked to a deployment metric is a system.
Vibes feel decisive. Systems are decisive. The difference is whether you can defend the decision a year later when the people who made it are gone.
Boyd's OODA loop — observe, orient, decide, act — is a useful frame, but it gets misread. Most readers focus on the speed of the loop. The deeper insight is that the loop has to close. You have to feed the result of your action back into your orientation, or you are running fast on stale assumptions. Without closure, fast is just confidently wrong, sooner.
Vibes-driven decision-making is a fast OODA loop with stale orientation. It feels decisive. It is losing.
What this looks like in practice
Three concrete examples.
An engineering decision. A team adopts a new architecture pattern. The vibe version: someone read a blog post, the architect liked the diagram, the team rolled it out. Six months later the pattern is causing pain and nobody can articulate which problem it was supposed to solve. The system version: every architectural commitment is documented with the source, the alternatives considered, the rollback plan, and the metric that will indicate success. Six months later the team can answer "is this still the right answer?" without re-litigating the original decision.
A consulting recommendation. A firm advises a client to adopt a particular GRC framework. The vibe version: the senior partner has had good experiences with that framework before. The system version: the recommendation is traceable to the client's specific risk profile, the framework's compliance coverage, the implementation cost model, the projected outcomes, and prior engagements where the same recommendation worked or failed. The recommendation can be defended. It can also be revisited when the underlying inputs change.
An AI-generated answer. A team uses an LLM to draft a response to a customer question. The vibe version: the LLM produces something that sounds right, the team ships it. The system version: the LLM is constrained to draw from a specific corpus of validated institutional knowledge; the draft is reviewed against the source; the response is logged with its sources and outcome; subsequent questions on similar topics see the previous response as a starting point. The same LLM, the same team, two completely different epistemic regimes.
The difference is not the technology. The difference is whether the work is structurally reviewable.
Why most AI rollouts produce vibes, not systems
Because the technology is fluent. LLMs produce text that reads like systematic thinking even when the underlying process is undocumented and unreviewable. The cosmetic similarity to expert reasoning hides the structural absence of expert review.
Most "AI-augmented" knowledge work today is the typewriter version of intelligence work. The output looks the part. The process is invisible. The next person to handle the same question starts from scratch, with no access to the reasoning that went into the previous answer.
This is the typewriter trap we have written about before. It is also the dominant failure mode of AI rollouts in regulated industries. The model is fast. The model is confident. The model is also generating outputs that nobody downstream can audit, and the regulator's first question is "show me the reasoning." There is no reasoning to show.
What ORCA does about it
ORCA is opinionated about this distinction. The product enforces it.
Every piece of knowledge that enters the system is sourced. Every recommendation is traceable. Every change is reviewable. Every action is reversible. These are not features we are adding; they are the operating discipline the product is built around.
We make this concrete with three commitments. First, every new piece of knowledge passes through review before it enters institutional memory — proposed, reviewed, and either promoted or rejected with a reason. Second, every customer install ships with the same review discipline applied to its own engagement work; the customer's own knowledge is structured the same way ours is. Third, the discipline is measured. Time spent in each phase of the loop is observable. Long orientation times signal indecision. Long action times signal execution friction. Long observation gaps signal the loop is not closing.
This is what we mean when we say ORCA is infrastructure for the Orient phase of Boyd's loop. We are not selling answers. We are selling the discipline that makes answers reviewable.
The market position
Most of our competitors are in the typewriter business. They are selling fluent text. We are selling the smallest set of explicit components that makes a piece of work auditable end-to-end.
If your priority is fluent text, you have many options. If your priority is reasoning that survives scrutiny — by your own future self, by your clients, by your regulator, by whoever buys you — there are fewer of us, and we operate on different first principles.
Vibes don't ship. Systems do.