Who Is Behind AB SoftLoop?

Bert Rendon did not learn to program in a classroom.

He learned by looking over someone else’s shoulder.

It was 1993, in Maracaibo, Venezuela. Computers were not where his story was supposed to lead. He had trained as a diesel mechanic and apprenticed at LagoVen, surrounded by the machinery of the oil industry. But by then he was working as a cost analyst in a large hardware store, and behind the numbers he handled every day was another kind of machine.

The store ran on an inventory system written in FoxBase on UNIX. Bert began studying it: first watching, then asking questions, then understanding what the code was doing beneath the screen. Before long, he had taken responsibility for its inventory module.

That same year, he built something of his own. He called it, with the confidence of youth, “La Inteligencia Artificial”: a decision tree that asked questions, weighed the answers, and could learn new ones. Primitive by today’s standards. But the instinct behind it would follow him for decades:

Could a machine do more than store information? Could it reason with it?

He earned an associate degree and began a bachelor’s in computer science, but his career moved faster than the curriculum. His skills took him into database architecture and production systems, where software stopped being an exercise. If something failed, somebody had to find out why.

In Tampa, he founded AB CleanRoom Services and entered a different world: pharmaceutical and biotechnology cleanrooms, where procedures must be followed, actions documented, and small deviations carry consequences far beyond a screen. Once again, he built software because there was a problem in front of him. VabcV-CRM and its GMP Training Portal were built for his own operations, crews, training, and audits. They had to survive reality before they could become products.

That philosophy carried into THE C.U.R.E., the subject of a U.S. provisional patent application, into SynLoop, and into everything AB SoftLoop builds today.

Then something fundamental changed.

Artificial intelligence had changed the tools. So Bert changed the method.

Bert does not build alone anymore, and AB SoftLoop does not build software the way it was built in 1993.

He is the human architect and master orchestrator behind every product. Before any code is written, he defines the world the software is allowed to exist in: its architecture, how information moves through it, what authority each part holds, how it behaves when things go right and, more importantly, what it must do when they do not. He sets the rules no line of code is allowed to break.

Then he orchestrates the builders.

Around him works a deliberately structured team of collaborating AI agents. Their responsibilities are divided. One translates architectural decisions into precise instructions. Coding agents, running in parallel in Claude Code, each build their own part of the system under the same architecture and the same rules. Other agents come at the work from the opposite direction. They test it, challenge it, and audit it, searching for what the builders missed.

The orchestra may be artificial. The conductor is not.

And no agent gets to grade its own homework. Every significant change must survive automated tests written before or alongside the code. Smoke tests make the system prove it actually starts and runs, not merely that it compiles. Independent audits check the work against the architecture and its rules. The most important promises become automated gates, so they cannot quietly disappear as the system grows.

So the evidence moves upward, until the decision returns to the human who defined the system. Nothing ships until Bert reviews it and approves it.

The method is fast: parts of a system that once had to wait their turn can now be built, tested, and audited at the same time. But speed is not the point. Integrity is: an architecture that holds when the product gets complicated, a flow that behaves as designed, a decision that can be traced to the evidence behind it, and a failure that announces itself instead of hiding behind a convincing answer.

AB SoftLoop does not claim software can be flawless. That promise is easy to make and impossible to prove. Its method is built around something harder:

• Find what is wrong before the customer has to.
• Test what can be tested.
• Mark what cannot be confirmed.
• And never call uncertainty certainty.

Nowhere is that more visible than in SIGNAL. When SIGNAL meets evidence that does not fully align, it does not discard the disagreement or polish it into certainty. It reports it, keeps the reason, and marks the case for deeper review.

SIGNAL is built according to the same principle by which SIGNAL is meant to think.

More than three decades after a young man in Maracaibo learned to program by watching over a developer’s shoulder, Bert works alongside teams of machines that write, test, audit, and challenge software with him. His role has changed. The principle has not.