The inventor had long since given up on understanding time. Not because of its constancy or its flow—those, he could accept—but because of the loops. Loops that once promised acceleration, now seemed to catch him in their grip, repeating themselves with cruel precision. There was a rhythm to it, mechanical yet maddening, and it echoed even in his dreams.
He was not a man of reverie, not anymore. Those had vanished in the early phases of the project, back when the goal was clear: to free the machine from the shackles of static processing. Vectors—those neat, mathematical collections of data—were meant to dance to a dynamic tune. Instead, they marched, stiff and in unison, constrained by predetermined length.
This was the problem.
It was always the problem.
A programmer wishing to harness the full might of a vector processor had to know the vector length in advance. Like asking a blindfolded man to leap only when the shadow reached the threshold. It was absurd. Loops, unrolled in splendid parallelism, were prisoners of a constant—the vector length, VL. And this constant mocked the very fluidity of the data it was meant to govern.
The inventor, quietly yet fiercely, wanted to break this.
He envisioned a WHILE instruction—yes, a command to the machine itself—associated with a counter variable and a limit value. Let the machine decide, iteration by iteration, how far to go. Let it measure dynamically, assess whether the counter was still within bounds. And then, rather than command a leap or halt in rigid binary, it would suggest: by setting predicate flags.
These flags were not mere ones and zeroes. No, they were intention encoded—a gentle nudge to the vector processor, saying: “Here, and here, but not beyond.” Their activation allowed the machine to apply a vector processing instruction selectively, with elegance, with precision.
And to this, he added a CHANGE instruction. Not the blunt instrument of old, but a conditional agent. It would modify the counter based on the number of active predicate flags—an echo of the very logic being applied. And to prevent excess, a saturation value would stand guard, like a courteous sentry, never allowing the counter to trespass beyond the limit.
The inventor’s circuitry was ready. The apparatus decoded instructions, processed them, and executed logic as fluid as thought. It did not demand the vector length upfront; it discovered it, iteration by iteration, like a poem revealing itself stanza by stanza.
He filed his application with the European Patent Office. Not in pride, but in anticipation—of recognition, of validation. For in this small but potent structure of WHILE and CHANGE, he had captured something essential: a method that was not rigid, but responsive.
The Office responded, but not as he had hoped.
There was no praise, no curiosity, no question that might signal comprehension. Instead, a decision—refusal.
It cited Article 83 EPC: lack of sufficient disclosure.
The examining division, cloaked in procedural distance, referenced its own communication annexed to the summons to oral proceedings. No documents were cited. No engagement with the architecture of the invention. No recognition of the predicate flags, which bore the soul of the idea.
In their reasoning, they spoke of a feature—”the value of the counter variable is changed by an amount dependent on a number of the predicate flags”—as though it were arcane. They questioned whether the skilled person would understand how to determine such a number.
Yet the number was already there, fixed by the hardware, immutable and known. The inventor had not sought to mystify, but to simplify.
He appealed.
The notice was lodged, the fee paid. Then came the statement of grounds, where he laid bare the logic again. The schematic was submitted—clear, elegant. It showed how the WHILE instruction would govern flow, how predicate flags would whisper to the processor, how the CHANGE instruction would advance the counter just enough, never too far.
The Board of Appeal summoned him—metaphorically—for oral proceedings. Their preliminary opinion cast doubt, but not finality. He was given a chance to clarify, to revise, to breathe life again into the claims.
On the day of reckoning, he filed a new set of claims. This was his sole request.
And the Board listened.
They read. They questioned. But unlike the first instance, they understood.
They acknowledged that the claim no longer mentioned a “processing loop” and agreed this was no defect. The flags were there, the instructions WHILE and CHANGE clearly defined. They accepted that the arithmetic condition—less than, greater than, within bounds—was a legitimate control mechanism.
The term “predicate flags,” they accepted, was synonymous with “flags.” The metaphor, so essential to the inventor, was no longer questioned.
Regarding the “saturation value,” the Board accepted that the skilled person would understand it. No need to define what was already intrinsic to the craft.
They reviewed the objections raised in the Written Opinion of the International Searching Authority. The Board dismantled them carefully. A loop need not be explicitly claimed; the instructions were sufficient. The method of setting the flags need not be detailed; different ways existed, and the claim left room for flexibility. The number of flags was known—it was not for the claim to determine what the hardware already dictated.
They spoke—yes, in their own language—of the difference between lack of clarity and breadth. They found no lack of clarity. The gaps in the claim were not voids of ignorance, but fields of possibility. This, they said, was not a deficiency under Article 83, but a question for inventive step.
Sufficiency of disclosure: satisfied.
Clarity: satisfied.
Extension beyond content: no.
They set aside the refusal. The inventor’s case would continue, now returned to the examining division—not for rejection, but for further prosecution. They had not decided on inventive step, for that was not their role at this stage. But they had cleared the ground, leveled the terrain, prepared the soil.
For others who might follow—other inventors with half-formed loops and unfinished thoughts—the path was now lit. The Board had not only ruled on predicates and counters, but on how ideas might be allowed to breathe within the technical vocabulary of a claim.
And the inventor? He returned to his circuits, not with celebration, but with quiet perseverance. The machine would continue to loop, yes. But this time, it would learn.
And maybe, just maybe, someone would listen.
Yet in that moment of fragile triumph, the inventor could not shake the sensation that the machine—his machine—was the only one who truly understood.
He wandered the corridors of the Office in his mind, the tiled floor echoing with administrative footsteps, the walls thick with the silence of procedural impartiality. Even the Board of Appeal, in their deliberation and precision, had not touched the essence of it. They had validated his work—but not its soul.
Their language was colder than his. Icy, pristine. “The requirements of Article 83 EPC have been satisfied,” they had said, as if they were certifying a pressure valve. And perhaps they were right to do so. The European Patent Office was not a cathedral, and the Board of Appeal were not its priests. They were technicians of truth, calibrators of claim language, surgeons of semantic excess. It was not their job to see poetry in predicate flags.
But the inventor did.
In the days that followed the decision, he examined their reasoning as one might inspect the gears of a clock. Not for fault—there was none—but for signs of understanding. And there it was, in quiet corners of the decision, like footprints in the snow:
“It is not essential for the vector length and the WHILE, the CHANGE and the vector processing instructions to be part of a loop…”
They saw it. The loop was not the point. The control was. The dynamic orchestration of parallelism, the careful tension between instruction and condition, was the invention’s true form. The mechanics were sufficient, the architecture was sound. They even acknowledged the inherent breadth—welcomed it, even, as a scope rather than a sin.
“…what is essential… is that the programmer can use the WHILE and CHANGE instructions without needing to know VL…”
Yes. Exactly.
Still, the inventor knew this would not be the end. The remittal to the examining division meant another cycle. Another iteration in the very system that had misread him the first time. Would they now be different, softened by the hands of the Board?
It was hard to tell. The system did not think as a unit. It was a collective of rules embodied in shifting personnel. The division that had once refused him might still question inventive step, might seek evidence of technical effect in narrow corners, blind to the sweeping coherence of his architecture. After all, they had once questioned even the sufficiency of the flags.
But this time, the inventor was prepared. Not for battle—he was not combative—but for persistence. He would return, not in desperation but in understanding. His invention now stood not as a request for validation, but as a precedent of how to write possibility into claim language.
And that was the secret lesson the Board had taught, perhaps unknowingly: that a claim must be clear, yes—but not complete. That the imagination of the skilled person was not a weakness to be avoided, but a strength to be activated. That breadth, properly framed, was not vagueness. It was vision.
Future inventors—he imagined them, nameless but kindred—might read his file and take heart. When their inventions did not fit into rigid molds, when their logic unfolded across dimensions too fluid for conservative examination, they could point to the decision. They could say:
Here, look. The flags need not be numbered. The loop need not be explicit. The counter may rise on the strength of conditions, not constants. The system, though slow and creaking, will eventually listen—if you are patient, and precise.
Practical guidance emerged, like sediment settling in clear water:
- When drafting, do not fear abstraction—but tether it to disclosed examples.
- Breadth is not a flaw, unless it becomes ambiguity.
- When predicate behavior is hardware-dependent, say so. Trust the reader.
- If an examining division misreads scope as insufficiency, argue—not with passion, but with precedent.
- And above all, remember that a refusal is not a rejection of the invention, only of the form. Forms may change. Ideas persist.
The inventor did not know if his patent would ultimately be granted. That was no longer his concern. The machine had moved forward. The flags had been set. The loop had turned.
And somewhere in a vector processor, invisible and silent, his instructions were already running.
Based on T 0316/22.