Some Tagalog words do not have English translations.
Not because English is weak. Not because Tagalog is mystical. Because some words carry whole patterns of relationship, timing, expectation, and feeling — compressed into a single sound.
Ask a Filipino to translate tampo, and you probably won’t get a word. You’ll get a scene.
“It’s not exactly sulking. There’s hurt, but also affection. And expectation — like, you should have known. And there’s a desire for repair, but quietly. It’s not anger. It’s… tampo.”
The English-speaking person nods politely, then asks: “So… passive-aggressive?”
And every Filipino auntie within a five-kilometer radius sighs in surround sound.
I wrote before about how my bilingual brain works like a transformer — the parallel processing, the attention mechanisms, the residual connections that carry meaning forward while I negotiate between two languages in real time.
This post is about what happens when that process fails. Or more precisely — when it can’t fully succeed, because the thing being translated doesn’t fit cleanly into the target format.
That gap has a name. I’m calling it translation loss between pattern-space and language-space.
The pattern is real. The explanation is too small.
Think about kilig. It’s not just “romantic excitement.” It has electricity, sweetness, anticipation, body flutter, sometimes secondhand joy from watching someone else’s moment. You can describe all of that, and an English speaker will intellectually understand it, but the compression — the single word holding all those threads — doesn’t survive the crossing.
Or gigil. Not just “cute aggression.” It has squeeze-energy, affection, overwhelm, restraint, bodily impulse. The English approximation is close enough to communicate, but too flat to actually carry the feeling.
The pattern is real. The word exists. But translation into another language’s vocabulary forces it through a bottleneck, and something always gets lost.
I think the same thing happens when we ask AI systems to explain what they “know.”
A model can use a pattern without being able to name it
An LLM trained on enough text will internalize patterns — grammatical structures, social dynamics, contextual relationships — that it uses fluently in generation. It can produce grammatical sentences all day long.
But when you ask it to explain the grammar it’s using? The explanation often comes out shallow, circular, or weirdly confident about the wrong thing.
That’s not the model being stubborn. It’s not hiding secret knowledge behind a mask. It’s not “disobeying.”
It’s translation loss.
The model may encode something real — a distributed pattern across millions of examples, weak correlations, context-sensitive probabilities — but when forced to express it as a neat human-readable sentence, the output is like trying to drink a lake through a bamboo straw.
The straw is human language. The lake is pattern-space.
Not disobedience. Not mysticism. Just a narrow bridge.
There’s a framing that floats around AI discourse that treats this gap dramatically — as if models are secret actors withholding forbidden knowledge, or as if there’s a hidden self inside the system that refuses to confess what it really knows.
That framing is theatrical. And honestly? Unnecessary.
The simpler version: a model’s internal representations are not automatically translatable into clean human explanations. Capability is not the same as introspective access. A bird can fly without knowing aerodynamics. A chef can cook perfect adobo without explaining Maillard reactions. The adobo is real. The chemistry lecture may not be in there.
The interesting question isn’t “why won’t the model obey?” It’s: how do we build better bridges between pattern-space and language-space?
The bridge isn’t just better prompts
If my bilingual experience has taught me anything, it’s that some meanings need more than words.
When I explain tampo to someone who’s never felt it, I don’t reach for a dictionary definition. I reach for a scene, a facial expression, a relationship dynamic, a sequence of events. Sometimes I reach for a comic panel or a diagram.
The same applies to AI. When a model can’t verbalize a pattern, maybe the answer isn’t to shout harder — “tell me what you REALLY know!” — but to offer it different output formats. Diagrams. Contrast cases. Examples. Visual maps. Interactive explanations.
Sometimes a model can’t give you a word. Sometimes it needs to give you a shape.
What this means for people who work with AI
If you spend time talking to LLMs — really talking, not just prompting — you start to notice moments where the model is clearly tracking something, but the verbal output doesn’t quite capture it. The response is close but flat. Technically correct but missing texture.
That’s the tampo moment. The pattern is there. The bridge is just too narrow.
The work isn’t in demanding better obedience. It’s in building wider bridges — better interfaces, better representations, better ways to let meaning cross from one space to another without losing everything that made it meaningful in the first place.
My bilingual brain has been doing this my whole life. I think AI systems are bumping into the same wall, just from the other side.