I’m going to start with a few examples that seem unrelated. They live in different domains, involve different actors, and are usually discussed in isolation. The connection will only become clear at the end.
1. The no-fly list
People are placed on no-fly lists by mistake with disturbing regularity. Once on the list, getting off is notoriously difficult—sometimes impossible.
This is often explained away in familiar terms: classified criteria, opaque intelligence sources, name-matching errors, or overbroad watchlists. These explanations are not wrong, but they are incomplete.
What matters here is not why an error occurred, but the mechanism that follows once suspicion is recorded. Some signal crossed a threshold. That signal was logged, propagated, and reused by downstream systems.
The individual rarely knows:
- what triggered it
- what data was used
- how it was transformed
- who to appeal to
The record does not have to be correct. It only has to exist.
Once it exists, it acquires authority.
2. Twitter archaeology
Years-old tweets are excavated, screenshotted, stripped of context, and redeployed in the present. A casual joke, a half-formed thought, or a moment of poor wording becomes evidence of character.
This is not historical inquiry. It is adversarial recompression — the deliberate extraction and re-presentation of a compressed artifact in a way that discards context and maximizes downstream leverage.
What remains is a compact, legible artifact that can be circulated and weaponized.
Again, accuracy is secondary. The tweet does not need to represent what the person believes now—or ever believed.
It only needs to be retrievable.
3. Scientific research
Science is often held up as the antidote to these problems, but it quietly relies on the same mechanism.
Raw reality is not preserved. Observations are:
- sampled
- abstracted
- reduced to variables
- summarized into models
This is necessary. Without compression, science cannot scale.
But problems arise when:
- models are treated as reality
- assumptions are forgotten
- uncertainty is collapsed into point estimates
- downstream users inherit conclusions without context
When that happens, disagreement is no longer a scientific question. It becomes a deviation from the record.
This is visible whenever a proxy or metric—an average temperature anomaly, a risk score, a confidence interval, a benchmark accuracy—is treated not as a model output with assumptions, but as a direct description of reality itself.
4. What these have in common
At first glance:
- security watchlists
- social media call-outs
- scientific models
seem unrelated.
But structurally, they share the same pattern:
- Reality is compressed
- Context is discarded
- The compressed artifact is stored
- The artifact becomes authoritative
- Decisions are made downstream
- Correction is difficult or impossible
This is not about malice. It is about lossy compression.
5. The thesis: lossy compression creates power
It is important to be explicit here: compression itself is necessary and unavoidable. No system that scales—scientific, technical, or institutional—can operate without reducing reality into manageable representations.
Lossy compression is not just an efficiency tradeoff. It is a power transformation.
Compression:
- makes information portable
- makes it cheap to store
- makes it easy to reuse
- makes it legible to institutions
But it also:
- removes contestability
- erases provenance
- hides uncertainty
- outlives its original context
Once compressed, reality cannot argue back.
A person can insist they are not a threat. A tweet can be explained. A model can be challenged.
But the record already exists.
And in modern systems, the existence of a record outweighs its truth.
6. The danger is not error — it’s irreversibility
There is a fundamental asymmetry at work: creating a compressed record is cheap and instantaneous, while correcting or unwinding it is slow, costly, and often institutionally discouraged.
The real failure mode is not that compression is wrong. All compression is wrong, by definition.
The failure is that compressed representations:
- are reused as if they were complete
- gain authority without accountability
- cannot be meaningfully appealed
- persist long after their assumptions expire
At that point, error stops being a mistake. It becomes a condition of life.
7. Closing
We often say “the internet is forever,” but that’s not quite right.
What is forever is not the raw data. It is the compressed trace.
And once compression becomes authoritative,
truth becomes optional—but power does not.