When I was a kid, the sources of information were few and well-defined. You had books, which were expensive and took a long time to produce. You had newspapers, which came out once a day. And you had the three TV channels/networks. It wasn’t that these sources were always right, but you knew what you were getting. The menu was short.
Now we live in a firehose. Information comes at you from every direction, all the time. And the most dangerous thing about this new world isn’t the lies, it’s the stuff that’s almost true. The things that are a mixture of fact and opinion, so cleverly blended that it’s hard to tell where one ends and the other begins.
If you’re trying to do anything important, like build a company or just figure out what’s actually going on, you have to get good at separating these things. You have to develop a kind of “smell test” for truth.
Most of us think of statements as being in one of two buckets: facts or opinions. “The unemployment rate is 5%” is a fact. “The government is doing a bad job” is an opinion. This is a good start, but it’s too simple. In the real world, information exists on a spectrum.
At one end of the spectrum, you have raw, verifiable data.
The server returned a 500 error at 10:32 AM.
This is the bedrock. It’s as close to ground truth as you can get. There’s no interpretation. It’s not debatable.
Move a little along the spectrum, and you get interpretation.
The server crashed because of the recent code push.
This is still close to the fact end of the spectrum. It’s a hypothesis based on evidence. An experienced engineer makes this kind of connection all the time. It’s useful, but it’s not the same as the raw data. The crash could have been caused by a sudden spike in traffic that just happened to coincide with the push. The interpretation is a new layer, built on top of the fact.
Travel a bit further, and you start to see personal views mixed in.
The server crashed again. This is what happens when we rush features instead of focusing on stability. We need to overhaul our entire process.
See what happened? We started with a fact (the server crashed), added an interpretation (it was because we rushed), and then bolted on a strong opinion about what to do (overhaul the process). This is where things get dangerous. The opinion part feels weighty because it’s anchored to a fact, but the chain of logic is getting longer and more fragile. Most of the content we consume lives here.
Keep going, and you get to a point where the opinion almost completely swallows the fact.
This project is a disaster; the whole system is constantly falling over.
The phrase “constantly falling over” is an exaggeration born of frustration. It contains a kernel of truth, the system did crash, but that kernel is buried under a mountain of feeling. This is more of a complaint than a report.
And at the far end of the spectrum, you have pure opinion, unmoored from any evidence at all.
Our whole tech stack is garbage.
This is a statement of feeling, not fact. It’s useless for diagnosis. You can’t fix “garbage.” You can only fix a specific error that happened at a specific time.
Why does this matter? Because if you’re anyone who has to make decisions, you want to operate as close to the “raw data” end of the spectrum as you can. You can’t afford to build on someone else’s interpretation, let alone their opinion. Your job is to take the raw data and build your own chain of logic.
This isn’t easy. It’s work. It means that every time you read an article or listen to someone talk, you have to be actively asking yourself: where on the spectrum does this live? Is this raw data, or is it an interpretation? Is there an opinion baked in here? Whose opinion is it, and what are their incentives?
Developing this filter is a form of intellectual hygiene. You’re training your brain to reject low-quality information. The more you do it, the better your mental models of the world will become, because they’ll be built on a foundation of reality, not on the shifting sands of other people’s views.
The world is awash in information. Most of it is a mixture of signal and noise. Your job is to be a brutally effective filter. Start by asking: how far is this from the raw data? The answer will tell you almost everything you need to know.