What Even Is Context?

When you say “they have the context,” what do you intuitively feel and think? That this person has the information required to have a meaningful conversation on the topic itself, no metacommunication required?

They know everything required to talk about the core idea/story. No background information required to “catch them up.”

When you catch someone up, you’ve given them the information and it changes the meaning of the information they already had or is net new.

All context is is information that changes the meaning of other information.

So when you’re doing context engineering, you are manipulating the structure and timing of when information changes the meaning of downstream information.

I recognize how insanely abstract that likely sounds, but let’s use a physical analogy to make it more tangible.

Imagine a physical assembly line, big robot arms welding big chunks of metal to one another.

Would it make sense to say all manufacturing is, is manipulating the physical structure of matter to change the structure of other matter?

Probably not day to day, but from a first principles perspective, that is completely correct. Systematic manipulation to achieve a given outcome.

Context engineering is this exact systematic manipulation to achieve a given outcome for manipulating meaning itself.

Sitting below context engineering is software engineering. Software engineering is the systematic manipulation of information to create meaning out of raw information. Information does not inherently have meaning. Meaning must be created and assigned.

Data -> Information -> Meaning

Ultimately, it is up to a subjective human being to make up and create that meaning, and only then can we process it in our meaning scaling machines (our computers) using deterministic logic. You could send meaning around using computers but the computers lacked the ability to interpret it.

That was it until the advent of probabilistic, context-aware, systems like large language models. Now our machines cannot just manufacture meaning, but interpret it, change it, and interact with it programmatically. This is pretty big.

Now we can encode our unique meaning interpreting rules into these system. We can imbue systems with our unique perspective on what a given piece of information means, and then not only will it store that definition, but be able to actually interpret it and relate it to other people’s definitions and personal meanings.

This interplay between subjective assignment of meaning is what creates space for the most elusive, uniquely, unquantifiable human creation, taste.

Taste is inherently subjective. You can’t buy it, you can’t automate it, and you either have it or you don’t. It sits in the negative space in between things. It changes and flows just like our lives do.

And now for the first time our systems can be encoded to reflect it. It can never be fully copied or distilled, but a negative can be imprinted on the systems itself like how you can stamp hot wax.

The more refined your taste AND how clearly you can communicate it, the more effectively you can program these meaning-interpreting machines to use your perspective and see the world through your lens, literally seeing the same information and coming to a different conclusion. This lets you amplify your taste.

The winners of the whole AI game are going to be those who:

1. Have a unique perspective

2. Can clearly communicate it

3. Others like it and think it is valuable/vibe with it

These people will be able to amplify their taste and scale their expertise for effectively zero marginal cost of reproduction.

This is why when everyone else defines context engineering with some paragraph-long technical enumeration ritual paying homage to this quarter’s AI buzzwords, all I think it is is a ruthless pursuit of clarity and obsession on the why.

Everything flows from the clarity of your thought and your ability to communicate it. Human thought is the most powerful force the world has ever seen, and I don’t see that changing any time soon.

So clarify your thoughts to bring meaning to these lifeless machines. Give them meaning yourself.