At face value, summaries just take lots of words and mash it down into less words.
Highlights say, “We know enough about you and your world to focus on what matters most in this big blob of text.”
It’s a subtle nuance, but an important one. Summaries shorten. Highlights curate.
For example, what if you used a summary of Gatsby to write a last minute paper for class. What’s the paper supposed to be about? A plot overview? The illusion of the American Dream? The fragility of identity? That probably matters.
This obviously requires context. You have to know enough to curate effectively.
It’s one of the reasons the average person struggles to work with LLMs. They haven’t taught the model about their world first (or stuffed it into a prompt). So when they feed it a document and it spits back a summary, they’re nonplussed. It’s not the machine’s fault – it doesn’t have sufficient context.
In Steady, we put the process of gathering context on rails. Context, by the way, is not just data, it’s a distilled mix of “hard” information from things like project management tools and “soft” information provided by humans–intentions, risk assessments, progress towards goals, and blockers.
That’s how we’re able to provide highlights: distilled and targeted context, derived from hard and soft sources, delivered on cadence.
Summaries without context are marginally helpful at best. We can do so much better with highlights.
