The Capture Layer | Seevali Rathnayake

The most important part of any knowledge system isn’t the structure. It’s the moment of capture — the 3-second window between having a thought and losing it.

Most note-taking systems fail here. They ask too much of the moment. Open an app. Choose a folder. Add tags. By the time you’ve done all that, the thought is half-formed, the context is gone, and you’ve talked yourself out of writing it down.

I solved this by making capture feel like texting.

The WhatsApp Journal Group

There’s a WhatsApp group on my phone with one member: me. Every raw thought goes there.

Not polished thoughts. Not thoughts that are ready to become content. Rough material — an observation from a meeting, a decision I made and why, a pattern I noticed, something that went wrong and what I’d do differently.

The messages look like this:

“integration platform — I’m not going to build this just for the current requirement. I won’t follow the same old way. My target is the future integration pipeline we will have to cater to in the next 2 years.”

“third incident this week and I keep seeing the same thing. it’s not really a code problem. the two teams have completely different mental models of how the system works. the bug is just where that mismatch happened to surface.”

“we measured first this time before touching anything. the bottleneck wasn’t the query like everyone assumed. it was the serialiser. if we hadn’t measured we would have spent a week optimising the wrong thing entirely.”

These aren’t blog posts. They’re not even notes yet. They’re raw — the kind of thing you’d say out loud to a colleague but never write down because it doesn’t feel important enough.

That’s exactly why they need to be captured.

What Happens Next

Every message in that group goes to Lyra, my AI agent powered by NanoClaw. She doesn’t rewrite it. She routes it.

Short messages become bullets in the day’s journal: 08_Journal/2026/2026-03-28.md. One line, timestamped, in context with everything else from that day.

Longer messages get a subsection with a heading. A decision gets a note in 01_Projects/[Name]/Decisions/. An incident goes under Notable Moments. An achievement becomes a file in 02_Areas/Career/Achievements/.

The routing is triggered by prefixes — but most of the time I don’t even use them. I just write naturally. Lyra reads the signal in the content and routes accordingly.

journal: [text]    → today's journal, most relevant section
achieve: [text]    → Career/Achievements/
incident: [text]   → journal, Notable Moments section
decision: [text]   → active project, Decisions/
p: [text]          → active project, Progress/ (daily aggregate)

Without a prefix — just a raw message — it defaults to the journal. That’s the right default. The journal is where everything lands before it finds its permanent home.

Why This Works

Three things make this capture layer actually stick:

Zero friction at the moment of thought. WhatsApp is already open. I’m already in the habit of sending messages. The capture behaviour borrows from an existing habit rather than building a new one.

Routing is automatic, not a decision. I don’t have to think about where something belongs. That’s Lyra’s job. My job is to get the thought out of my head while it’s still sharp.

The bar for capture is intentionally low. Half-formed thoughts count. Observations that don’t yet have a conclusion count. The system doesn’t require polish — it requires presence. The structure comes later, in the KB. The capture moment is just about not losing it.

The Output of Capture

What that accumulation looks like in the vault:

Daily journals that read like a real working log — not a polished retrospective
Decision notes that capture the actual reasoning, not the cleaned-up version
Achievement records that exist because someone remembered to write them down
A thread of incidents that, when read together, show patterns I wouldn’t have noticed individually

None of that came from sitting down to write. It came from texting.

That’s the capture layer. The structure that holds it is next.

Signal 2 of 7 in The Second Brain That Publishes Itself.