For the past few years, I’ve been wrestling with a problem that’s surprisingly hard to articulate:
As someone who plays multiple roles every day, the challenge isn’t “being busy”, it’s being cognitively split.
Slack channels stack on top of Telegram groups.
Telegram groups stack on top of Discord communities.
Emails sneak in between everything else.
Dozens, sometimes hundreds, of messages across all these surfaces.
Not because I’m “popular.”
Not because I enjoy noise.
Simply because every role I hold connects me to a different network of people, expectations, updates, and decisions.
And no matter where I am, boarding a flight, running between meetings, or finally winding down at night,
the messages don’t stop.
At some point, I realized:
We don’t have a messaging overload problem.
We have a cognitive architecture problem.
This essay is my attempt to break that down—and to explain why Lumis begins where the real gap is.
The Five Layers of Human Communication
Most tools think communication is about sending and receiving messages. But when I look at my own day, the real process is much more complex.
I break it down into five layers—a cycle every one of us goes through, whether consciously or not:
Sensing (Input)
The world comes in fragmented pieces.
Some updates happen in Slack threads.
Some happen in Telegram groups.
Some happen in meetings, tone, or implicit context.
Understanding (Meaning)
Who is who?
Which role am I acting in?
Where are things now?
What changed?
What matters?
This is the core cognitive challenge.
And no tool truly helps us do it.
Expressing (Communication)
Once things make sense, we can articulate them, write updates, answer questions, draft reports, clarify decisions.
Acting (Execution)
Updating tasks, aligning people, moving projects, following up.
Agency (Operating on Your Behalf)
In a mature system, once understanding is deep enough, the tool can begin to act for you,
nudging, reminding, filtering, protecting your time, even communicating where safe.
This is the full cognitive cycle.
And the real fatigue we feel comes from the fact that today, humans have to manually stitch these layers together.
Why Most AI Tools Haven’t Solved the Real Pain
Over the past two years, I tried nearly every popular AI tool.
They helped me:
- searching
- polish English
- translate emails
- summarize long threads
But they all shared the same limitation:
They worked on the “cut edges” of communication, not the core.
They didn’t know:
- what role I was in
- who mattered to me
- the history of a project
- the long-term context behind a decision
- the implicit dependencies
- the interpersonal dynamics
- what had happened earlier in the day
So in the end, I still had to:
- read everything,
- think it through myself,
- understand the situation,
- then ask AI to help with phrasing.
The hard part, understanding, was still mine.
AI solved expression.
It solved formatting.
It solved summarization.
But it didn’t solve cognition.
When Lumis Finally Felt “Different”
Recently, while using an internal build of Lumis, something clicked.
Lumis didn’t only show me messages.
It showed me states:
- Here’s what progressed today.
- Here are the things your role cares about.
- Here’s what resurfaced.
- Here’s the new risk.
- Here’s what actually needs your judgment.
No scrolling.
No reconstructing.
No trying to piece together scattered threads.
It felt like someone pre-processed the day for me.
Then another moment hit me:
With long-term memory, Lumis could reply on my behalf—while preserving my context, tone, history, and relationships.
Not in a “template + fill-in-the-blanks” way.
But in a “this is actually what I would say” way.
That was the first time I felt an AI tool doing something structural, not just cosmetic.
Of course, this is early.
There’s a long road ahead.
But the direction feels right.
Why Start With Understanding?
If you map the five layers again:
- Sensing → Many products can do this.
- Expressing → ChatGPT handles this well.
- Acting → Automation tools exist.
- Agency → Everyone is imagining this future.
But Understanding is the layer that almost no tool meaningfully solves, yet it’s the center of the entire cycle.
Without understanding:
- automation is brittle
- summaries are shallow
- replies feel generic
- decisions lose coherence
- attention keeps fragmenting
We chose to start here for one simple reason:
If AI can’t understand you, it can’t help you.
Most people today operate in a constant state of reactive overload.
If we can reduce that—even partially, it’s genuinely a public good.
We’re not trying to build another productivity hack.
We’re trying to rebuild the cognitive layer underneath modern work.
The Long-Term Vision: The Full Cycle
Even though we’re starting with Understanding, our long-term goal is clear:
Rebuild the full cognitive cycle—end to end.
A system that can:
- capture your world
- understand your roles
- model your context
- express your reasoning
- help you act
- eventually operate as a trusted extension of you
Not replacing humans, but restoring their capacity to think, to choose, and to stay whole across all the roles they inhabit.
That’s the direction Lumis is heading toward.
And honestly, I’m building it because I need it myself.
Top comments (0)