About
Stable, defining, foundational. This is where identity and operating constraints live.
Identity Values Goals Rules Boundaries
How it works
I built this because agents deserve better context
About Now Data is my take on agentic memory. A simple Markdown-based system that adapts to you, not the other way around. I kept identity, priorities, and knowledge separated so they stay legible—for you and for any agent.
My goal wasn't just better memory inference. I wanted memory you can inspect, edit, move, and trust.
About.md carries stable context: values, rules, and identity. Now.md tracks
what matters today: active priorities and next steps. Data stores structured details:
projects, references, and the tools you work with.
A memory you can't read isn't a memory—it's just noise. That's the rule I started from.
- Base unit
- Three semantic units:
About,Now,Data. That's it. - Governance
- Human-curated, explicit, inspectable. I don't believe in silent consolidation.
- Temporal model
- Clear separation between what is stable and what is current.
- Main risk
- It becomes thin or stale if you stop updating. That's the honest trade-off.
- Main strength
- Full auditability and human legibility, even with multiple agents involved.
Three Layers
One system, three roles.
I kept the framework small on purpose—each layer has a precise job. That separation is what makes it usable for both a person and an agent.
Now
Brief, current, actionable. This is the fastest-moving layer and the easiest one to refresh.
Top of mind Active projects Recent activity Next steps Risks Opportunities
Data
Structured detail and project knowledge. This layer absorbs supporting depth without polluting the core.
Projects/ Notes.md Skills.md Tools.md References Archive
Workflow
How to start.
Drop the files into your project and fill them with what matters to you.
Point your agent to the folder.
Audit the reasoning by opening the Markdown files. That's the whole point.
1
Read About.md first
Start with a structured map, identity, values, and boundaries. Ensure behavior is aligned before any action begins.
2
Act from Now.md
Use the current state to prioritize correctly. Avoid acting on stale assumptions or outdated context.
3
Use Data for depth
Pull project details, notes, and references only when needed. Keep the "working memory" lean.
4
Write back explicitly
When the system state changes, update the relevant file—manually or automatically via your agent. Never invent hidden memory or "hallucinate" context outside the Markdown structure.
Comparison
The difference is where context lives.
Most memory systems optimize for automatic capture, latent inference, and silent consolidation. That feels convenient—until the stored memory is wrong, outdated, or impossible to inspect.
I designed About Now Data from the opposite position. Memory should be legible. Stable context and changing context should never be mixed. And human control always matters more than hidden cleverness.
The Usual Setup
Clutter and opacity
Opaque and Fragmented
Context is often trapped in black-box databases, hidden chat histories, or proprietary software that humans can't easily audit.
Context Drift
Stable values and temporary tasks get mixed together, leading to "hallucinations" and agents that lose track of who they are representing.
Productivity Theater
Systems often prioritize complex automation and "completeness" over actual utility, creating a maintenance burden that feels like work instead of progress.
Memory as a file
Information is stored to preserve it, organized by type or date, and retrieved through search from a mostly static archive.
With About Now Data
Clarity and control
A Minimal, Open and Portable Architecture
One About file, one Now file, and one Data folder. A universal structure that works anywhere Markdown does—no heavy software required.
Human-Centric Control
Auditability beats "black-box" convenience. You can see, edit, and move your memory at any time, ensuring you—not the algorithm—stay in charge.
Agentic Alignment by Design
By separating stable identity from current priorities, the system provides a clean foundation for agents to act with precision, without polluting your core context.
Memory as a model
Information is stored to improve reasoning, organized around what an agent needs to know, and recalled through context-aware inference that evolves with each interaction.
Future
Start small. Adapt as agents evolve.
Thanks to recent model improvements, refreshing Now.md is no longer the hard part. An agent
can summarize recent activity, extract next steps, or rewrite your current state in seconds—regardless of
the ecosystem you use.
That's what makes this more than a local format. It's a transversal agent memory—a single structure that different agents can read, update, and respect without being locked into a single vendor or product.
The real risk isn't that agents can't update the system. The risk is that you stop keeping it alive.
My design goal was to make updating Now.md cheap enough that the system stays effortlessly useful.
You don't need a perfect personal operating system to begin. A short About.md, a minimal
Now.md, and an empty Data/ folder are enough.
It works with zero automation. I built it for the future, but you own it today.