Human–AI collaboration
- Draw clear lines between human judgment and model output.
- Turn ad-hoc prompting into reusable flows and review loops.
- Express roles, tone, ethics, and visibility as explicit "settings."
Fragment Practice — a quiet systems & human–AI work protocol studio.
We help people design human–AI work protocols, document structures, and everyday rhythms so that AI improves their work without eroding judgment, safety, or shared context. Our raw materials are the fragments you already produce: notes, 1:1s, meetings, chats, logs, drafts, and unresolved questions.
From those fragments, we co-design quiet operating systems for work: simple, repeatable ways that people and AI think together, make decisions, and keep track of what matters—without turning everything into dashboards or surveillance.
This page is a concise English overview for founders, leads, researchers, and practitioners exploring collaboration. Detailed service descriptions and many examples currently live in Japanese.
Who this is for — people already thinking seriously about their work
Most people who contact us are not asking for "AI productivity hacks." They are already carrying complex work and want clearer structure, better shared reasoning, and safer ways to use AI inside that reality.
- Founders and early teams trying to keep their thinking coherent while shipping, hiring, and talking to users.
- Engineers, researchers, and designers who manage long-lived systems and need better ways to capture decisions and assumptions.
- Teams introducing AI into knowledge work but wanting to keep their own language, ethics, and boundaries intact.
- Leaders facing ongoing, high-stakes decisions (governance, security, product, care) and wanting a calmer way to track them.
- People who prefer small, structural changes over large transformation programs.
What we are — a studio for structure, protocols, and quiet reasoning
Fragment Practice sits at the intersection of systems design, writing, and applied AI. We treat documents, conversations, and AI interactions as one knowledge system—and design the structures that help them reinforce rather than undermine each other.
Documents & work rhythms
- Design note and decision structures that survive hiring and handovers.
- Connect notes → AI summaries → actions → documentation.
- Reduce drift between text, tools, teams, and what people think is happening.
Governance & operational clarity
- Align AI usage with security, risk, and BCP realities.
- Handle sensitive data via clear rhythms and boundaries.
- Replace constant "firefighting" with quiet, preventative structure.
What we do — Services, Research, and Fragment
Our work moves across three connected lines. Many collaborations start in one and naturally touch all three over time: applied consulting (Services), structural research (Research & Writing), and our text-first notebook (Fragment / fragment.place).
Services — applied collaboration
We work directly with you to redesign workflows, AI usage, and document systems around the work you actually do, not an idealized org chart.
- Map existing fragments and surface real constraints.
- Prototype small protocol changes using tools you already have.
- Co-design sustainable rhythms and decision flows.
Research & Writing — structural notes
We translate recurring project patterns into reusable frameworks and language. This is where Fragment System, Prism Protocol, and QFS live and evolve.
- Cross-project pattern extraction and framing.
- Documentation of layers such as Prism, QFS, Health & Rhythm, and Streams.
- ZINEs, essays, and applied research logs.
Fragment — the notebook model
Fragment is a calm, AI-native notebook built on Markdown plus small YAML blocks. It is the page model we use internally and with some collaborators.
- One page ≈ one Fragment: a unit of context and intention.
- Prism / Scene / Flow metadata for AI-aware reading modes.
- Public examples and a prototype UI are available at fragment.place.
Typical situations — what people bring to us
You do not need a polished brief. Most collaborations start from a small, uncomfortable fragment of reality:
- "Our decisions live across slides, chats, and people's heads. AI makes it faster, but not clearer."
- "We want to use AI with sensitive work, but our security and BCP people are understandably cautious."
- "New teammates cannot see how we think, only what we decided."
- "We tried dozens of prompts and tools. Some worked, none became a shared protocol."
- "We are juggling work, care, and health. We want AI to help without turning life into a KPI dashboard."
How we work — small loops, structural change
We prefer short, concrete loops over large programs. A typical collaboration moves through three repeating phases:
Listen & map fragments
We read what already exists—notes, logs, drafts, Slack threads, internal docs—and map the actual decision points, bottlenecks, and fragile places in your system.
Propose & test protocols
We prototype the smallest useful change: new note structures, meeting flows, Prism roles, checklists, or AI interaction patterns—ideally inside tools you already use.
Write & embed
We convert what works into durable templates, internal docs, or reference pages colleagues can return to—forming a quiet OS that can be extended later.
Principles — the constraints we design within
These principles shape how we write, design, and collaborate. They are deliberately simple and operational.
Quiet rhythm first
Good reasoning needs slack and safety. We reduce cognitive noise and alert fatigue instead of adding more dashboards, bots, or mandatory rituals.
Fragments → structure
Half-written notes and scattered chats often contain the real signal. We treat them as raw material for structure, not as failures of discipline.
Clear human–AI boundaries
Governance begins with boundary setting. We make it explicit where AI may help, where it must not, and when humans must pause, read, and decide.
Tools as partners, not replacement
We aim for tools that extend attention and care, not replace them. Protocols should make it easier to be thoughtful, not easier to disengage.
About Fragment Practice
Fragment Practice is led by Yasuhiro Shinsho, a Human–AI Work Protocol Architect and consultant based in Takamatsu, Japan. His background spans information engineering, cybersecurity consulting (KPMG / NRI Secure), and long-term experiments in note-taking, governance, and AI-assisted writing.
Since around 2012, the threads behind Fragment System, Prism Protocol, QFS, and Quiet Systems have grown through research on networks and access control, incident response and IT-BCP, internal security programs, and everyday attempts to keep work livable under pressure.
The studio now focuses on a simple question: How can we let people and AI think together, without losing ourselves? Most of our work is small in scope and deep in texture—one team, one set of documents, one set of recurring decisions at a time.
How to start — a small fragment is enough
If you're considering working together, you don't need a full project scope. One of the following is usually enough for a first conversation:
- A document or folder that feels important but hard to maintain.
- A recurring decision or meeting where AI is involved and you're not sure it is helping.
- A tension between security / governance requirements and how people actually work.
- A personal or team rhythm question (health, care, remote work) where AI might play a role.