← Spark Story

Behind Spark's Reflection Tool

The iPhone Notes Problem

I was the teacher scrolling through iPhone Notes trying to find what a child said three weeks ago.

It was in there somewhere — between a grocery list and a voice memo I never transcribed. The observation was real. The moment mattered. But by the time I found it, or didn't, the conversation with the parent had already moved on.

That's what this tool is solving. Not documentation itself — educators already know how to notice. It's the after. Where does it go? How do you find it again? How do you hold observations across a whole semester without carrying all of it in your head?


What the first version got wrong

The earliest version could generate reflections. It could take a single observation and return patterns, questions, and connections. An educator who was used to blank-page Sunday evenings would have found it useful. I almost shipped it as that.

But single-entry reflection is the wrong unit. The most important pedagogical move isn't noticing one moment — it's seeing across many moments. When you look at three observations about the same child together, things emerge that are invisible in any single entry. An interest that keeps returning. A frustration with a shape. A progression from one kind of thinking to another. That's where the real understanding lives.

So the tool was rebuilt around the cross-documentation view — reflection across the full body of observations, not entry by entry. That change made everything harder to build and much more worth using.


The three lenses — and why they're separate

The AI output is structured into three distinct sections — patterns, questions, and connections — and this wasn't a design flourish. It was a pedagogical constraint.

Early outputs mixed all three into flowing paragraphs that sounded thoughtful but were hard to act on. An educator reading them couldn't tell what was observation, what was interpretation, and what was invitation. The three-section structure forces the tool to stay in its lane — patterns describe what's present, questions hold open space, connections offer direction without prescription.

What was rejected
"Rod demonstrates a growth mindset in the face of technological challenges."
What the tool now surfaces
"Frustration with outputs is a recurring emotional response. There is a progression from building to troubleshooting."

One is a conclusion — it belongs to a professional who has applied a theory. The other is a pattern — something to look at, not something to believe. The difference matters more than it might seem. The tool should not be closing the space where educator judgment lives.


The quiet children problem

One of the things that emerged during refinement was something I hadn't fully planned for: the equity visibility problem.

When you can see all your students in one place — with their observation counts shown as dots beside each name — something becomes very visible very fast. Some children have eight entries. Some have one. Some have none. The child with none is almost always the quieter one. The one who doesn't push for attention. The one who, in a busy classroom, can go weeks without being specifically documented.

The tool now shows this directly in the student list — dots proportional to entries, and a red nudge for any child who hasn't been documented in fourteen days or more. Not as a judgement. Just as a mirror. The professional instinct does the rest.

The student list — equity dots, observation counts, and a red nudge for undocumented children.
The student list — equity dots, observation counts, and a red nudge for undocumented children.

What was added, and what was deliberately left out

The tool grew through use. Each feature came from a real problem that showed up in practice.

The observation form — starter prompts, NEL disposition chips, and a voice memo record button.
The observation form — starter prompts, NEL disposition chips, and a voice memo record button.
📸
Photo and voice memo attachments
Because observations aren't only written. A photo of a child's construction, a voice memo recorded in the moment — these belong alongside the text, not in a separate app.
👥
Group observations
In real classrooms, children don't learn alone. One observation written about three children working together can now be saved to each child separately — each gets their own reflection thread from the shared moment.
Group observation — one entry written once, saved separately for each child with their own reflection.
Group observation — one entry written once, saved separately for each child with their own reflection.
📅
Backdated observations
Educators document after the fact. A moment observed on Tuesday often gets written up on Thursday. The date of observation is now separate from the date of entry — so the record is honest.
Photo and voice memo attachments, backdated date picker, and
Photo and voice memo attachments, backdated date picker, and "Also document for" — one observation saved to multiple children.
🤝
Co-educator access
Reggio documentation is a team practice. A second educator can be invited to view and add observations for the same class. The documentation belongs to the classroom, not just one person's phone.
📄
PDF and clipboard export
Because documentation eventually needs to leave the tool — for parent meetings, for portfolios, for team conversations. One tap copies the full reflection as structured text, or opens a clean print view.

What was left out matters as much as what was added. There are no learning outcomes. No curriculum tags. No developmental milestones to tick. No "next steps" generated by the tool. Those belong to the educator — shaped by what the child does tomorrow, by the relationship that has been building for months. No language model has access to that, and any tool that pretends otherwise is not supporting pedagogical thinking. It's replacing it.


The thing it keeps coming back to

Every decision in this tool was made against the same question: does this support the educator's thinking, or does it do the thinking for them?

The pull quote that sits at the top of every reflection is not decoration. It is the actual constraint the whole thing is built around:

"The educator remains the thinker. This companion offers patterns, questions, and connections to revisit — you interpret what they mean for each child."

What the Documentation Reflection Companion generates is the result of holding that line. Not AI doing the thinking. A companion that has absorbed enough to notice alongside you — and knows when to get out of the way.

Bring your documentation. See what you notice when you look again.

Try the Documentation Reflection Companion →