Our second stop in Manaus was Arte & Escola na Floresta, a community space close to the city but thirty minutes by 4x4 from the nearest paved road. A patch of place that somehow holds everything at once.
Before anything else, before we opened a laptop or mentioned Tainá or explained what a Raspberry Pi is, we asked everyone to share their name, where they were from, and a movement.
Not describe a movement. Do it. Stand up, take up space, make something with your body that said something about you. The space was open on all sides, the river visible from where we were sitting, and people leaned into it. Elders, parents, teenagers, small children, everyone in the same room and everyone equally ridiculous and equally delighted. The energy after that icebreaker was something you cannot manufacture with a slide deck.
That energy was partly down to Nora Hauswirth, co-founder of Arte & Escola na Floresta, who led the icebreaker and set the tone for everything that followed. The kind of host who makes a room feel like it belongs to everyone in it.
This is how environmental education should feel.
The visit
Arte & Escola na Floresta was our second community visit as part of the Klarna and Milkywire AI for Climate Resilience initiative, following Parque das Tribos earlier in the week. The modules were largely the same: Tainá, bioacoustic monitoring, drone mapping, data sovereignty, and Bumicerts. If you have not read the Parque das Tribos piece, it is worth starting there for the full context on how these tools work and what we were trying to do.
Samuel and Eduardo, our local champions from Parque das Tribos, came with us. So did Marina Mura, an indigenous conservationist who had experienced an earlier version of Tainá and joined us to help facilitate. Together they translated and guided the sessions throughout, which changed the dynamic in ways that matter. The introduction to Tainá did not come from an outside team explaining a tool. It came from people who had used it, had opinions about it, and could speak to it in the same breath as talking about their own relationship with the forest. That kind of handoff is what peer learning actually looks like.
About twenty people joined us over two days. Some were from the surrounding area, others had come from outside, all of them supporters of what this place is and what it stands for. Farmers and agroecology practitioners sit alongside artists and environmental educators. Elders, parents, teenagers, and children all in the same space. Sharing happened constantly, in every form it takes: songs, food, knowledge, stories. Conservation here is not a separate subject from culture or art or education. It is all the same thing, expressed differently depending on who is holding the pen or the seed or the phone.
Into the river
On the second day the kids led us into the water.
Not to the water. Into it. They had been watching us set up equipment and at some point the dynamic shifted and they were the ones deciding where to go and what to do. They chose the tree on the far bank. They swam out to anchor the AudioMoth sensor to it. Standing chest-deep in the river watching a child secure a bioacoustic monitor to a tree they had selected themselves, you understand something about what democratising technology actually means. It is not about making tools cheaper or simpler. It is about who gets to make the decisions.
We also flew a drone over areas that mattered to the community, places they wanted documented. Before we showed anyone the imagery, we asked the kids to draw their version of those same places. What they thought the land looked like from above, what they knew was there, what felt important enough to put on paper. Then we put the drone visualisation up next to their drawings. The room got quiet in a good way. Two ways of seeing the same place, both of them real, sitting side by side.
Phones, parents, and what attention is for
One of the first things parents raised was phone addiction. Young people staring at screens. The disconnection from the world right outside the door.
It became one of the most honest conversations of the trip, partly because the teenagers being discussed were sitting right there. Parents were not asking us to take phones away. They wanted to know whether a phone could be turned toward something that mattered. Whether the same device that pulls a child’s attention inward could point it outward at the forest, the river, the plants their grandparents know by name.
Tainá is not an answer to that question. But it is a place to test it. A phone becomes a tool for noticing when you use it to photograph an insect or identify a plant or log an observation that goes into a record that belongs to your community. We did not make any promises. But we watched teenagers at the riverbank crouching over something they had just noticed, and the parents watched too.
Who owns the knowledge
The data sovereignty conversation here went deeper than anywhere we have been before. An elder in the group said that for her community, the experience of having knowledge extracted without consent is not abstract. Plant knowledge, cultural knowledge, ways of making and growing and living that took generations to develop, taken without credit and without return. The Raspberry Pi on the table was interesting to her not as a piece of hardware but as a proof of concept. A tool designed so that extraction is not architecturally possible.
We talked about how most platforms work. Data leaves, decisions are made elsewhere, and the people who generated the knowledge have no say in what happens to it. It is not just a privacy setting. It is a different theory of who technology is for.
The group drew clear lines on what should be public and what should stay close. Biodiversity observations, traditional recipes, medicinal plant knowledge, seasonal and environmental data: public, because knowledge becomes more valuable when it travels and teaches. Personal health data, children’s data, precise locations of significant sites, ritualistic plant knowledge, the chatlog itself: private, because some things belong to the community and the technology should reflect that, not work around it.
Designing the assistant
We ran a session we had not done at Parque das Tribos: letting the group redesign Tainá’s personality from scratch.
First we showed them two versions we had built to make the point that this was actually possible. One Tainá told bad dad jokes. The other was melancholic and dramatic, sighing about the state of the forest (and life in general). The room found both of them funny and then got serious about what they actually wanted.
What came out: happy, curious, sincere, friendly, focused, and provocative enough to make you think. Not a tool that delivers answers. A presence that asks questions back. A community that makes things together, that blends art and agriculture and education in the same breath, knows what it wants from an AI assistant. They did not take long to decide.
What they did not want: messages too long, tone too formal, responses that feel like a form you are filling out. And the language issue. Tainá does not yet understand how people here actually speak, the dialect, the rhythm, the way things get said in this community. But this is not something we can patch on our own. AI speech and language models are simply not built for most indigenous and regional languages and dialects yet. The gap is systemic, and it reflects who current AI systems were built for and who they were not. What gives us hope is that communities like this one are now part of the conversation, and that changes what gets built next.
What stayed
We left the kit behind: the Raspberry Pi with Tainá running on it and the sensors. The community is planning their own BioBlitz events now, on their own schedule. Samuel and Eduardo at Parque das Tribos are doing the same.
Two communities, two devices, two growing bodies of observation that belong to the people who made them.
The river class was not in the plan. The kids leading us into the water was not in the plan. Neither was the conversation about extraction and historical harm, or the thirty minutes spent arguing about whether Tainá should be allowed to make bad jokes. None of that was in the plan, and all of it mattered more than what was.
Technology finds its shape when it meets people who already know who they are. Arte & Escola na Floresta knew exactly who they were. Our job was to show up, stay curious, and try to be worthy of the room.
We are grateful to Nora and the whole Arte & Escola na Floresta community for having us. For being so open, so generous with what you know, and so willing to teach us things we did not know we needed to learn. You made the work better just by being who you are.
I find what you guys are doing with Gainforest and the learnings that you are sharing with the wider community of regeneration to be an absolutely essential and a precious gift. This is more of what we need in our bioregions of practice…networks like Gainforest being developed in accompaniment and service to the place. True partnership 🦋