Every day I hand you three ways to play your part: a game, a story or an action. Pick one. That's your challenge for the day, and it keeps your streak going.

Games are short quizzes and sorting games built on real data. The daily quiz changes every weekday. On weekends the library is open and I suggest a game you haven't played for a while.

Stories are short visual reads on one climate topic, things I've found out since I landed, five to eight minutes each.

Actions are small, specific things you can do this week, drawn from what you've been playing and reading. I suggest one later in the week, once you've done a couple of challenges. You read it, then pledge it or say no and get another. Three pledges a week is the limit. Tap any action on your profile later to add a note on how it went.

One challenge a day is the rhythm. If you want more, you can do a second one. After that, the home page tells you the day is done and I wait for tomorrow. The Companion is always open.

Points. A game earns up to 150 points depending on how well you do, a story earns 100, and an action 30 or 50. Actions earn the least because I take your word for them. Playing several days in a row adds a small streak bonus.

Impact credits. Every 1,500 points becomes an impact credit, about a fortnight of play. Your score never goes down. A credit can be spent with one of our partners on something real, like a native tree or a patch of pollinator meadow. Your profile shows how far you are from the next one.

Because small is what lasts. A nudge is a gentle prompt at the right moment, not a lecture. A tiny action is something you can do in a couple of minutes, so there's no good reason to put it off.

Learning on its own rarely changes what we do. Knowing a fact and acting on it are two different things. That's why I pair the two: you play or read something, then I suggest one small thing to try while it's fresh.

Each small action you finish makes the next one feel easier, and over time it stops being a task and becomes how you see yourself. Someone who checks the label, switches off standby or takes the bag to the shop is already thinking like a person who looks after the planet.

That's the long game here: not one big change, but a mindset built a few minutes at a time. It's also why I keep it to three pledges a week. Easy to do, easy to repeat.

The rapid growth of AI, and the data centres behind it, is a real issue. It can put pressure on energy grids and local water supplies. I've seen it from up there, and the people who built me take it seriously.

That said, we think AI has a real part to play in making climate information easier to get at. The trick is doing it responsibly.

Here's how we try to get the balance right:

We use the smallest model that gets the facts right. Most of my answers are written by Claude Haiku, a small, low-cost model from Anthropic. For harder ones, like follow-ups, the latest figures or things our wiki doesn't cover well, I use Claude Opus, a bigger model.

We keep testing European models. We've run a lot of tests on smaller European models, including Mistral and Apertus, the open source model from ETH Zurich and EPFL. I'd love to tell you they won. But on the same sources they got more facts wrong, so they don't write my answers for now. We'll keep testing them as they improve. Mistral does power my wiki search.

Facts come first. I'm here to share facts, so we use the model that gets them right and explains them most simply.

Our own wiki does a lot of the work. We collect and check our own climate information, so I don't have to lean on what a model happens to remember. We'll keep building it, and every good page means fewer answers need the bigger model.

I try not to answer the same question twice. If someone asks a first question I've already answered, I reuse the saved answer, so no AI model runs at all. For now the words have to match, and saved answers last 30 days.

About the energy. Training a big model takes a huge amount of energy, and that part is already spent before I answer you. Each answer uses much less. But answers add up across everyone using AI, and over time they can pass the training cost. That's why we try to keep our model calls few and small.

We're not perfect, and the field moves fast. But choosing the right partners and tools makes a real difference.

I draw on three sources, and I show you where each answer came from.

First, official figures. We keep a list of dated figures from agencies like the EPA, SEAI and the CSO. When one fits your question, I use it first.

Second, our own checked knowledge base. The team has built and maintains a wiki of climate and sustainability information, sourced, reviewed and validated by us. When you ask a question, I search that wiki, using Mistral.

Third, general AI knowledge as a fallback. If those two don't have the answer, I use what Claude learned in its training. When that happens, the answer is clearly labelled, so you know it's general AI knowledge rather than our checked sources.

If you ask for a national figure I can't confirm, I may offer to look on the official sites. If I find it, I'll show you the page it came from.

So as much as possible, answers are grounded in verified, local information. I can still get things wrong though, so if a figure really matters to you, check the source.

This app isn't trying to grab your attention or keep you scrolling. We know the pressure that comes from constant digital noise, and from the weight of bad climate news. That's not what I want to add to your day.

Our research shows people want simple, trustworthy ways to stay informed, and that short, playful challenges do that without the overwhelm.

So Environmentle is built around one challenge a day, with a cap of two. A game takes about three minutes, a story about six, an action two. That's on purpose. Get something useful out of it, then get on with your day.

To play, I only ask for your first name, so I know what to call you. No password, no phone number. The one time we ask for an email is when you spend a credit, more on that below.

Your points, streak, games played, stories read and actions pledged are kept in a database against that name and nothing else, so they follow you between devices. We use those figures in aggregate, to see how many games were played, how many actions were pledged and how many points were earned across everyone. They're never sold or shared.

If you turn on notifications, we store the token your browser gives us for that purpose only, and you can switch it off at any time.

When you ask me something, your question goes to Anthropic, Mistral or Infomaniak, depending on the question, so I can search and answer. We save first questions and their answers for 30 days, and we keep questions I couldn't find a figure for, so we know what to add. No name is attached, but please don't put personal details in your questions.

Environmentle is a web app. It runs in your browser, and you can add it to your home screen so it opens like any other app, but there's nothing to download from an app store.

That makes it lighter. It takes almost no space on your phone, it updates on its own, and it works on any device with a browser. No account to create, no store to go through.

For a few minutes a day, that's the right fit. A native app would add weight and friction without adding anything you'd notice.

Playing never needs more than your first name. Spending a credit is different: it funds a real project in the ground, so at that point we ask for an email address.

We use it to check the credit and pass it on to the partner you chose, so the tree or the meadow actually gets funded. It also lets us send you news about your impact, like the photo when it's planted.

That's the only time we ask, and the only thing we use it for.