One year ago, Earthly Insight was only an idea. Today, thousands of people use our text first AI while we donate thousands of dollars each month to environmental nonprofits. Here is what we built, what we learned, and what comes next.

OpeningOne year ago, Earthly Insight was a question, not a product. Could we build a useful AI platform that is honest about its footprint, designed to use less compute, and structured to fund real ecological restoration at the same time?

Twelve months later, we are still asking that question. We are also answering it with action.

What we set out to do

• Make AI helpful without pretending it is clean. Data centers use energy, water, and hardware. We take that seriously.
• Design for less harm by default. Text first, efficient features, careful with context size, no image generation.
• Use revenue to fund restoration directly. A built in commitment, not a marketing add on.

Where we started

• Month 0. A conviction that responsible AI should be simple. Keep it text based. Be transparent about costs and trade offs. Share progress publicly.
• Month 1 to 3. Prototyping and early user conversations. We heard a clear message. Make it fast, reliable, and straightforward. Avoid feature bloat. Be clear about what happens to user data.
• Month 4 to 6. First public launch. Multi model integration for reliability and choice. Clear statement of values. Early donations begin. We started small and shipped often.

Choices that still guide us

• Text first by design. We intentionally do not offer image generation. It costs far more compute per output than text and it does not serve our core use cases.
• Integrate models rather than train new ones. We select the most efficient model for the task. This lets us deliver quality while keeping our own compute footprint lower.
• Right sized context and retrieval. Shorter prompts, focused documents, and careful context windows reduce work for the model while improving results.
• Privacy and transparency. We do not train models on user chats. We keep clear retention policies and are working toward end to end encryption.
• Revenue shares that matter. We donate a meaningful portion of subscription revenue to environmental nonprofits focused on restoration.

What changed in a year

• From idea to thousands of users. We grew through word of mouth, careful feature releases, and a clear mission.
• From first donation to thousands of dollars each month going to environmental nonprofits. As our revenue grew, so did our ability to support restoration.
• From a single model to a multi model platform. Users can pick the right model for the job, which often means lower cost and fewer resources for the same outcome.
• From assumptions to receipts. We now publish clear updates and invite scrutiny. The work is imperfect and ongoing. That is the point.

What we learned

• Simplicity wins. Text first and lean features create better outcomes for users and reduce unnecessary compute.
• Transparency builds trust. Clear language about energy, water, and hardware costs resonates more than claims of clean AI.
• Constraints improve product quality. When you choose not to chase heavy features, you focus on speed, clarity, and reliability.
• People will support a different kind of tech company. Many users want useful tools that also fund restoration. They do not need perfection. They expect honesty.

Milestones worth noting

• Public launch with a clean, text first interface
• Multi model routing with careful defaults
• File uploads with documented limits and rationale
• Clear privacy policy with plain language
• Regular donations to environmental nonprofits, increasing with growth
• A growing library of transparent posts about AI’s footprint and our design choices

Numbers at a glance

• Users. Thousands, and growing
• Donations. Thousands of dollars per month to environmental nonprofits that restore ecosystems
• Product. 100 percent text first, no image generation

Notes on transparency

• Figures are rounded and current as of publication. We update totals on a regular cadence.
• We do not claim to offset our footprint. We fund restoration directly while we also work to cut compute where we can.

What is next

• Improve the user experience. Bring Earthly Insight closer to what people expect from tools like ChatGPT and Claude. Faster responses, clearer reply structure, inline citations, better code formatting, smarter follow ups, improved file handling with previews and summaries, conversation memory controls, and mobile parity.
• Make text even more efficient. Smarter context management, better retrieval, and clearer prompt tools that reduce tokens without losing quality.
• More transparency. Publish regular impact and product updates that anyone can audit.
• Better guardrails. Continued work on data privacy and security, including progress toward end to end encryption.
• A smaller, domain tuned model when the timing and funding align. The goal is lower energy per query without sacrificing quality.

Using Earthly Insight is not an act of environmental purity. It is a choice to support a tool that tries to do less harm and more good. Thank you to everyone who tested early builds, offered tough feedback, and chose to support restoration through your subscriptions. This first year convinced us that practical, honest, and lighter AI is not only possible. It is necessary.

‍