From hackathon teammates to co-founders: the climate-reporting startup we built on evenings and weekends

Context

Three teammates, one deadline

We didn’t start with a market. We started with three people who had already shipped seven winning prototypes together.

David Harvey, Fernando Rabasco and I met in a Discord channel for web3 hackathons, then kept meeting in person — three hackathons across Europe, seven first prizes between us. The pattern never changed: pick a hard problem, split the work, ship in 48 hours. By 2024 the question was what we would build with more than a weekend.

The answer came from a regulation. The EU was phasing in mandatory sustainability reporting — the Corporate Sustainability Reporting Directive — from large companies down to listed SMEs. Large companies have ESG teams and consultancies. The small and mid-sized businesses that make up 99% of EU companies have neither: the software was built for enterprises, and the consultants were priced for them too.

Research

Work backwards from the report

Before drawing a single screen, we wrote down the report we wanted the software to produce.

We read the CSRD and ESRS texts and studied published reports from Ørsted, Yara, Philips and AkzoNobel to see what “compliant” looks like on paper: twelve sections, a long list of data inputs behind each. Then we designed a simpler method for companies without a sustainability team — and that method became the feature list.

1. Materiality first
   Pick the five to ten topics that matter for this business, and extract only the relevant parts of ESRS.
2. Use what you have
   Utility bills, HR exports, existing policies — data the company already tracks.
3. Phase the hard parts
   Scope 1 and 2 emissions first; Scope 3 and the supply chain later.
4. Iterate
   The first report doesn’t have to be perfect. It has to exist.

Two hypothesis personas, modelled on real companies, framed the research: a head of sustainability at Catawiki, stuck on Scope 3 emissions from shipping auctioned objects, and an ESG officer at Busuu measuring the footprint of its cloud infrastructure. The competition — climate-accounting software, Microsoft’s Emissions Impact Dashboard, consultancies — wasn’t built for a 150-person company.

Product design

A guided path to a compliant report

Four steps, in the order a small company actually experiences them.

The dashboard, tasks and data screens above carry the first three steps. Two more ideas did the heavy lifting. Integrations made “use what you have” real: QuickBooks for energy, SAP for materials, Workday for workforce demographics — data pulled in instead of typed in. And AI sat where a consultant would: ask in plain language how much carbon you emitted and why, get a chart, a breakdown and recommendations for the next report.

We deliberately didn’t build an MVP first. A high-resolution Figma prototype, a demo video and a live website were enough to test the story with real people before a line of production code; user tests on the clickable prototype were the next step.

The pitch

A deck that opened doors

The deck wasn’t only an ask for money. It was the fastest way to find out whether the idea held.

Fourteen slides: why now, a €44 billion ESG-services market by 2028, what Atmos looks like, pricing at a fraction of what a company saves on consultants, and the road from customer discovery to product-market fit. I planned and designed it with David; every number came from the research above.

Two months in, we applied to Y Combinator with it. The application forced us to write down every assumption, and it became the best planning document we had. Then doors started opening: David was invited to YC’s Startup School Europe in London, conversations at Devcon turned into incubator introductions at Santander and Protocol Labs, and we began product discovery with an ESG manager at Allianz.

The CPO hat

What a CPO does in a company of three

In a team of three every hat is shared, but someone has to own the product, the plan and the story. I owned all three.

• Product: research → personas → flows → high-res prototype → demo video → website copy → deck design.
• Plan: what we would show at each milestone — mockups, demo video, website, application, investor conversations — across seven months of evenings and weekends.
• Fundraising: with David — which numbers to lead with, which slide answers which investor question, and what to change after each round of feedback.
• Build: stack decisions with Fernando — Next.js and Tailwind, FastAPI and PostgreSQL, Vertex AI and OpenAI — so the design could become code without translation.

Where it landed

Why we stopped, and what we kept

In early 2025 the EU rewrote the rules for the very customers we had designed for.

The Omnibus package pushed CSRD deadlines back and took most small and mid-sized companies out of scope altogether — the SME that had to report next year suddenly didn’t. Staying with SMEs would have meant redesigning the product around a moving target, hiring regulatory experts and working much more closely with policy-makers. Moving up to large enterprises would have meant far more competition and far more fundraising pressure. For three people with full-time jobs, neither was an evenings-and-weekends decision, so we made it deliberately: we chose not to pursue Y Combinator. We did rebuild heyatmos.com around a free CSRD reporting toolkit early that year, to keep CSRD experts and consultants in the conversation.

Atmos never took investment and never charged a customer — by design. What it gave us was proof that the three of us could go from a hackathon table to a company-shaped thing: a researched problem, a product people recognised, a deck investors wanted to talk about. It is the clearest example of the work I like most — deciding what to build, why, and how to tell it — with nobody upstream to hand the decision to.

Reflection

What I took with me