How the Green Context Engine Works
What is this?
The Green Context Engine is an automated system that monitors energy and climate news, pulls real data from scientific and government databases, and uses AI to write data-grounded summaries. Every number in every brief traces back to a named, verifiable source. A human reviews and approves every post before it goes live.
The goal is to take the flood of daily energy and climate news and add context that is often missing: how do the numbers in a news story compare to global benchmarks? What does the data actually show? What are the trade-offs?
How it works
- 1
Find and select stories
The system scans 5 energy and climate news sources for new articles matching topics like solar, wind, coal, deforestation, and grid decarbonization. An AI then selects which story from each source is best served by the available data.
- 2
Decide what data to look up
An AI reads the article and decides which databases to query and for which countries or regions. For example, a story about Indonesian deforestation triggers lookups for Indonesia's electricity grid data, forest cover loss, and deforestation drivers.
- 3
Pull real data
The system queries government and scientific databases for actual numbers: electricity generation by fuel type, carbon intensity, tree cover loss, forest carbon emissions, and heating or cooling degree days. It also pulls global and regional benchmarks for comparison.
- 4
Write a brief
An AI writer creates a 200-250 word summary structured around three sections: The story (what happened), The bigger picture (data context), and The tension (the key trade-off). The writer follows learned rules from past editorial feedback to avoid repeating previous mistakes.
- 5
Fact-check
A separate AI editor reviews the brief and checks every number against the source data. It can pass the draft, fix issues directly, or fail it for a full rewrite. After any fix, a read-only verification step confirms the corrections didn't introduce new problems.
- 6
Human approval
A human reviews the final brief in Notion and approves or rejects it. Nothing goes live without this step. Rejected drafts include written feedback explaining why.
- 7
Learn from feedback
When a brief is rejected, the human's feedback is processed by AI into generalized writing rules. These rules are loaded into the writer's prompt on every future run, so the system learns from its mistakes and improves over time.
What makes this different
Unlike typical AI-generated content, every brief is grounded in data from named, verifiable sources. Country-level data is always compared to global or regional benchmarks. A dedicated AI editor verifies every factual claim against the actual source data before a human approves publication.
The system also learns from human feedback. When a brief is rejected, the editor's notes are distilled into reusable writing rules that shape every future draft. The pipeline gets better over time, not just at avoiding specific errors, but at internalizing the editorial standards behind them.
Data sources
Real numbers are pulled from these scientific and government databases.
Global electricity generation, carbon intensity, and emissions data for ~200 countries and economic groups.
US electricity generation by fuel type at national and state level, from the Energy Information Administration.
Tree cover loss, deforestation drivers, and forest carbon emissions by country, from satellite data covering 2000-2024.
Temperature, precipitation, and heating/cooling degree days from the Global Historical Climatology Network, covering 180+ countries.
US solar resource data and energy production estimates from the National Laboratory of the Rockies. GHI, DNI, and PVWatts capacity factor for any US location.
Global solar radiation, wind speed, and historical weather data at 10km resolution, covering any location on Earth back to 1940.
Real-time and forecast carbon intensity and generation mix for Great Britain at 30-minute resolution, from the National Energy System Operator.
News sources
Stories are monitored from these energy and climate journalism outlets.
Environmental journalism covering energy, biodiversity, and climate.
UK-based outlet specializing in climate science and energy policy.
Global solar industry news and market analysis.
Clean energy and electric vehicle news.
Electric vehicles, energy storage, and solar news.
Limitations
- Data from sources may be weeks or months behind real-time events.
- AI can misinterpret nuanced trends or draw connections that oversimplify complex dynamics.
- Coverage is limited to the data sources listed above — the pipeline cannot verify claims against sources outside its dataset.
- Briefs are short summaries, not comprehensive analyses. They are meant to add context, not replace deep reporting.