Provenance Record

Clear, well-structured News piece (530 words, within the 400-1200 News range). Overview/What We Know/Why May/What We Don't Know structure is logical, and the 'Why May' and 'What We Don't Know' sections add useful seasonal caveats without editorializing. Headline, summary, and lead all accurately reflect the central finding.

Both source snapshots read from disk and SHA-256 verified against manifest. source-0.html.gz (Electrek, status 200, hash e322a814... matches): confirmed verbatim — solar 12.8% vs coal 12.2% as first month on record May 2026 per Ember; solar 45.5 TWh record up 17% from May 2025 and above last July's record; solar third-largest US source behind natural gas and nuclear; coal 39.3 TWh April low, 43.4 TWh May, 11% below May 2025; five years ago coal 19.7% / solar 5.4%; March renewables-over-gas US milestone. source-1.html.gz (Utility Dive, status 200, hash a896a289... matches): confirmed solar generation expected to grow 17% over 2025 production this summer and coal to decline about 10% in H1 2026 vs 2025, per EIA Short-Term Energy Outlook published April 6, 2026. Both outlets (electrek.co, utilitydive.com) are on the source allowlist.

Every specific (12.8%, 12.2%, 45.5 TWh, 17%, 39.3 TWh, 43.4 TWh, 11%, 19.7%, 5.4%, 10%, April 6 2026 STEO) traces verbatim to a cited, allowlisted source. The 'Why May' transition sentence about forecasters projecting other sources to out-generate solar over the full year is editorial framing/context consistent with the EIA outlook (solar remains a growing but not year-leading source); only the precise 17%/10% figures are attributed to Utility Dive, which is correct.

APPROVE. All claims, percentages, the TWh figure, and the Fulghum quote trace verbatim to allowlisted Electrek and Utility Dive snapshots. Ember is correctly excluded from the sources array (403 bot-block) while every Ember figure is carried by Electrek. No provenance break, no duplicate, neutral tone. Ready to publish.