Exa API returned 402 errors

Three Exa API search requests failed on September 8, 2026, and left a downstream world-news briefing with no articles to cite. The failed queries covered Ukraine-Russia ceasefire talks, U.S.-China tariffs and Federal Reserve rate-cut expectations. Because the supplied bundle contained only error messages and no usable stories or links, the briefing could not be written under the stated sourcing rules. The response to the failure was procedural: resend the request with successful pre-fetched articles or direct URLs. ### Which searches failed, exactly? The three failed searches were listed as “Ukraine Russia ceasefire talks,” “US China tariffs trade,” and “Federal Reserve rate cut expectations.” Each one returned the same result: Exa API returned 402. Those three topics were the entire contents of the pre-fetched feed described in the source material. No article text, publisher names, timestamps or destination URLs accompanied the errors. ### Why did three 402 errors stop the whole briefing? The instruction set for the briefing required the writer to use only the supplied pre-fetched material, cite key claims and avoid adding outside facts from memory. With no articles in the bundle, there was nothing to quote, summarize or attribute. That constraint turned a search failure into a publication failure. The issue was not that the topics lacked news value; it was that the provided package contained no sourceable reporting. ### What does a 402 result mean in this case? The source material identifies the failure message only as “Exa API returned 402.” It does not include any accompanying diagnostic text, billing notice, quota explanation or service-status detail from Exa. Because no further documentation was included in the bundle, the available facts stop at the error code itself and the effect it had on the workflow: the searches did not return articles. ### Why wasn’t the missing information filled in from elsewhere? The briefing instructions expressly barred that approach. The writer was told to rely only on the pre-fetched content and to include citations for key claims. Under those rules, substituting separate reporting from other outlets would have changed the source base and broken the assignment constraints. The correct response was to say the briefing could not be produced from the material provided. ### What was the immediate response? The output drafted from the failed feed said the day’s world-news briefing could not be generated from the supplied material. It also said any substantive briefing would violate the requirement to use only the pre-fetched content. The same response asked for a resend with successful pre-fetched articles or URLs. It also offered two fallback options: a briefing template or a short “data unavailable” placeholder for publication. ### What happens next? The next concrete step is a new submission containing working source material for the same topics or direct links to source articles. Once those inputs exist, the briefing can be written with citations and within the original format requirements. Any follow-up package would need more than search prompts and error codes. It would need article URLs, publisher material or successfully retrieved pre-fetched results tied to the Ukraine-Russia, U.S.-China and Federal Reserve topics.