Morning industry intelligence brief from X to Gmail

Build me a morning industry intelligence brief that lands in my inbox every weekday at 7am, summarizing what the smart accounts in my niche said overnight.

Trigger: cron, Monday through Friday at 7am in my local timezone.

Inputs the workflow should expose as configurable variables: (1) a source mode that is either "handles" or "list", (2) a list of 15 to 30 X handles to track when source mode is "handles" (start with a placeholder list I can edit), (3) a Twitter list ID when source mode is "list", (4) the lookback window in hours (default 18), (5) the niche label used in the subject line (for example "AI", "DevTools", "Climate"), and (6) the recipient email address for the brief.

Step 1, fetch overnight posts.

If source mode is "handles", iterate over the handle list and for each handle call TwitterAPI.io "Get User Last Tweets" with a sinceTime filter set to now minus the lookback window in hours. Paginate until the returned data is empty or the timestamps fall outside the window. Capture text, tweet URL, createdAt, like count, retweet count, reply count, quote count, view count, isReply, and the author block.

If source mode is "list", call TwitterAPI.io "Get List Tweet Timeline" with the configured list ID. Paginate the cursor until you have collected every tweet posted within the lookback window, then stop.

Watch credits. With 30 handles a run uses roughly 450 credits for the read calls plus the Gmail send. If a single handle's call fails, log it and skip that author for the day, do not abort the whole run.

Step 2, filter.

Drop any tweet where isReply is true unless it is the head of a self-thread. Drop pure retweets (where the text starts with "RT @" or the tweet has a retweeted_tweet block and no original text). Drop tweets posted outside the lookback window. Keep quote tweets, since they often carry the actual take.

Step 3, cluster and rank.

Cluster the surviving tweets by topic. A cluster can be a single big tweet from one account or a near-identical take from several accounts on the same news item. Within each cluster, dedupe near-identical takes: pick the strongest single post (highest authority plus engagement) and treat the others as supporting signal that the topic is trending, not as separate items.

Then pick the 5 most important clusters using a blended score across these three factors:

- Author authority: follower count and verified status of the author or authors involved.

- Engagement: a combined like + retweet + reply + quote score, with view count as a tiebreaker when present.

- Topical novelty: prefer clusters that are not just routine personal posts. A take backed by several independent accounts in the same window counts as higher signal, not as a duplicate to drop. A lone hot take from a high-authority voice also counts as high signal.

Step 4, write the briefings.

For each of the 5 picks, write a 2 to 3 sentence briefing in the form of "what they said and why it matters". Lead with the substance of the take in your own words, then explain why it is worth noting today (new product, shift in narrative, contrarian read on a known story, fresh data point, etc.). Include the author's handle, the direct tweet URL, and the engagement numbers. If multiple accounts are in the same cluster, mention the supporting voices by handle in a single line under the main briefing.

Do not use em dashes anywhere in the email body. Use periods or commas instead.

Step 5, render the HTML email.

Render a clean, inline-styled HTML email designed to be skimmed in under a minute. Structure:

1. A short header with the niche label and the date.

2. A one-line "overnight in one sentence" summary across all 5 picks.

3. Five numbered briefing cards. Each card shows the author name and handle, the 2 to 3 sentence briefing, the direct tweet link, the engagement numbers, and any supporting voices in the same cluster.

4. A small footer with the source mode (handles list size, or list ID) and the lookback window, so I can sanity check what was actually scanned.

Step 6, send.

Call Gmail "Send a Message" with the rendered HTML body, the configured recipient in the To header, and the subject line in the form "[{niche}] morning brief, {date}" (for example "[AI] morning brief, May 28"). Send from my connected Gmail account.

Edge cases. If after filtering there are fewer than 5 quality clusters, send a shorter brief with however many you have and say so at the top. If there is nothing notable overnight, send a one-line email confirming the scan ran and that nothing crossed the bar, so I know the workflow is alive.