Daily Intelligence Brief.
A 9am SMS with yesterday's market/competitor moves, be informed on what actually matters for your business.
End of run
A real SMS lands on my phone at 9am with the day’s intelligence brief for the target vertical: competitor moves, M&A in the relevant space, geopolitical/policy items that touch the supply chain, market signals worth knowing, filtered hard for relevance and explicitly checked against prior briefs so the same story never gets surfaced twice. A longer-form version simultaneously posts to Slack for the team.
Six minutes of compute, $2.20 per brief, running daily since January. Cheaper than a part-time analyst and more reliable.
Challenges it solves
Five sources of signal, one phone notification. The agent runs parallel searches across M&A news, geopolitical news, general Google news, market conditions, and policy news, then synthesizes everything into a single SMS-readable brief plus a fuller Slack post. Same source data, two delivery formats sized for two reading contexts.
Don’t repeat yesterday’s news. Before the day’s searches even
start, Load Previous Context pulls the running memory of what
has been covered this week. After delivery, Update Memory
appends the new items. Today’s brief leads with what’s actually
new.
Relevance over recency. The Comp Query Builder + Deep Research pass narrows everything to the lens of the target vertical, not generic tech news. Random tech news doesn’t make it through. Mentions of competitors, suppliers, regulatory posture, and supply-chain disruptions do.
Daily reliability. It runs on a 9 AM schedule and has for months. The end of every run is a successfully delivered SMS, not a “successfully invoked the SMS API,” because if the message doesn’t arrive, the agent failed.
Under the hood
Daily-scheduled workflow. Loads prior-day memory, fans out five parallel news searches, synthesizes with Claude and Gemini Deep Research, generates the brief in two formats, sends SMS, posts to Slack, then writes the new coverage back to memory before ending. Two models share the work: Claude 3.7 Sonnet for synthesis and writing, Gemini 2.5 Pro for the deep-research pass.
Example output, a real SMS the agent sent.
