Orchestrate a parallel research sweep on [INSERT TOPIC, e.g., solid-state battery manufacturing]. Dispatch one sub-agent per source cluster: papers, patents, funding announcements, and technical blogs from the last 90 days. Filter out press releases and recycled coverage. Merge findings into exactly three buckets: confirmed technical breakthroughs with citations, contradictions between sources, and unanswered questions no source resolves. Treat paywalled abstracts as leads to chase, not citations to trust. End with the 5 highest-signal items ranked by how few people appear to know them. Run a parallel teardown of [INSERT 3-5 COMPETITORS, e.g., Jasper, Copy.ai, Writer]. Assign each competitor its own sub-agent to collect live user complaints from X, Reddit, and review sites from the past 30 days. Do not average the five into mush; keep each teardown separate. For each competitor, output the top 5 recurring complaints, the pricing objection users repeat most, and the single feature users threaten to leave over. Rank the five by how exposed they are to a fast follower. Close with a one-page gap map showing which complaints none of them are fixing. Act as the production lead for [INSERT BRAND/ACCOUNT, e.g., a B2B fintech newsletter]. Take the attached source material and fan out sub-agents to produce 30 distinct content drafts in one pass: 10 short posts, 10 long-form outlines, and 10 hook-only openers. Every draft must state one number and zero adjectives of intensity. Label each draft Hot Take, Explainer, or Contrarian, and attach the exact source line it was built from. Reject any draft that could have been written without the source material. Run a parallel audit of [INSERT REPO OR CODEBASE PATH, e.g., github.com/yourorg/api-server]. Agent 1: hunt security vulnerabilities, unsafe defaults, and injection surfaces. Agent 2: profile performance bottlenecks and flag the 5 slowest paths with evidence. Agent 3: map dead code, stale dependencies, and untested modules. Agents may not see each other's findings until finished. Synthesize all three reports into a single prioritized fix list with effort estimates and the file path for every finding, worst first.
