Weekly funded-companies agent
The assignment
Build a weekly agent that scrapes last week's funded startups from the web and proposes your way in — a tailored application path.
✳What Vaishnavi shipped
PM Radar is a BD-to-PM job-search pipeline that surfaces newly funded companies each week, complete with round size, amount, and date. Every company card carries a personal status pipeline (Interested, Contacted, In Conversation, Passed) so a user can track outreach at a glance, and opens into a seven-tab dossier that includes cover-letter generation. The board refreshes automatically every Tuesday, with This Week, Last Week, and Archive views for browsing past cycles.
✳Our read on this build
Vaishnavi shipped a deployed web app you can open. This read comes from the artifact and Vaishnavi's own account of it — no scores, just what the build does and does not show.
What it demonstrates
- Fresh input is part of the machine: live search.
- It goes past listing companies into proposing a way in — outreach drafts.
- And it is built to keep running: automation.
Take this awayThe transferable pattern is a three-stage pipeline — fresh signal in, judgment in the middle, a draft out. That is most useful agents, whatever the domain.
✳Why this quest matters
Agents earn their keep when they run without you. The bar isn't 'can AI write an outreach note' — it's 'does fresh data flow in on a schedule and come out tailored,' which is the real shape of an autonomous workflow.
✳What separated the strongest builds
Real companies, funded recently
Data pulled from a live source. A hardcoded list of startups everyone already knows about is a demo of a UI, not of an agent.
An application path that changes per company
The specific round, the sector, the role that funding implies. One paragraph with the company name swapped in is the failure mode here.
It runs without you
A schedule, a cron, a workflow trigger — something that makes 'weekly' true rather than aspirational.
Outreach you would actually send
Notes with judgment behind them. Volume is the easy part and the part that gets people ignored.
No numeric scores are published here. These pages are written reads of what each builder shipped — the build is the point, not a mark out of a hundred.