The Ten-Day Dev Loop is a composite story of one customer request moving through three new agents in the fleet. Monday, 4:02pm: a CSM at Acme Corp tags a Slack DM asking for bulk task reassignment. Ten days later the feature is live, announced across six marketing channels, with the customer's quote on the landing page. The old loop would have taken eight weeks. This one took ten days.
What each agent does
The Instant Prototype Agent watches the #product channel, ingests the tagged message, and searches the last 90 days of signals for anything similar. In the story it finds four related signals, reads the design system, and generates a working prototype. At 4:09pm, seven minutes after the DM, there is a draft GitHub PR, a clickable Vercel preview, a Linear ticket, a Notion doc, and a Slack message. The customer clicks the preview from her phone that evening and sends two specific changes back.
The Auto Bugfix Agent runs the same pattern for defects. In the story it ingests a Globex ticket about a hanging CSV export at 2:41pm, reproduces it with a 12k-row fixture, locates an O(n squared) string concatenation, writes a streaming fix, adds a regression test, and opens a PR by 2:53pm. The on-call engineer reviews it over a sandwich and merges at 3:14pm. Total engineering time: six minutes.
The Launch Comms Agent fires on production deploy. It pulls the Linear ticket, the Notion doc, release notes, a Figma export, and the repo's voice config, then generates six channel drafts in parallel: landing page copy, a LinkedIn post, a segmented customer email, an in-app banner, a changelog entry, and an X thread. Total PMM time from draft-ready to published: about 90 minutes, versus a week of writing plus a week of coordination before.
Why it works: handoffs, not steps
The thing that changed is not the speed of individual steps. It is that the handoffs stopped costing days each. Prototype to customer feedback: minutes, not a week. Feedback to updated scope: the same morning, not the next sprint. Merge to launch: 18 hours, not ten business days. Humans still make every decision. They just make them with better raw material, faster. The bulk-reassign feature hit 84% first-week adoption among eligible accounts.
What stays human
Deciding whether a request is worth prototyping, judging the prototype before it goes to the customer, engineering review and merge (always), customer empathy, PMM voice work, and business decisions like pricing and permissions. Two lessons matter most: context is the whole game, since output quality tracks the quality of the context an agent can read, and you should shadow-run any agent for two weeks on real data before it touches customers.
Pick one thing to try this week: wire up the Instant Prototype Agent on a single Slack channel and shadow-run it for two weeks before it forwards anything to a customer.