We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Big Ideas
AI leverage compounds when teams productize their context. Sachin Rekhi’s “Compounding OS” shifts AI from individual productivity to team-wide productivity that improves with use. His playbook is to standardize an agentic platform, build a shared skills library or marketplace, and connect AI to repositories, the design system, data layer, and code so it can contribute to a growing company brain. Apply this by taking one recurring PM workflow and turning its prompt, context, outputs, and review criteria into a reusable skill; improve it from failures instead of leaving expertise in one person’s chat history.
Reliability must be specified, not inferred from demos. Hiten Shah’s warning is practical: one great run proves only that a task could work, while a system that succeeds 90% of the time has roughly a 59% chance of succeeding on all five independent runs. Build evals from normal, messy, and known-trouble examples; define what “good” means; rerun the same set after changing prompts, models, context, or tools; and save important failures as regression cases. For agents, evaluate state changes, rule-following, recovery, and cost—not only the final answer—because the full workflow, not a public benchmark, owns the behavior. Retry, refusal, escalation, and help-seeking thresholds are therefore part of the product specification.
Tactical Playbook
Make stakeholder input decision-grade. Shreyas Doshi’s framework rejects seniority as the shortcut: input quality depends on who you ask, how you frame the problem, whether you know the real goal, whether you listen without filters, and whether you understand the reasoning behind the input. Before asking for opinions, write the decision and desired outcome; select people with relevant context; ask for reasoning and constraints rather than a vote; then replay what you heard and identify what evidence would change the choice.
For cross-sell, start with a trigger—not a campaign. Validate the adoption journey with current adopters and comparable non-adopters, then define the buyer, problem, trigger, value proposition, and positioning; confirm whether the buyer is even the same role across products. The first test should target one observable trigger and one adjacent use case. Assign the hypothesis and message to product marketing, the conversation to sales or CS, and the activation event to product; use a phased rollout or holdout, tracking exposure, activation, time-to-expansion, and support burden.
Case Studies & Lessons
Skydio’s lean PM model ties platform investment to customer outcomes. Its roughly 20-PM organization combines a core platform team with vertically focused PMs; product leaders must get field signal, inspect data in “product data day,” and pair with engineering. Its DFR business reports 55,000 911 responses per month and 25 million Americans within two miles of a dock. The outcomes dashboard tracks whether the drone arrived first, response time, useful information, and calls where the drone prevented an unnecessary officer dispatch, with customers reporting the data. After years of R&D without delivered value, Skydio made DFR the number-one goal; only after X10, the dock, and remote-operations software created customer value did it expand investment elsewhere. Make the wedge’s outcome metric the gate for adjacent bets—not internal enthusiasm or platform completeness.
Career Corner
Nontraditional paths can compound into product leadership. Skydio’s product leader has a history degree, military experience, self-taught software and drone experience, and entered the company by building customer success from zero before moving into product. The actionable signal for career changers: seek roles where you can own a customer problem and show operational and technical learning, not just acquire a PM title.
Tools & Resources
- Evals 101: Hiten Shah’s free live session is Friday, August 14 at 10 AM PT.
- Compounding OS webinar: Sachin Rekhi’s free product-leader session is August 20 at 10 AM PT and covers the three-step playbook.