Instinct's Flight Problem and LangChain's 64% Cost Cut

Noah Shinn is rebuilding Instinct's "boring" core use cases (flights, hotels, customer service calls) even as the company chases flashier autonomy wins, while LangChain quietly published the exact kind of result that justifies an agent harness investment: a 64% cost cut with no quality loss on real coding tasks. ### Instinct - Rebuilding the Flight and Hotel Experience Instead of Chasing New Autonomy Wins **Source:** https://x.com/noahrshinn (multiple posts, Oct 7 & Oct 10, 2026) **Credibility:** High (founder, first-party product commentary) **What happened:** Noah Shinn posted a short series on Instinct's product priorities this week. The headline number: more than 20% of Instinct's users had the agent help with flights last month, and the team just rebuilt that flight experience from scratch. On the same day, Shinn also posted about extending "Instinct Concierge" so the agent can make customer service calls on the user's behalf, going beyond the original scope of simply handling the call. Then, this morning, he announced a new hotel booking experience built around a specific philosophy: "You shouldn't have to work around someone else's app or website to get something done. The interface should be built for you, every time." **Key capabilities:** - Flight booking/assistance used by over 20% of Instinct's user base in the last month — the clearest usage number Instinct has shared publicly - Instinct Concierge now extends to making calls on the user's behalf for customer service tasks, not just navigating menus - New hotel booking experience, explicitly designed to avoid routing users through third-party apps/websites - Pattern across all three posts: Instinct is choosing to go deep on its "boring," high-frequency use cases rather than only expanding into novel categories **Why it matters for PMs:** This is the tension every agent product eventually hits: the sexy pitch is "autonomous AI that does anything," but the usage data pulls you back to a handful of unglamorous, high-frequency jobs. Shinn is explicitly naming that tradeoff ("we spend a lot of time on things that go well beyond classic use cases, but the classic use cases still have to work"). The 20% flight-usage number is also one of the only concrete adoption stats we've seen from Instinct, and it's a signal for the open question about how much autonomy and delegated action users actually want from a personal agent: they want it to handle logistics, not have philosophical reach. **Critical questions:** - Is 20% of users touching flight booking a strong signal of retention, or a one-time novelty action (most people don't book flights weekly)? - What does "rebuilt from scratch" actually mean here — was the original flight flow broken, or is this positioning a UX problem as a technical one? - If Instinct Concierge is now making calls autonomously, what's the error-recovery path when it books the wrong flight or misstates something to a human agent on a support call? - Does building a custom interface for every vertical (flights, hotels, customer service) scale, or does Instinct end up re-implementing travel-booking UX that Expedia/Kayak already solved? **Action you could take today:** If you're building an agent product, pull your own usage data and check: are users concentrating on 2-3 "boring" jobs-to-be-done, or spreading thin across many features? That ratio should drive your next quarter's roadmap more than your competitive positioning does.