Amplitude runs a proactive product agent called Wave that ships product improvements in one to two days, according to a customer story OpenAI published on November 12, 2026 S¹P⁵. The 1-2 day figure is a vendor's own claim with no error bars, no released code, and no independent confirmation S¹.
My read: The 1-2 day figure is the kind of headline number that sells a customer story, and OpenAI has every commercial reason to promote it. What I would watch is whether Amplitude publishes its own engineering blog with before-and-after cycle times. Until then, this is a vendor case study, not a measured result. The interesting part is the loop: Wave decides what to build, then checks whether it worked. That full-cycle framing is where agent stories are heading.
Wave, built primarily on OpenAI models, does three things according to the case study: it surfaces feature recommendations, helps teams ship changes with coding agents, and tracks the impact of those changes P⁵. The OpenAI product named in the story is Codex P⁵. Amplitude is a North American technology company P⁵.
The case study is one of several OpenAI published on the same day, alongside stories about Gondola, Holiday Extras, and Gilbert + Tobin P³. The batch release fits a pattern of OpenAI promoting enterprise adoption of its models.
Wave's three-step loop compresses manual product work
The three-step loop maps to a product development cycle that most teams run manually across days or weeks. Wave identifies what to build, which implies it analyses user behaviour data, Amplitude's core product, and surfaces feature recommendations. It then helps teams ship changes using coding agents, and finally measures the impact of those changes P⁵.
The case study does not specify which OpenAI models power Wave, beyond naming Codex as a product P⁵. It does not say what percentage of Amplitude's improvements ship in the 1-2 day window, or whether the figure represents best-case or typical performance S¹.
The case study leaves key metrics unspecified
The story rests on a single source: OpenAI's customer story page S¹P⁵. Google News carried the same article through an aggregator link, adding no independent reporting S². OpenAI has a direct commercial interest in promoting Amplitude's success as evidence that its models deliver business value.
For a product team evaluating whether an agent like Wave could compress their own shipping cycles, the missing pieces are concrete. The case study provides no baseline cycle time, no sample size of shipped improvements, no error rates from the coding agents, and no breakdown of which steps are automated versus human-reviewed. It also does not indicate whether Wave is available to external customers or remains an internal Amplitude tool S¹.
For a product manager running a SaaS roadmap, the practical shift is that an agent could take over the work of triaging user behaviour data into feature recommendations and drafting the corresponding code, turning a process that currently spans a sprint planning meeting and a separate engineering ticket into a single automated pass P⁵. The next checkable milestone is whether Amplitude publishes its own technical writeup of Wave's architecture and measured outcomes. OpenAI's customer story page, last updated November 12, 2026, is the only public record so far P³.
Sources: S1 — How Amplitude ships product improvements in 1–2 days with OpenAI · S2 — How Amplitude ships product improvements in 1–2 days with OpenAI - Ope · P3 — Customer stories | OpenAI · P4 — openai/gpt-oss · P5 — How Amplitude ships product improvements in 1–2 days with OpenAI | Ope · P6 — TCL606/WAVE
Written from 6 sourced items, 5 of them primary.
