Amplitude’s proactive product agent Wave uses OpenAI models to find opportunities, produce implementation plans, hand work to coding agents, and measure outcomes. Amplitude says Wave is roughly 80% powered by OpenAI models, including GPT‑6.1 Sol, GPT‑6 Luna and GPT‑Image‑2.5 Flare.
What Wave does
Wave is designed to turn product analytics and user feedback into shippable changes. Teams specify a product area and the goals or metrics they want to improve; Wave combines Amplitude data, session replays, customer feedback and code to identify and prioritize opportunities, propose solutions and technical plans, and coordinate implementation with humans and coding agents. It then tracks results and incorporates those learnings into future rounds.
Amplitude estimates Wave can reduce the time to ship a product improvement to 1–2 days, down from roughly a month. In an internal homepage experiment, Wave’s personalized suggestion cards drove a 160% lift in card-driven sends relative to control after one day.
Which models handle which tasks
Amplitude assigns models to tasks based on capability, latency and cost. GPT‑6.1 Sol supports the opportunity manager and deep code and data investigations. GPT‑6 Luna handles frequent, lower-cost checks like scanning code and small evaluations. GPT‑Image‑2.5 Flare generates mockups and wireframes using session replay screenshots as references.
When an opportunity is ready to build, teams can send it from Wave to coding agents (including Codex) with a technical plan and acceptance criteria to guide implementation.
Homepage experiment: personalizing suggestion cards
Amplitude tested Wave on its own homepage. In a sample of 100 session replays, 80 users typed their own questions while only 12 selected a suggested card. Wave recommended personalizing cards with recent charts, dashboards or the page context, provided an illustrative mockup and a technical plan (including preserving free typing and falling back to generic suggestions if no context is available). After a short review the work was given to a coding agent and A/B tested.
The experiment’s readout shows messages sent via homepage suggestion cards increased 160% versus control after one day. Eric Carlson, Senior Principal AI Engineer at Amplitude, called this an example of how small, otherwise-ignored changes can unexpectedly increase feature adoption.
Model evaluations and trade-offs
Amplitude measures quality, latency and cost for each task and assigns models accordingly. In the Opportunity Investigator evaluation, GPT‑6.1 Sol scored 4.06 out of 5 overall, compared with 3.62, 3.52 and 2.94 for three alternative models. GPT‑6.1 Sol led on all five investigation criteria and had the fastest median investigation time at 54 minutes.
In a separate Data Science Agent evaluation, GPT‑6.1 Sol scored 67.5 out of 100 versus 69.3 for another leading model, while the mean model cost per dispatch was $1.28 for GPT‑6.1 Sol versus $2.63 for the other model. GPT‑6.1 Sol scored higher on tool‑contract compliance, estimation and uncertainty handling, and grounding of claims in evidence.
For high-volume tasks like code scanning and small checks, Amplitude uses GPT‑6 Luna to keep latency and cost low, reserving more expensive reasoning for the tasks that need it most.
On visual generation, OpenAI’s image model scored 80.3 out of 100 for solution wireframes versus 63.9 for another image model; Amplitude evaluated 15 opportunities for communication quality, product fidelity, instruction adherence, text and UI legibility, and visual coherence.
Customer use and automation levels
Customers across industries are using Wave in beta to identify and ship improvements. Wave analyzes session and analytics data and produces a recommendation with a proposed fix, technical plan and acceptance criteria.
Teams control automation levels: most teams send bug fixes directly to coding agents, while new features and larger bets receive human review. Among customer-launched experiments, flags and guides, one in four went live the same day Wave surfaced the opportunity, half went live in under four days, and two in three within a week.
Ambition and outlook
Amplitude’s stated ambition is to make self‑improving products the norm using OpenAI models. As models improve at reasoning and coding, teams will be able to delegate more investigation and implementation work to agents, accelerating delivery of better products. Amplitude reports that agents are already finding and helping deliver improvements in its own product while the human team focuses on other priorities.
Jacob Newman, Principal AI Product Manager at Amplitude, summarized the concept: Wave provides a unified platform with agents that comb through large volumes of session replays, analytics and code to find product improvement opportunities. Skates added that teams may be surprised by how much they can accomplish when their software works alongside them to improve the product.
