
engineering leader, Wave at @Amplitude_HQ. figuring out how self-improving products actually work. recovering functional programming enthusiast. he/him
Bay Area, CA
Joined November 2019
- Tweets161
- Following232
- Followers79
- Likes291
As a “growth person” by trade, I’m intrigued by self-improving products… and very skeptical of the hype.
We have plenty of ideas. The hard part is knowing which ones deserve engineering time. A drop in conversion tells you where to look, but there’s a lot of digging between that chart and a change you’re willing to ship.
Wave is really good at that digging. It brings customer behavior, replays, feedback, and code into the same investigation, so you can inspect why it’s recommending something. And it follows the change through to measurement, which is where plenty of convincing ideas fall apart.
Watching a customer’s eyes light up because it found something urgently worth building is what gives me confidence in this.
The Wave team has been cooking. Proud to be part of it. We launched today!
Introducing Wave.
The world's first AI that reads your customers' mind.
$1.47 trillion will be spent on software this year, but only a fraction of software features are used by customers.
AI can build anything you want, but it doesn't know what your customers need.
Wave fixes that.
2
8
279
I’m an engineer building Wave, @Amplitude_HQ’s proactive product agent for helping teams build self-improving products.
This diagram gets at what we’re trying to build. The pod sets the outcome, shares the context, and defines the guardrails. Wave senses what’s happening, helps prioritize opportunities, and carries work across product, design, and engineering.
Even a small customer problem can need all three, and lining them up is often harder than it should be. Wave can keep the same thread intact from customer evidence to a proposed design to a code change. The PM decides what’s worth doing, the designer shapes the experience, and the engineer owns what ships.
What I want is for Wave to take on more of the coordination and execution, so PMs can spend more time talking to customers, designers on user research and design systems, and engineers like me on product architecture and long-term technical decisions.
Wave takes on more of the execution load, so product teams can spend more time where their judgment matters.
That’s the kind of agent I’d want on a team, and it’s what we're building with Wave.
Agent observability tools tend to overcomplicate things. Amplitude is treating this data simply as product behavior (prompts, evals, traces, tool calls, queryable alongside your standard Amplitude events).
Congrats to the team for the launch🎉
Today we're launching Agent Analytics
Every team shipping an AI agent has the same blind spot. Offline evals pass, you ship, and then you have no idea what's happening in production. AI fails silently. Users ask a question and get different answers. They all look 'engaged' in a classic dashboard. You don't know who got a great response and who got a terrible one.
Agent Analytics solves it:
- Every session scored out of the box on task completion, response quality, friction, safety, and negative feedback
- Topic clustering across thousands of conversations, so you know if a failure hits 1 user or 10,000
- Eval agents that watch for regressions, and if you want will file a Linear ticket or the pull request themselves
- Agent quality sits next to product data, so 'payment scheduling fails 31%' becomes 'which renewals did that cost us?'
The Economist got their agent to a 96.9% task success rate and cut weekly failures 84%.
Included on every plan. Free tier included. amplitude.com/agent-analytic…
5
305
ok this is the good stuff. @Jacob__Newman pulled together some early results we've seen running Wave on Amplitude's products and website. go read this one 🌊
(1/6) i'm the pm on Wave at @Amplitude, our agent for self-improving products. for the past few months we've had it running on our own product. it finds problems, ships fixes, and measures whether they worked. here are some highlights:
1
4
301
More on Wave:
amplitude.com/wave
amplitude.com/docs/wave
16
Kelson Warner retweeted
We're now regularly shipping new features again with @statsig
It's amazing to see how excited customers are by the product!
Kelson Warner retweeted
You can't just give a model the ability to run CDP commands and expect it to perform well in production.
Real browser agents need a harness that provides security, identity, caching, credential brokering, and memory.
Read about how we think about the browser agent harness.
This quoted post is unavailable.
Kelson Warner retweeted
Introducing llmtrack.org, the largest catalog of brand visibility across GPT, Claude, Gemini, and Google AI Overviews.
We collected 44k+ prompts to track which brands AI models recommend across every category.
Agents that can read the web and the product behind it. The loop is closing. Stoked for @Amplitude_HQ , and your agents, to have an official launch partner skill on Browse.sh.
Today, we're launching 4 official partner skills on Browse.sh to improve agent capabilities:
- Get an inbox with @agentmail
- Make any payment with the @link CLI
- Deep research people & companies with @ExaAILabs
- Analyze product insights with @Amplitude_HQ
4
172
Kelson Warner retweeted
Today, we're launching 4 official partner skills on Browse.sh to improve agent capabilities:
- Get an inbox with @agentmail
- Make any payment with the @link CLI
- Deep research people & companies with @ExaAILabs
- Analyze product insights with @Amplitude_HQ
We @Amplitude_HQ shipped an AI agent that reads your codebase and sets up Amplitude in one command.
Here's a little story about what broke along the way.
Super proud of our team for everything that went into launching Amplitude as an early provider for Stripe's new projects.dev. Fully agentic provisioning _AND_ payments!
Just released provisioning Amplitude through Stripe Projects.
It's a great use case of of Amplitude's new CLI setup where you can sign up, manage, use, and pay for an account fully headless.
projects.dev/
Kelson Warner retweeted
Just released provisioning Amplitude through Stripe Projects.
It's a great use case of of Amplitude's new CLI setup where you can sign up, manage, use, and pay for an account fully headless.
projects.dev/
Just launched some great new providers in Stripe Projects (projects.dev):
@huggingface
@Cloudflare
@OpenRouter
@firecrawl
@flydotio
@Amplitude_HQ
@mixpanel
@inngest
Many more coming later this month.
Kelson Warner retweeted
we’ve shipped a bunch of incredible new tools + skills using @Amplitude_HQ's mcp and i wanted to highlight a few killer use cases i’ve been running as a pm/engineer/builder!
/replay-ux-audit + /debug-replay 👀🎥
we added two new session replay tools (list_session_replays and get_session_replay_events) which allows you to more effectively search across your session replays AND provides the agent access to the underlying events and rrweb data.
this allows your agents to query session replays then analyze what’s happening in your replays.
/monitor-ai-quality + /analyze-ai-topics 🤖📊
if you’re using our new agent analytics product, we’ve added mcp tools to query your LLM sessions, spans, conversations, and tagged traces.
this enables your agents to easily browse your agent interactions, synthesize wha the top error categories are or main topics customers are talking to your agents about, then quickly iterate in your ADE/IDE on agent + prompt improvements.
/review-agent-insights 💡
if you’re a heavy amplitude agent user, we’ve added a get_agent_results tool that lets you query all of your amplitude agent previous conversations and the artifacts attached to them.
so if you have a bunch of dashboard and session replay monitors, you can stack workflows on top to triage and fix those issues.
/diagnose-errors 🚨
fun fact: if you’re using autocapture, we track a bunch of rich web vital data like network requests, status codes, logged errors, and error clicks data as different amplitude event types.
so you can automate workflows using an agent to auto-investigate web failures then fix those issues.
get_flags, create_flags, update_flag 👨🏻🔬🧪
if you’re using amplitude for feature flagging, you can create, edit, and retrieve your feature flags all directly via MCP instead of having to manually configure in the UI anymore.
we’ve been using this to scale out the number of experiments we’ve been running internally!
give these a try and lmk if there’s other tools or skills we should be adding! 👀
I just bought tickets to Wax + DJ Hoppa @ The Independent ticketweb.com/event/wax-dj-h… via @TicketWeb
18
Just got my ticket for @vercel Ship - claim yours!
vercel.com/ship/ticket/8Pw70…
1
32

















