# Human Behavior：AI 原生产品分析新方向

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-08-11 00:18
- AIHOT 分数：35
- AIHOT 链接：https://aihot.virxact.com/items/cmsngruln06garohf6tlygspa
- 原文链接：https://x.com/rohanpaul_ai/status/2086849753212682598

## AI 摘要

Human Behavior 推出 v2，获 Paul Graham、Y Combinator、General Catalyst、Vercel 等 500 万美元融资。其 SDK 持续捕获会话、点击、网络调用与错误，AI 智能体可自动分析变化原因并执行操作，如开 PR、提交工单、发 Slack 消息。用户可直接提问“注册量为何下降”，智能体基于完整产品上下文作答并流转至后续任务。

## 正文

This is probably how AI-native product analytics should work.

Dashboards was for humans. For AI agents, they're a pretty bad interface.

That's the direction Human Behavior is taking.

It watches sessions, funnels and errors continuously, works out what changed and why, then hands that context to agents that can act on it: open a PR, file the ticket, post it in Slack, pull the numbers.

Its SDK captures sessions, clicks, network calls, and errors.

You can ask questions like "why did signups drop?" and its agents work across the available product context rather than treating the question as a standalone prompt. The same context can then flow into tickets, PRs, Slack messages, reports, or follow-up investigations.

### 引用推文

> amogh：hey! super excited to announce human behavior v2 ;) we raised $5M from @paulg, @ycombinator, @generalcatalyst, @vercel to put your company on autopilot product ...