Ce que PostHog a publié en juin 2026

Why we rebuilt on data warehouse on DuckDB over ClickHouse

PostHog replaced its ClickHouse-based data warehouse with a DuckDB-powered system to address multi-tenancy limitations, lack of query optimization, and tooling gaps. The new architecture mirrors all P…

The best AI observability tools for developers, compared

PostHog published a detailed comparison of nine AI observability tools, highlighting features, pricing models, and integration methods. The article emphasizes the importance of LLM observability for d…

Pricing change detected for Posthog

Pricing updated for Posthog: - Free: allowance changed from 1 project, 1-year data retention to 1 project, 1-year data retention, Community support - New tier: Usage-based (Custom pricing) - Removed: …

Web app

Realtime analytics data exports

PostHog expanded its real-time data export feature to include pre-configured destinations for popular tools, a webhook option for custom endpoints, and SQL-based filtering to refine exported data. The…

Start here

What is a Scout? A technical deep dive

PostHog launched 'scouts'—small agents that monitor product data and emit actionable signals as part of its Self-driving feature. Scouts use markdown-based skills for flexible logic, reducing code dep…

What if your product built itself?

PostHog introduced an AI-driven pipeline that ingests product signals (errors, logs, session replays), groups them into actionable reports, and automatically generates GitHub PRs to fix issues. The sy…

Pricing updated for Posthog: - Free: allowance changed from Usage capped at free tier limits to 1 project, 1-year data retention - Pay-as-you-go: allowance changed from Usage-based pricing after free

Cheapest AI observability tools for developers, compared

PostHog published a guide comparing free tiers of AI observability tools, highlighting PostHog’s 100K LLM events/month free plan and Langfuse’s 50K units/month plan as the most generous. The post emph…

Everything (and everyone) is build mode now

PostHog introduced AI-driven features that allow product managers, marketers, and support teams to build, test, and deploy code directly from Slack or its AI tools without traditional engineering invo…

Replay Vision

PostHog introduced Replay Vision, an AI tool that automatically analyzes session recordings by watching video alongside events to extract visual and behavioral cues like hesitation or attention focus.…

PostHog Slack app

PostHog introduced a Slack app in beta that integrates its AI agent into Slack channels and threads. Mentioning @PostHog allows teams to query product data, request code changes, or open draft PRs dir…

MCP analytics

PostHog released a TypeScript SDK (@posthog/mcp) to instrument Model Context Protocol (MCP) servers with analytics, currently in alpha (0.1.x). The SDK tracks tool calls, errors, latency, and agent in…

Pricing updated for Posthog: - Free: allowance changed from 1 project, 1-year data retention to Usage capped at free tier limits - New tier: Pay-as-you-go ($0/mo) - Removed: Usage-based (Pay-as-you-go

Pricing updated for Posthog: - Free: allowance changed from Usage capped at free tier limits to 1 project, 1-year data retention - Scale: now includes 6 projects, 7-year data retention - New tier: Usa

All social media metrics are bad

Picture this: your company's social media team excitedly informs you that stonks are up this month. You can see the screenshots. They're green. Holy smokes. We are killing it. Social media h…

Distributed tracing

What is AI observability, how it works, and which tools to use

AI observability is a new monitoring category for tracking LLM calls, agent workflows, costs, errors, and output quality, distinct from traditional APM tools. PostHog’s guide details its six-layer arc…

Pricing updated for Posthog: - Free: allowance changed from 1 project, 1-year data retention, free tier limits to Usage capped at free tier limits - Pay-as-you-go: allowance changed from Pay nothing i

We put PostHog in Slack and now everyone's an engineer

PostHog released a Slack app in beta that allows users to tag @PostHog with requests to generate code, fix bugs, or analyze product data. The bot creates sandboxed PRs, runs tests, and addresses revie…

In-depth: DuckDB vs Postgres

In-depth: DuckDB vs SQLite

Evaluations

Prompt Management

PostHog introduced a Slack app in beta that integrates its AI agent into Slack channels. The bot can answer product data questions, execute code tasks, and manage follow-ups in-thread, leveraging Post

Tracing

PostHog introduced Tracing, a distributed tracing solution integrated with OpenTelemetry (OTLP) that allows users to capture and explore traces without vendor-specific SDKs. Currently in alpha, it act…

AI Observability

PostHog introduced AI Observability features to capture traces, generations, and spans from AI and LLM products. The update includes guides for integrating with OpenAI, Cohere, and Anthropic's Claude …

Minimum viable AI observability: what to set up after shipping your first AI feature

PostHog published a guide for small teams on implementing minimal AI observability, focusing on tracing, cost tracking, and error capture as foundational elements. It advises against over-engineering …

How autoresearch found a 3-year-old bug in our query engine

PostHog’s AI agent discovered a three-year-old bug in their ClickHouse query engine where timestamp filters failed to use the primary key correctly due to a timezone wrapper, causing 62% more granules…

Karpathy's Autoresearch found a 3-year-old bug in our query engine (and improved performance by 11%)

PostHog’s AI-driven autoresearch system, inspired by Andrej Karpathy’s method, uncovered a three-year-old bug in their ClickHouse query engine where timestamp filters failed to use primary keys correc…