How we built user behavior analysis with multi-modal LLMs (in 5 not-so-easy steps)
Beta10 months agoUnverifiedAdded Sep 24, 2026
PostHog released Session Summaries in beta, a tool that uses multi-modal LLMs to analyze user session recordings and highlight issues automatically.
Topics: LLMs, Observability
More PostHog releases
Every PostHog release8 learnings from 1 year of agents – PostHog AI
PostHog AI is an integrated agent that can access product data, write SQL queries, manage feature flags, and perform multi-step analyses of product usage.
Meet Logs (beta) – more debugging context, all in PostHog
PostHog has introduced a logs feature in beta, allowing developers to view backend logs directly alongside session replays and error tracking without switching platforms.
Workflows graduate to beta! Product data, meet automation
Workflows has moved to beta, allowing users to automate product-led actions like emails and Slack alerts based on live event data.
Your product data just got a job: Workflows is now out
Workflows allows users to automate actions like sending emails or Slack alerts directly from PostHog event data. It eliminates the need for external syncing tools or custom API integrations.
Workflows are now in Alpha and I already broke mine
PostHog introduced Workflows in alpha, an automation builder that triggers actions based on live user behavior and event data.
Stop AI slop: Run evals with LLM-as-a-Judge
PostHog now provides automated evaluation tools that use LLMs to score generative AI outputs based on criteria like relevance and helpfulness. It includes pre-built templates for common use cases.
Your logs' final destination is in GA. You always end up here anyway
PostHog Logs is now generally available, providing a centralized destination for logs alongside errors, session replays, and product data.
How we made feature flags even faster and more reliable
PostHog has rewritten its feature flag evaluation service in Rust to improve speed and reliability by moving evaluation logic from database queries to in-memory processing.
Also shipped on Dec 3, 2025
PostHog in December 2025Introducing Seven New Email Templates for Supabase Auth
Supabase Auth introduced seven new email templates to notify users of security-sensitive account actions.
The new Supabase power for Kiro
The Kiro IDE now provides deep integration with Supabase, assisting developers with database migrations, edge functions, and security policies.
EU Endpoint Now Generally Available
The Deepgram EU endpoint ( api.eu.deepgram.com ) is now generally available for customers requiring data processing within the European Union.
eAPM General Availability and eBPF Network Metrics Public Preview
Background Our customers are faced with these challenges: The first challenge is the traditional observability gap. Traditional APM tools lack kernel-level network and system metrics. This means that during an incident, it is difficult to pinpoint the root cause between application code issues and network-related issues.
NeMo NVIDIA Neural Modules 2.6.0
Highlights Speech Add Timestamps to streaming ASR PR Add Streaming decoding policies (Wait-K and AlignAtt) for Canary model PR Add NeMo Voice Agent PR Hybrid RNNT-CTC Prompted Parakeet Model support PR [New] MT-Parakeet Streaming Models release Removed the Automodel module. Automodel is available in the repo https://github.com/NVIDIA-NeMo/Automodel .
Supported APIs
The EU endpoint supports the following Deepgram APIs: Speech-to-Text : /v1/listen and /v2/listen (excluding Whisper models) Text-to-Speech : /v1/speak Voice Agent : /v1/agent/converse Text Intelligence : /v1/read