# Changelog #17: Wake up your AI agent before getting paged
Every second matters when an incident happens. AI SRE investigates your incidents, pulls all the evidence, and gives you a root cause analysis. Focus on decision making and the fix itself, and leave the data analysis to AI SRE.
Read on to make the most of AI SRE incident analysis.
## Wake up to a root cause, not an unknown incident
**Paged at 3am with nothing but an alert name?** AI SRE starts investigating the moment the incident happens. It checks what's failing and what changed around the start time, then posts a short summary to the incident timeline and Slack. By the time you've opened your laptop, the root cause analysis is ready.
Enable AI SRE investigation for all incidents.
## Tell AI SRE how to investigate
**Every team debugs differently.** Write your own prompt to control what AI SRE checks, how deep it goes, and how it reports back.
**Want different prompts for different incidents?** Set up multiple AI SRE integrations, give each its own prompt, and trigger the integrations in your escalation policies.
Customize your AI SRE prompts and use them in escalation policies.
## Give AI SRE the context it needs
**AI SRE builds its root cause analysis from your logs, traces, metrics, errors, and recent releases.** Connect GitHub, GitLab, Sentry, Grafana, Datadog, Linear, and Notion, and let AI SRE dig through them during the investigation.
Connect your existing tools to AI SRE.
## Not using AI SRE yet?
**Investigate your first incident.** Trigger an investigation with a single click for any of your incidents.
**Want to analyze your data even before an incident is opened?** Start a new AI SRE chat and ask about your telemetry, errors, recent commits and changes.
Start using AI SRE.
---
## Fresh community content
### GPT-5.6 Sol and Ultra Mode: What You Need to Know
OpenAI’s GPT-5.6 ships a three-model family and Ultra Mode, which lets the flagship Sol model spawn and coordinate specialized sub-agents internally. What used to take external orchestration frameworks now happens inside one API call.
### GitHub Agentic Workflows: Automating DevOps with AI-Powered GitHub Actions
Write a workflow in plain Markdown and let an AI agent run it as a GitHub Action: performance audits, issue triage, code cleanups. A practical look at continuous AI in the pipeline you already have.
### [How They Fixed AI’s Biggest Local Hardware Problem (with FreeToken)](https://www.youtube.com/watch?v=8n4Uo47CadY";_blank")
FreeToken is an open-source inference engine that runs Mixture-of-Experts models dramatically faster than Ollama once they outgrow your GPU’s VRAM. A look under the hood at expert caching and bandwidth-adaptive loading.
### [I replaced my entire tech stack with Postgres](https://www.youtube.com/watch?v=MZfhhl6ej-k";_blank")
Caching, vector search, full-text search, geospatial, cron, and document storage, all inside the one database you already run. Not because you should, but because you can.