TL;DR

1️⃣ Use both Prompts + UI to adjust to your product surfaces
2️⃣ Create a lightweight design system to balance usability with portability
3️⃣ Use ‘Prompt Fidelity’ to set a bar for UX and UI
4️⃣ Build trust and increase success rates with ‘Spare Citations’

Building Delightful UX in the Age of AI AI breaks many of the best practices and benchmarks for building a delightful user experience. Here’s what we’ve learned while building Flightcrew

AI challenges the traditional tools and benchmarks for building a delightful UX. User journeys are open-ended, user experience is probabilistic, and intent is no longer declared but inferred.

We've learned a few things about AI Usability while building Flightcrew

www.flightcrew.io/blog/buildin...

AI SRE vs AI Platform Engineer We get asked if we're building an AI SRE. That sounds useful, but here's why we're building something new and different

I was in SF last week and was asked a few times if @flightcrew.io is an 'AI SRE' which seems to be a tool that

1. Ingests logs + jira context
2. Identifies root cause
3. Recommends a fix

This seems like a useful tool but we’re not building that.

Here’s why

www.flightcrew.io/blog/ai-plat...

Behind the Scenes: Building a GitHub Copilot Extension We built a Copilot Extension so that everything Flightcrew about your cloud is accessible via GitHub Copilot. Here what we've learned about building and integrating with AI products.

Building a GitHub Copilot Extension requires solving hard problems like
- Prompt routing
- Authorization
- Context management

Here's what we learned at @flightcrew.io from the process.

Announcing our Integration with GitHub Copilot Everything Flightcrew knows about scaling, maintaining and securing cloud infrastructure is accessible via GitHub Copilot

Read more here:
www.flightcrew.io/blog/github-...

With Flightcrew + Copilot you now can ...

- Catch and fix infrastructure errors before you deploy
- Generate IAC refactors or ‘magic numbers’ straight into a PR or IDE
- Give every engineer a coach for SRE, FinOps and Compliance

Why build a copilot extension?

First, we’re heavy users of Copilot and and so this started as an internal hack.

Second - we’ve found that a Copilot chat is better for tasks where there's a high degree of ambiguity about current + desired state

Flightcrew integrates with Github Copilot

Flightcrew now integrates with GitHub Copilot

We're building the AI bot that catches and fixes the thousands of things that can go wrong in cloud infrastructure.

And now … (almost) everything Flightcrew knows about cloud infrastructure is accessible via Copilot.

Scaling Kubernetes with KEDA and RabbitMQ Setting KEDA to scale on RabbitMQ extends horizontal autoscaling to message-based systems. Here’s a quick guide

Platform Engineers love #KEDA because it extends Kubernetes autoscaling to message-based architectures. This is critical for managing the cost and dependencies of today's AI workloads.

Here's a quick guide for scaling KEDA on #RabbitMQ:

www.flightcrew.io/blog/keda-ra...

An Honest Look at Vertical Pod Autoscaling We're giving our honest read on VPA, what it's missing for enterprise workloads, and how it interacts with Flightcrew.

We try to have honest conversations with engineers about when they should use an AI Agent vs an existing open source tool.

This gets tricky when Agents can both supplement, and replace existing tools and workflows.

Here's a look at the Kubernetes VPA in 2025

www.flightcrew.io/blog/an-hone...

API, ChatGPT & Sora Facing Issues OpenAI's Status Page - API, ChatGPT & Sora Facing Issues.

Read more here: status.openai.com/incidents/ct...

tldr - a new telemetry service tested fine in staging, but overwhelmed the Kubernetes API on larger production clusters. Live remediation was rough because of cascading errors / lockout

Quote from OpenAI: “...  the fix for this issue required us to remove the offending service. In order to make that fix, we needed to access the Kubernetes control plane – which we could not do due to the increased load to the Kubernetes API servers ...”

OpenAI published their postmortem from Wednesday's outage and it perfectly summarizes the difficulty of production engineering: cloud environments are volatile and any change can have unintended consequences

We're at re:Invent this week to chat all things AI + IAC.

Let's meet up at Booth, Bar or Pool 😎

👋