Fabio Oliveira

AI-DLC v2, Part 1: What Actually Changes

Thu, 16 Jul 2026 10:00:00 +0100

AI-DLC v2, Part 1: What Actually Changes Link to heading

Most teams have settled into a comfortable arrangement with AI in their software delivery: the model writes code, a human reviews it, and everyone moves on. It works, but it quietly caps how far the automation can go — every stage still waits on a person to say “yes, continue.” The AI-Driven Development Life Cycle (AI-DLC) was built to challenge that ceiling, and its second version, now in preview, sharpens the challenge considerably.

The AI Act Deadline Didn't Move — Not the Part That'll Bite You

Wed, 15 Jul 2026 09:00:00 +0100

The AI Act Deadline Didn’t Move — Not the Part That’ll Bite You Link to heading

Half my DACH customers relaxed last month. They read “EU AI Act high-risk rules delayed to 2027” and quietly stood their compliance programs down.

That’s a mistake, and it’s worth being precise about why.

The Digital Omnibus split the August 2, 2026 deadline into two piles. One moved. The other — the one most likely to touch a mainstream business — did not. And this is no longer a proposal: the Council gave the amendments final green light on June 29; the delay is binding law, not a rumor you can plan around.

The 59% That Never Ship: Why AI Agents Die Between Demo and Production

Mon, 06 Jul 2026 10:00:00 +0100

The 59% That Never Ship: Why AI Agents Die Between Demo and Production Link to heading

Tian Pan framed this better than anyone I’ve read. In The Demo-to-Production Cliff, he makes a claim that sounds wrong until you do the arithmetic: a 90%-accurate per step agent isn’t 90% of a product. It’s an unshippable one. His diagnosis is airtight, so I’m not going to re-litigate it — I’m going to take it as given and push on the part that leaves practitioners stuck: how you make failure cheap on real infrastructure. That’s the question I put to 16 CTOs who deploy agents for a living, and their answers are what most of this post is about.

The AI-Powered SDLC Landscape: Where AI-DLC Fits

Mon, 06 Jul 2026 10:00:00 +0100

The AI-Powered SDLC Landscape: Where AI-DLC Fits Link to heading

AI-DLC v2 just hit preview (awslabs/aidlc-workflows). To mark it, I’m running a short series on AI-DLC over the coming weeks. This first post steps back to map the whole AI-powered SDLC landscape and show where AI-DLC fits; the follow-ups go deeper into what v2 changes and how to put it to work.

Figure: The journey to an autonomous SDLC — Autocomplete → SDD → Human-managed SDLC → Agentic SDLC, with a fully-autonomous aspirational endpoint. The single axis — how much of the lifecycle does AI drive? — sorts every framework.

Why Autonomous Cost Governance Is FinOps's SRE Moment

Mon, 06 Jul 2026 10:00:00 +0100

Why Autonomous Cost Governance Is FinOps’s SRE Moment Link to heading

SRE went from pagers to auto-remediation in a decade. FinOps is about to make that same leap in months — and platform teams who see the pattern have a massive head start.

The Tipping Point Link to heading

On June 9, 2026, AWS launched the FinOps Agent in public preview at FinOps X in San Diego — the first AWS-native agent that doesn’t just show you cost anomalies, it investigates them autonomously.

AWS re:Invent 2025 — Accelerating Development and DevSecOps with Amazon Bedrock and Kiro (PEX303)

Wed, 03 Dec 2025 10:00:00 +0100

Accelerating Development and DevSecOps with Amazon Bedrock and Kiro Link to heading

My AWS re:Invent 2025 session (PEX303) on accelerating software development and DevSecOps practices with Amazon Bedrock and Kiro.

Improving Reliability and Performance for Teleoperations for Autonomous Driving Vehicles on AWS

Fri, 21 Jul 2023 10:00:00 +0100

External post on the AWS Industries blog: how automakers can use WebRTC and AWS services (Kinesis Video Streams, Direct Connect) to build secure, low-latency communication between autonomous vehicles and remote teleoperation centers.

Scalability and high availability rules

Sun, 27 Jan 2019 20:53:16 +0100

Objectives Link to heading

This page describes a set of rules and a methodology that, once applied to a system, can help you evaluate how well your system is regarding scalability and availability, can help you map your weaknesses and plan actions to improve them. In this context, scalability, availability and performance measured by low latency and high throughput are highly intertwined, as a high available system need to scale to be available regardless the load, a system with a high throughput scale more easily than one with a low throughput, and as a consequence is more available.

The belonging clues

Sun, 13 Jan 2019 18:39:50 +0100

Don’t forget the belonging clues Link to heading

A reminder to myself, from the book Culture Code, a book I can recommend:

Posture and expressions are important
Avoid interruptions
Spotlight your weakness (I can be wrong here)
Embrace the message (hug the messager)
Preview future connections
Overdue thank yous
Design a painstakingly interview
Eliminate bad apples (no dickhead policy)
Design spaces that maximaze collisions
Everyone has a voice
Pick up trash
Capitalize in threshold moments (day one)
Avoid to give sandwich feedback
Embrace fun

AWS Observability Handbook

Mon, 01 Jan 0001 00:00:00 +0000

AWS Observability Handbook

Monitor, trace, and alert your cloud applications with AWS' myriad observability tools

By Fabio Oliveira and Phani Kumar Lingamallu · Published by Packt

Get the book on Amazon

About the book Link to heading

As modern application architecture grows increasingly complex, identifying potential points of failure and measuring end user satisfaction, in addition to monitoring application availability, is key. This book helps you explore AWS observability tools that provide end-to-end visibility, enabling quick identification of performance bottlenecks in distributed applications.

Contact me

Mon, 01 Jan 0001 00:00:00 +0000

Want to get in touch? Drop a message below and it lands straight in my inbox. I read everything, though I can’t always promise a fast reply.