DEV Community: Flora Brandão

Why we moved from LVM to Ceph for container storage 📦

Flora Brandão — Fri, 22 May 2026 10:00:00 +0000

Managing storage that is glued to your compute nodes makes scaling a nightmare. We found that LVM worked for a while, but it eventually stopped our VMs from being truly disposable and flexible.

To fix this, we migrated our container storage to Ceph. Here is what we learned and why it changed our workflow:

Separating compute from storage allowed for much better flexibility.
Data cloning turned out to be the essential feature that drives our setup.
We gained the ability to perform instant volume migrations.
Our VMs are finally disposable again.

The storage layer you choose shapes every other part of your system. It is worth looking at the trade-offs and what worked for us before you commit to an architecture.

Check out the full breakdown here:
https://developer.upsun.com/posts/how-it-works/why-upsun-moved-from-lvm-to-ceph

Centralising Upsun Technical References with Meta-Upsun

Flora Brandão — Thu, 21 May 2026 14:00:00 +0000

Managing technical reference data is a grind when information for runtimes and services is scattered everywhere. It leads to inconsistent automation and inaccurate documentation.

We built Meta-Upsun to fix this. It is a read-only and versioned API that centralizes runtime, service, region, and PHP data into one public source.

Why it matters for your workflow:

Provides a single source of truth for consistency
Automates documentation to keep it accurate
Builds in reliability and validation from the start
Aligns registry status with upstream support

This shift improves the developer ecosystem and leverages agility for better automation.

Check out the full article here:

Centralising Upsun Technical References with Meta-Upsun - Upsun Developer

How Upsun centralised runtime, service, region, and PHP reference data into a single public API to improve consistency, automation, and documentation accuracy.

developer.upsun.com

Why AI agents need cloud sandboxes 🤖

Flora Brandão — Thu, 21 May 2026 10:00:00 +0000

Running AI agents on your local machine is a security headache and a validation trap. While local isolation works for simple coding tasks, it falls apart when you need to see how an agent handles real systems and production data.

Here is why the local approach is failing your workflow:

Agents execute real commands that require real infrastructure.
Local containers lack the environment depth needed for proper validation.
Testing without production data means you are not seeing the full picture.

The solution is moving to cloud sandboxes. Using preview environments gives you a full stack sandbox with access to real data. It allows you to use the Git workflow as a safety net. This approach ensures your agent workflows are safe and your systems remain secure.

Read the full article for a deeper look at agent security and cloud sandboxing:

Why AI agents need cloud sandboxes - Upsun Developer

AI agents run real commands on real systems. Learn why cloud sandboxes with production data beat local isolation for safe AI agent workflows.

developer.upsun.com

[Guide] Making coding agents (Claude Code, Codex, etc.) reliable 🛠️

Flora Brandão — Tue, 19 May 2026 14:00:00 +0000

Model capability is rarely the problem when AI coding agents fail. The real bottleneck is your verification infrastructure. When humans cope but agents crash, it is usually because the intent is not locked before implementation. This leads to the phenomenon of AI slop.

To build reliable agents, you need to shift your approach:

Focus on the asymmetry of verification.
Use spec-driven development to lock intent.
Build a robust infrastructure using the eight pillars of verification.
Create a flywheel that makes reliable Software 2.0 possible.

This shift helps prevent agents from crashing and ensures your code remains highly verifiable.

Check out the full technical breakdown for the 8 pillars and the self-assessment guide:

Making coding agents (Claude Code, Codex, etc.) reliable - Upsun Developer

The bottleneck for AI coding agents isn't model capability. It's your verification infrastructure. Here's how to fix that.

developer.upsun.com

Automating runtime and service tracking with Meta Version Updater 🤖

Flora Brandão — Tue, 19 May 2026 10:00:00 +0000

Keeping your documentation in sync with the reality of your services and runtimes is a constant manual grind. When your reference data is outdated, you end up fighting drift instead of building features.

We built Meta Version Updater to act as a shared reference layer for runtimes and services. Here is how we handled the automation:

We implemented a nightly ingestion pipeline to pull data automatically.
Human review remains in the loop to maintain high data quality.
The system detects drift before it causes issues for developers.
This layer now powers our documentation and internal tooling.

Automating these updates means less time spent on manual tracking and more time on actual engineering. We are even looking at adding Slack summaries via n8n to keep everyone updated.

Check out the full article:

Automating Runtime and Service Tracking with Meta Version Updater - Upsun Developer

Meta is Upsun’s shared reference layer for runtimes, services, regions, and extensions. Learn why we built it, how it works, and how it powers the next generation of Upsun documentation and tooling.

developer.upsun.com

[Guide] Sandboxing AI agents on Upsun 🛡️

Flora Brandão — Mon, 18 May 2026 14:00:00 +0000

Prompt injection is a massive headache when you are giving AI agents access to your codebase and credentials. The real vulnerability usually is not the AI itself but the harness that connects it to your environment.

The problem:

Prompt injection can trick agents into revealing sensitive secrets.
Many developers focus on the model and ignore the security of the execution environment.
Without isolation, an agent has too much reach into your system.

The fix:

Sandbox your AI agents inside Upsun containers.
Use Linux primitives to build a secure and isolated environment.
Leverage tools like Claude Code and Codex within these protected boundaries.

By isolating the agent harness, you can use powerful automation without the constant fear of a malicious prompt compromising your infrastructure. It is a practical way to keep your development workflow fast and secure.

Check out the full technical write-up for the deep dive on building these sandboxes:

Sandboxing AI agents on Upsun - Upsun Developer

Sandbox AI agents in Upsun containers to protect secrets from prompt injection. Covers Linux primitives, Claude Code, and Codex.

developer.upsun.com

How product managers use preview environments and AI to prototype 🏗️

Flora Brandão — Mon, 18 May 2026 10:00:00 +0000

The distance between a product idea and actual validation often feels like a long grind of writing briefs. Most of the time, those documents do not prove if an idea actually works in the real world.

At Upsun, product managers are changing the workflow to get faster feedback:

Use AI coding assistants to quickly build out feature ideas.
Leverage preview environments to see how code behaves with real data.
Share interactive demos with stakeholders instead of static documents.
Validate concepts early and find out if an idea is bad before over investing.

Moving to a git push workflow means you can stop guessing and start testing. It is a practical way to ensure your prototypes are grounded in reality and ready for feedback.

Read more to understand how this workflow can help your team:

How product managers use preview environments and AI to prototype - Upsun Developer

Learn how product managers at Upsun use AI coding assistants and preview environments to prototype feature ideas, validate concepts with real data, and share interactive demos with stakeholders.

developer.upsun.com

Git worktrees for parallel AI coding agents 🛠️

Flora Brandão — Fri, 15 May 2026 14:00:00 +0000

Managing parallel AI coding agents is a nightmare when you are stuck with single branch limitations. We have all struggled with isolation problems and the friction of switching contexts while trying to keep multiple agents productive.

The problem:

AI agents need isolated environments to work effectively
Standard workflows fall apart when orchestration tools lack the right structure
Context switching creates overhead and slows down parallel development

The fix:

Use Git worktrees to enable true parallel agent workflows
Implement preview environments to solve the isolation problem
Move toward orchestration tools that handle worktree logic natively

Using worktrees allows your agents to operate in separate, concurrent spaces without stepping on each other. It is a practical way to scale your AI assisted development without the typical environment headaches.

Check out the full breakdown:

Git worktrees for parallel AI coding agents - Upsun Developer

Learn how git worktrees enable parallel AI agent workflows, their limitations, and what an ideal orchestration tool needs.

developer.upsun.com

Composer 2.9 lands on Upsun PHP Images: here’s what you need to know 📦

Flora Brandão — Thu, 14 May 2026 14:00:00 +0000

You didn't change a single dependency and yet your build is suddenly failing. It is a frustrating situation that often points to a critical update in your underlying environment.

A remote code execution vulnerability in Next.js 15.x and 16.x App Router has been identified and it requires immediate patching to secure your applications. Additionally, Composer 2.9 has landed on Upsun PHP images.

Here is the breakdown:

Security patches are now required for specific Next.js versions to prevent remote execution.
Composer 2.9 is being shipped as is to ensure environments stay current.
Fresh dependencies are essential for maintaining a secure and functional workflow.
You have options to resolve build failures and get back to shipping code.

If you are seeing errors or want to understand how these updates impact your specific setup, check out our technical breakdown:

Composer 2.9 lands on Upsun PHP Images: Here’s what you need to know - Upsun Developer

Remote code execution vulnerability in Next.js 15.x and 16.x App Router requires immediate patching. Learn how to secure your applications.

developer.upsun.com

Speed vs. Security? Why not both: the Fastly CDN and WAF on Upsun 🛡️

Flora Brandão — Thu, 14 May 2026 10:00:00 +0000

Choosing between a fast CDN and a robust WAF often feels like a compromise that leaves your infrastructure vulnerable or your users waiting. You should not have to sacrifice global performance just to keep the threats at bay.

The solution is to pair the Fastly CDN with its Next Gen WAF on Upsun to get the best of both worlds. Here is why this setup works for your workflow:

It optimizes global performance for a better user experience.
It protects your infrastructure from active threats.
It integrates security directly into your delivery layer.

Check out the full guide to see how to implement this on Upsun:

Speed vs. Security? Why not both: The Fastly CDN and WAF on Upsun - Upsun Developer

Learn how to pair Fastly’s CDN for speed with its Next-Gen WAF for security on Upsun. Optimize global performance while protecting your infrastructure from threats.

developer.upsun.com

[Event] Scaling secure fintech: Meet us at Money20/20 Europe 🇳🇱 💶

Flora Brandão — Wed, 13 May 2026 10:00:00 +0000

The financial sector is facing a massive shift. Between new regulations like DORA and the pressure to modernize legacy stacks, the "standard" way of managing infrastructure isn't cutting it.

Upsun will be at Money20/20 Europe in Amsterdam (June 2–4) to show how we solve these challenges:

30-Second full-stack clones: Spin up isolated, bit-for-bit environments (data included) in less than 30 seconds for risk-free testing and high-velocity innovation.
Built-in compliance: We integrate DORA, SOC 2 Type 2, ISO/IEC 27001, and BDSG directly into your pipeline. Security is built-in, not bolted-on.
Unified multi-cloud toolchain: Deploy across AWS, Azure, or GCP with a standardized Git-driven process that supports PHP, Go, .NET, Python, and more.
24/7 Expert human support: Access proactive architecture guidance and engineering support to ensure the 100% uptime your financial apps demand.

If you’re building in the #fintech space, come find our booth in Amsterdam. 👋

Let’s talk about harmonizing your stack and achieving true financial resilience. 🚀

Money20/20 Europe in Amsterdam | 2 - 4 June 2026

Join us in Amsterdam, where the industry's leading Financial Services innovators come together to connect and create the future of money. Buy your pass today!

europe.money2020.com

Building an AI code review agent for our self-hosted GitLab 🤖

Flora Brandão — Tue, 12 May 2026 14:00:00 +0000

Managing code reviews on a self-hosted GitLab instance can be a constant grind as the volume of merge requests starts to pile up.

We used Claude to vibe-code a custom review agent to handle this workload and the results were surprising.

The agent consists of 40,000 lines of Python code.
It has already reviewed 1,000 merge requests for our team.
Building a custom tool provided the specific control we needed for our self-hosted environment.
This approach automates the initial feedback loop and keeps the development workflow moving.

You can find the full technical write-up covering the architecture and the lessons learned here:

Building an AI code review agent for our self-hosted GitLab - Upsun Developer

I vibe-coded a GitLab code review agent last month - 40K lines of Python written by Claude - and it has reviewed 1000 merge requests.

developer.upsun.com