iiinova — where your digital self keeps evolving

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 07/10/2026

The simplest improvements are often the best! You can now see the details of your workloads right from your cluster "workloads" tab. It's GA!

A dashboard displays workload metrics, showing CPU and memory usage percentages along with environment details and namespace information.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 29/08/2026

My Ukrainian teacher asked me "Що потрібно для успішного бізнесу?" - What is needed for a successful business - I replied from the most important to the least: luck, a strong network, money (capital), hard working, talent. Do you agree?

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 20/08/2026

At @Qovery, we’re working to treat agents as first-class citizens—yes! Alongside your traditional apps, you can have agents working and interacting with them. If you’re a Qovery user, contact us to get early access!

A digital console displays a Slack bot service with operational details and enabled agent workflows for AI interactions.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 14/08/2026

Now in Beta: Deploy agents natively with Qovery. Agents are first-class citizens alongside your apps and databases. The use cases are endless - and we’ll be sharing more in the coming days 🔥

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 10/08/2026

In 2027, Platform Engineers should be AI Platform Engineers: delegate the grunt work to AI agents, then supervise them. That’s the way to go - and that’s what we deliver at Qovery for the best companies out there.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 06/08/2026

Deploying an app on your infrastructure should be as simple as asking your favorite agent: “Can you deploy this project with Qovery?” That’s it! P.S.: Yes, I’m going to try n8n :)

A terminal interface displays a deployment plan for an application using Qovory services, highlighting project details and resource allocations.

010

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 01/08/2026

Some news about @Qovery_ : We’re building the infrastructure control plane for coding agents: one where every operation is policy-checked before execution, logged, attributable, and reversible. The missing piece for safely letting agents autonomously operate infrastructure.

020

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 31/07/2026

AI agents still haven’t made it all the way to infrastructure. Platform engineers use them to check what’s going on - but not to take action. What’s missing is a trust layer that lets agents operate safely, with real guarantees and peace of mind.

120

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

Agents will never be reliable enough for production, for the same reason application code never was. Make the operations deterministic instead. This is what we build at Qovery.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

The agent decides what: probabilistic, and getting better every quarter. A governed API executes how: deterministic, policy-checked and scoped before execution, logged and reversible after. The agent stays probabilistic. The blast radius doesn't.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

We solved this problem 50 years ago. Application code was never reliable enough to manage shared data, and we didn't fix it by making programmers smarter. We built database transactions. The code stayed buggy while the operations became safe. Same answer for agents.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

An agent that's 95% reliable per action completes a 10-step remediation 60% of the time. At 99% per action, a 50-step workflow succeeds 60% of the time. Probabilistic systems compound their errors, and no model upgrade fixes compounding.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

The rip-and-replace trap. "Abandon your stack, start fresh on our AI-native platform." Enterprises correctly said no. Vendors respond by making the agent smarter. Better model, better prompts. The problem is math.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

The YOLO trap. The agent gets admin credentials and a shell. It works until one hallucinated command runs at machine speed against the wrong cluster. Security writes the incident report, and the rollout is over.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

The advisor trap. The product reads everything and touches nothing. A human still executes every action, so the bottleneck never moves. You bought a faster way to read logs.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 28/07/2026

Every AI SRE product demos beautifully. Almost none survive production. I've watched dozens launch and die over the past two years. It happens three ways:

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 22/07/2026

Huge thanks to the EMEA Anthropic team for hosting such a high-quality, intimate dinner with top founders last night. Brilliant conversations, sharp minds, and genuinely exciting AI topics. Appreciate it.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

It's the same shift platform engineering already went through once (make people autonomous) - just extended to agents. If your platform team is still measuring itself by uptime on the tools instead of how safely the org can move at agent speed, that's worth a hard look.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

The platform engineers who'll matter most over the next few years are the ones asking "how do I let agents ship safely" - not "how do I manage my cluster." That's a different job.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

They generate PRs by the dozen, they need environments on demand, and they can't be handed a laptop's worth of production credentials. Infrastructure is where this shows up first, because it's where speed meets real risk.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

The problem isn't that those tools are wrong. It's that the whole stack was built for humans to operate carefully, one change at a time. Agents don't work that way.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

But a lot of DevOps and platform engineers are still fighting the same Kubernetes, CI/CD, Terraform.. battles as if nothing changed upstream.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

That's the gap I keep seeing. Software engineers stopped writing every line of code themselves a while ago - they're operating from AI agents now, reviewing, directing, shipping.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

Meanwhile, on the same call, his CTO told us they're already running agents like Claude Code and Codex to write and ship code. Not "evaluating AI." Already doing it, daily.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 16/07/2026

My team had one more call this week with a platform engineer at a fast-growing company. He spent 45 minutes asking them about Terraform state locks, cluster management, and whether he'd get "locked in" to a vendor. Fair questions. Ten years ago 😬.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

We're living through something incredible - but it puts real pressure on C-levels. If you're feeling it, let's talk.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

4. Every function inside an organization is being redefined by AI

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

3. Platform Engineers now serve non-tech people who want to build - not just other engineers

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

2. Everyone is a potential builder - which means governance is mandatory

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

What that means inside organizations: 1. Software Engineers need to sit closer to business. Deeply understand what they're building and why. That's no longer just Product's job

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

He has a freelancer who reviews his work - keeps him from breaking things. But the core? He built it himself. This is the reality now: anyone can be a builder. The people who used to be blocked by technical limitations? They're shipping.

200

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

A proptech platform. He knows the space deeply. Built the entire thing with Claude Code. Zero technical background. It works. He has live users. Paying customers coming soon.

110

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 15/07/2026

I'm at a co-working space today. I just met an entrepreneur who spent years selling for other companies - now he's using Claude Code and launching his own business.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 14/07/2026

Happy Bastille Day to all French and European partners fighting for freedom. 🇫🇷🇺🇦🇪🇺

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 12/07/2026

Side note: as a proud French citizen, it was especially great to see so many talented founders in Paris for the event - and choosing Paris for the long run. That momentum is real.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 12/07/2026

I’m excited about what we’re each building - and what’s coming next.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 12/07/2026

Great to finally meet @ivanburazin from @daytonaio in person at RAISE Summit last week. We shared the stage, and I’m genuinely energized by how our two companies approach agentic infrastructure in deeply complementary ways.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

That's the problem we came to solve - how to go fast and see what's happening. Photo caption: I met Jonathan from Tint meeting at Le Train Bleu in Paris, laptops open, we discussed the sandboxed AI capabilities he built on the Qovery API. Full story: www.qovery.com/blog/how-th...

qovery.com

How the market pulled us into becoming an agentic platform - Qovery Blog

We didn't decide to become an agentic platform - our customers pulled us there. Here's why infrastructure breaks first when agents show up, the problems nobody planned for, and how to move fast with control.

020

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

Once we saw it from the infrastructure side, the pattern was obvious. Agents wired directly into your cloud means shared credentials on every laptop, no audit trail, and no control over who touches what. It works for one person. It breaks the moment the agents aren't just yours.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

We met in Paris, sat down together for half a day, and it changed how I saw the whole market.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

Teams at Tint, Alan, Prezi, and Talkspace hit that ceiling before anyone else. One founding engineer at Tint had already built governed, sandboxed AI environments on top of our API - he didn't ask us to build anything new, he built it himself on what was already there.

110

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

The moment our customers put AI agents to work, infrastructure was the first thing that broke.

110

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 08/07/2026

TL;DR: You knew Qovery for deploying apps. We're now the control plane where agents run infrastructure - safely. We didn't plan this. The market pulled us there. Here's the full story.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 07/07/2026

Happy to discuss if you’re C / Exec level and you feel you’re late. It's the right time to react now!

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 07/07/2026

And I don’t think you have 12 months to catch up - Michael is being “generous” with that timeline.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 07/07/2026

I usually delete cold emails. This one had a sentence I can't stop thinking about: AI is no longer your competitive edge - it's table stakes. Which means if your company does not ship AI (with AI) into production yet, you're not behind. You're already losing.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 06/07/2026

The window to build this foundation before it becomes a crisis is closing. Most teams will figure that out the hard way.

000

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 06/07/2026

The companies getting this right aren't buying more AI tools. They're building a control plane. One that lets business teams move fast while giving security, compliance, and finance the visibility they need. On your cloud. Under your rules. Auditable from day one.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 06/07/2026

Companies are investing heavily in AI at the model layer. Almost nobody has invested in the layer that governs how agents actually run - where they execute, what they can touch, who authorized it, what it cost, and whether it can be rolled back.

100

Romaric Philogène 🇫🇷 🇺🇦 @rophilogene.bsky.social · 06/07/2026

🟠 Without governance, compliance eventually says "no AI in production, period." Teams then bypass controls entirely. The risk gets worse, not better.

100