Guillaume Gelin

Yesterday, Terramate joined the Cloud Native Paris Meetup at free.fr’s offices, together with around 80 attendees. To kick off the year, the event focused on concrete, real-world experiences beyond theory, featuring two in-depth talks: 1️⃣ GitOps Reimagined: How to Fix State Drift Before It Even Happens Presented by Damien DASSIEU from Scaleway, this talk introduced SynGit, an open-source Kubernetes operator exploring how to reconcile manual kubectl or API changes with GitOps by capturing them directly into Git. 2️⃣ AI & Kubernetes: How the Institut Pasteur Enabled Large-Scale AI Adoption in Production Teams from Institut Pasteur and Enix shared how they built and operated on-prem Kubernetes infrastructure to run AI workloads locally, accelerating research workflows and feedback loops. Great discussions, practical insights, and a strong community atmosphere. Thanks to the organizers, especially Smaïne KAHLOUCH, and to all the speakers for a great evening.

15

Romain Huet at OpenAI DevDay demo‘ing an idea that I think will be HUGE. Like super huge. I think this will be pervasive in 6-12 months: Users of your app will have a new “Codex” tab that will allow you to modify the software and add new features in-place. Inside the app itself. No code editor, no PRs, no deployment. This demo was for a weight loss management app. But I think we will have a lot of enterprise / internal software add in-place app edits. They’ll gate it to admin users.

305

44 Comments

Anthropic just launched "Agent Teams" in Claude Code — multiple AI agents working in parallel, coordinated by a team leader, each with its own context. Great innovation. For software development. We demoed this architecture for network operations months ago on the Cisco DevNet Podcast. Five specialized agents, each powered by its own fine-tuned reasoning model, connected to live data center infrastructure. One virtual Team Leader orchestrating the virtual team. Something breaks in the network — the right agent picks it up, diagnoses it, and proposes a fix. Seconds, not hours. Here's what's different from what Anthropic shipped: → Our models are private. Fine-tuned. On-prem. No network telemetry, configs, or topology data leaving the customer's environment. → Our agents are domain-specific — trained on networking, not general-purpose LLMs with clever prompts. → Our system is autonomous. It doesn't wait for a developer to type a command. It watches, detects, and acts. Multi-agent orchestration is now mainstream. The real question is: can your agents actually understand your infrastructure? 🎙️ Full demo: https://lnkd.in/eq_YFyxR 🎧 Presentation : https://lnkd.in/e53eybfk #NetworkAutomation #AI #MultiAgent #AIOps #CiscoDevNet

44

1 Comment

How was Codex built, why is it open source, how is it used inside of OpenAI, and how has it changed software development within the company? The single best person to answer: Thibault Sottiaux (Tibo), who was on the team that started Codex, and heads up the org building Codex. • YouTube: https://lnkd.in/esUEhaSP • Spotify: https://lnkd.in/eYHfnUiQ  • Apple: https://lnkd.in/esAFZqsb Brought to you by: • turbopuffer – a vector and full-text search engine built on object storage. It’s fast, cheap, and extremely scalable. https://lnkd.in/erQW8sEY • Antithesis – verify your system’s correctness without human review or traditional integration tests – and avoid bugs or outages. https://lnkd.in/eyWErUbm • Entire – Git hosting, rebuilt for the agentic era. Entire hosts your code in-region, and is up to 89x faster than any other competitor. https://lnkd.in/eN2mCs5B Three interesting parts from Tibo: 1. Google had a “ChatGPT-like” project a year before OpenAI. At the start of the decade, DeepMind was largely focused on games and reinforcement learning, but some members, including Tibo, reckoned that scaling language models might lead to artificial general intelligence (AGI). Their “chat with an LLM” project spread like wildfire internally, but for some reason the product was never launched. Who knows what would have happened if Google had beaten OpenAI to the public launch of AI chatbots in 2022. 2. Open source is a big reason why Codex supports working with other AI models. Claude Code can only be used with Anthropic’s models, whereas Codex is usable with any model, not just OpenAI’s. Being open source means that even if Codex were locked down to a model, anyone could still fork the harness and change a few lines of code to support a different model. Tibo believes in winning by letting users use the best models; the Codex team themselves also try other models in the same harness. Personally, I appreciate this approach of encouraging competition from a frontier lab! 3. Maintaining and re-architecting code is becoming very cheap with agents, especially time-wise. Maintenance tasks like dependency upgrades can be handled by a model blasting through the codebase within a couple of hours. Meanwhile, re-architecting for new tradeoffs that used to take years can now take days, at most. Tibo adds a caveat: quality code, good abstractions, and good test suites greatly affect how easy – or not – a codebase change is to make.

74

14 Comments

The next platform decision won't be about scaling one architecture — it'll be about knowing when to split it in two. On September 10, Consensys announced it is separating into two independent companies. MetaMask stays consumer-facing — self-custody, everyday spending and saving — still under Joe Lubin. A new institutional entity takes over the underlying Ethereum infrastructure: the Layer 2 networks and rails banks and asset managers build on. Same origin, same stack, same brand family — until the two customer bases pulled hard enough in opposite directions that one roadmap couldn't serve both. I see this pattern more often than the crypto framing suggests. 🔹 Divergent economics, not divergent tech. Retail wants simplicity, speed, and zero friction. Institutions want compliance, custody guarantees, and audit trails. Same rails, incompatible product requirements. 🔹 One roadmap can't serve two masters. Every consumer feature ships faster when institutional hardening waits — and every enterprise compliance sprint slows the pace retail users expect. Eventually the trade-off shows up in both products. 🔹 Splitting is a strategy, not a retreat. Consensys isn't shrinking — it's letting each business raise capital, hire, and set its own roadmap without dragging the other along. My takeaway: platform strategy isn't only about what you build. It's about recognizing the moment one architecture is quietly running two different businesses — and giving each the operating model, funding logic, and pace it actually needs. Miss that moment and you either over-engineer the consumer product for requirements nobody there asked for, or starve the trust infrastructure your biggest clients need to scale. It's the same question I bring into TRECOS advisory conversations before we talk features or funding: who is this platform really serving, and can one team credibly serve all of them at once? Have you seen a platform — yours or one you've watched from outside — reach the point where splitting made more sense than scaling? #PlatformStrategy #RDLeadership #OperatingModels #TechLeadership #BusinessModelInnovation #DigitalWallets #Fintech

3

When AWS goes down, the world pauses. But maybe that’s exactly the reminder we need. Cloud computing gave us scale and speed — but also centralized fragility. When a single cloud outage can take down half the Internet, we have to ask: What happens when the “center” fails? That’s where decentralized computing — edge, mesh, and peer-to-peer — comes in. Not as a rebellion against the cloud, but as an evolution of it. A resilient, distributed fabric where every node contributes, and where downtime in one region doesn’t mean silence everywhere. It’s time we stop thinking of the cloud as the “brain”, and start seeing the network itself as a living, thinking organism — from cloud → edge → device. Edge computing isn’t the future. It’s the missing balance. #EdgeComputing #Decentralization #Cloud #Resilience #P2P #OrpheLink #ZeroTrust #AWSOutage https://lnkd.in/gybsf-m5

7