Nos numéros
Toutes les éditions
This week, Anthropic went on a full offensive — shipping agents for coders and non-coders alike while the open-source world fired back. If you're a CTO not rethinking your developer tooling strategy right now, you're already behind.
Analyse
Anthropic's Agent Blitz: Claude Code, Cowork, and the Race to Own the Dev Stack
Anthropic didn't just ship a feature this week — it launched a platform play. Claude Code's creator Boris Cherny publicly revealed his agentic workflow, Cowork extended Claude's file-level autonomy to non-engineers, and the developer world collectively lost its composure. This matters because Anthropic is no longer selling a model — it's selling an operating layer that sits between your team and your codebase. For CTOs, the strategic question isn't whether to adopt agentic coding tools, it's whether you let one vendor own that layer or invest in open alternatives like Goose or NousCoder-14B that keep you portable. My advice: run a 2-week internal bake-off between Claude Code, Goose, and at least one open-source model on a real sprint. Measure not just speed, but context retention, security surface, and lock-in risk. The winners of this wave won't be the fastest adopters — they'll be the ones who kept optionality.
Perspective CTO
“I've seen this movie before — one vendor ships fast, everyone onboards, and 18 months later you're negotiating from a position of zero leverage. Build your agentic coding stack like you build infrastructure: with an exit plan on day one.”
Top actualités
Railway raises $100M to build an AI-native cloud alternative to AWS
Railway, which grew to 2 million developers with zero marketing spend, just raised a $100M Series B to challenge legacy cloud providers with infrastructure purpose-built for AI workloads. For CTOs running AI-heavy pipelines on AWS or GCP, this is worth watching — not because Railway replaces your cloud tomorrow, but because the AI-native infrastructure layer is now attracting serious capital and talent. Evaluate whether your current cloud abstractions are costing you latency and money on inference-heavy workflows.
Lire →NousCoder-14B matches proprietary coding models — trained in 4 days on 48 B200s
Nous Research dropped NousCoder-14B, an open-source coding model that benchmarks competitively with much larger proprietary systems, trained in just four days. This is a concrete data point for CTOs evaluating self-hosted coding agents: the performance gap between open and closed models is collapsing faster than pricing is dropping. If you have GPU capacity or a cloud GPU budget, this changes the build-vs-buy math on agentic coding tools.
Lire →Salesforce rebuilds Slackbot as a full AI agent for enterprise workflows
Salesforce gutted the old Slackbot and rebuilt it as a genuine AI agent — one that can search enterprise data, draft documents, and take action, not just ping you about meetings. This is the Microsoft Copilot vs. Google Gemini vs. Salesforce Agentforce three-way fight playing out in real time inside your collaboration tools. If your org is on Slack, test the new agent immediately — it may reduce the need for standalone AI workflow tools you're currently evaluating or paying for.
Lire →Outil de la semaine
Goose (by Block)Open-source, free alternative to Claude Code for agentic coding in your terminal.
Goose is Block's open-source AI coding agent that runs locally in your terminal — same autonomous write-debug-deploy loop as Claude Code, but at zero cost and with full model flexibility. It supports multiple LLM backends, so you're not locked into any single provider. For CTOs who want to experiment with agentic coding without committing $200/month per developer seat or handing code context to a single vendor, this is your starting point.
Explorer →Rejoignez les lecteurs