
Beyond the hype.
News explained, in-depth analysis and real-world experience in AI, cybersecurity and software engineering.
Insights from engineers building and running real-world systems.
Latest articles
Data & security6 min read
Crédit Mutuel Arkéa cyberattack: how far can your service providers' access go?
Fraudulent transfers sent from the bank's own account, through a service provider's platform. What we know about the attack on Arkéa, and the controls that limit what a compromised access can do. November 12, 2026Read Data & security7 min readBureau Vallée: customer data was also held by the service provider
Two incidents at the provider that runs Bureau Vallée's point-of-sale and management software, customer data copied, the online store spared. What we know, what we don't, and three questions to ask your own provider. November 5, 2026Read Data & security5 min readYour AI automations work. But what can they reach?
A demo proves that a tool works. It says nothing about what it can reach. In between are the access rights you granted it, which nobody reviews anymore. Updated November 5, 2026Read AI in business3 min readAutomation and AI: what costs you is what you forget
An automation runs smoothly for months. Then the person who built it changes jobs, a supplier changes its prices, and nobody knows who looks after it anymore. Four questions to take stock. Updated November 5, 2026Read AI in business4 min readAI and invoices: the real risk is when everything looks green
Reading an invoice is something AI can do. The difficulty starts with the exceptions, where it can produce an answer that is plausible and wrong. A control circuit keeps reliability and cost in hand. Updated November 5, 2026Read Data & security5 min readCyberattack: the questions to ask your teams before the crisis
Ten people on a video call, contradictory proposals, and a breach that is still open. What you do in the first hours depends mostly on decisions made months earlier, in calm conditions. Updated November 5, 2026Read
Data & security5 min read
A data breach can cost you your customers and your sleep
A data breach does not close with the vulnerability. It ties up your teams, exposes your customers and feeds further attacks, which AI makes easier to personalize. How far it reaches is decided beforehand. Updated November 5, 2026Read AI in business5 min readAre you sure you know how much your AI projects cost you?
The quote, the licenses, the provider's invoices: you know those. Your teams' time, the rework, the checks and the bill once in production, much less so. That is often where the budget of an AI project changes. Updated November 5, 2026Read AI in business3 min readDeploying AI isn't enough
AI assistants are being used, the licenses are paid for. But have the processes actually changed? Often, you save time on one task without touching the work around it. Updated November 5, 2026Read Method & engineering3 min readWhy are the AI giants sending their engineers to their clients?
OpenAI, Palantir, AWS and Microsoft are putting forward engineers who work directly with their clients' teams. What this role says about engineering in the age of AI, and what to look for in a technical partner. Updated November 5, 2026Read Method & engineering4 min readOur finance manager built her own software with AI
She does not develop software. In one day, with AI, our finance manager had a working forecasting prototype. What we then had to rework to turn it into a team tool. Updated November 5, 2026Read AI Agents2 min readAI agent: behind the word, what are we talking about?
Very different tools get called “AI agents”. Chatbot, copilot or agent: what the software actually does, and why the right level of autonomy depends on the cost of a mistake. Updated November 5, 2026Read Method & engineering3 min readWe measured our development lead times. Here's what we found.
We connected our Jira tickets to a Grafana dashboard. Over two months, the median cycle time went from 2.8 to 2.17 days. What that figure tells us, and what it does not yet allow us to conclude. Updated November 5, 2026Read AI Agents3 min readWorkflow or AI agent: who decides what happens next?
A workflow follows a sequence planned in the code. An agent can choose the next step itself. A simple criterion to tell them apart, and what autonomy costs in testing and control. Updated November 5, 2026Read AI Agents3 min readHow far can you let an AI agent act on its own?
An agent can look things up, prepare, modify, send, refund. These are not the same decisions. How to set, action by action, what an agent does on its own and what still requires approval. Updated November 5, 2026ReadLet's talk about your architecture with us
Building complex systems takes a partner who understands the nuances of long-term application health.
