If you had told me 12 months ago that @thomasdesfrancs who has no software engineering background would be able to build this as code, I would not have believed it

Open source isn't dead but it's true that our security model is. Right now most SaaS products put all tenants on shared infrastructure. A free signup from a throwaway email often sits on the same database as a paying customer with sensitive data. If all code becomes more vulnerable then maybe the question is: when a breach does happen, how do we mitigate the scope/impact? Infrastructure is probably the most under-utilized lever outside of Enterprises. At least SaaS will need to adopt standards inspired by how AWS handles service limits on new accounts, or how Stripe graduates merchants through risk tiers. We're currently exploring progressive tenant isolation on our cloud offering. The idea is simple: tenants earn trust over time and get moved into increasingly secure pools. A job evaluates security scores based on signals — KYC completed, payment history, usage patterns, anomalous behavior. New signups start sandboxed. Verified tenants graduate. Enterprise gets private endpoints. But not sure if shared infrastructure even makes sense for most B2B applications (especially if you don't have a free tier). Starting at $20/month you can probably get a dedicated instance for each customer and give full network isolation from day one. Then open source makes even more sense becaus we're back to the traditional self-hosting model. But the reason SaaS won over on-premise is that we were not able to build the right tooling to make upgrades and development easy. Our ecosystem is designed around shared databases with tenant_id columns. Managing a fleet of isolated instances, deploying migrations to thousands of tenants, monitoring across all of them, we don't really have the good framework for that. But it's a problem worth solving!

Amazing time to start a robotics startup or just tinker for fun. And it’s open source!

Google trends for Exit Tax in France after Macron dissolved the National Assembly and the next prime minister is likely to come from far left or far right

I think the most common mistake in today's AI excitement is thinking all software must become “AI-first”. Construction didn’t become “electricity-first” when electricity arrived. Paris didn’t need to be rebuilt. Haussmannian buildings and high-quality constructions evolved and still stand. Starting with weak software foundations and a direct focus on LLM is perhaps the worst way to build a defensible software in the era of AI. The best way to future-proof your SaaS is to focus on deep engineering problems. The core of what makes software great isn't changing; it's getting more important.

Proof that dogs are the best babysitters

Guys Elon finally did it! Bot problem is solved. We’re just between real humans now