Software Architect and Founder of @strumenta, a consulting studio specialized in Language Engineering. We build DSLs, parsers, editors, languages, compilers
Learn everything you need to parse, from scratch: you can start building a parser today. Nobody dreams of building a parser. You need a parser for something. We can help get you there faster. Discover our ANTLR Video Course at the following link: https://t.co/PGWUlmDNax
Mistral now gives you a way to build and deploy agents, workflows and apps, super EASILY.
The best part: you can deploy it ANYWHERE, so AI brains can run on your infrastructure
Give it a try and let us know how we can improve it 🙏
Our AI marketing agent, Aurélie, found some very good conferences for us. But all of them were in the past! This is video 4 of our AI agent series.
Thinking about bringing AI agents into your team? Book a discovery call with us. Links in the reply.
It is surprising how, for many tasks, I get measurably better results, in a much quicker time, at a fraction of the cost using DeepSeek v4.1 Flash hosted by @MistralAI (and therefore with better data protection) compared to frontier models like Sol 6.1
BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the exception list.
Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per million tokens, roughly a tenth of what they paid Anthropic and OpenAI. Engineering lead time improved 70%.
Here's the part that matters: the gains showed up when tokens stopped being rationed. Engineers quit saving AI for the hard tickets and pointed it at everything.
Frontier models become the exception, open weights become the default. That's a pricing problem for every lab heading to IPO.
h/t @OliverMolander
Would you move your team off frontier APIs if the bill dropped 90%?
@corey_robi90189 Spot on! We are indeed adding these checks in the future episodes. In this series we show how to build working agents step by step (or should I say, error by error)
Roy, our IT helpdesk agent, said it had created a new agent, Arturo. It hadn’t: it misread a failed tool result and the claim went into its memory.
Video 3 of our AI agent series. Thinking about bringing AI agents into your team? Book a discovery call with us. Links in reply.
When you migrate a legacy system, it is tempting to fix things along the way. Don’t do it! Short version of my video “Migrate First, Re-architect Later”.
Thinking about a legacy migration? Book a discovery call with us. Links in the reply.
@Hermessino_AI Thank you @Hermessino_AI we will implement this in future episodes. With this series I aim to show how things go wrong because I think that often we just show the final solution and not the steps to get there
Legacy code is simply code without tests (Michael Feathers). Migrate a 20-year-old system with zero tests: how do you check that the new one behaves like the old one? I call it the oracle problem. https://t.co/0EfMDs3jBB
#rpg#legacymodernization#cobol
Our AI helpdesk agent, Roy, remembers our conversations, but not the ones he has with my colleagues. But we found a trick. I explain it in video 2 of our ongoing AI agent series.
Bringing AI agents into your team? Book a discovery call. Links and full episode in the reply.