Some companies make noise. Others build systems. We try to stay in the second category.
This summer marks a dense phase of development at Terlouw AI. Under the surface, a lot is shifting—inside our models, our partnerships and how we think about applied intelligence. We’re sharing what we can. The rest will follow when it’s ready.
Contact update
Our new phone number is 088 – 6543 130.
A formal partnership: Inmeso Artificial Intelligence
We’ve entered into a strategic collaboration with Inmeso Artificial Intelligence. Based in Amsterdam, Inmeso has earned a strong reputation for real-world AI systems that are as practical as they are technically sound.
The focus of this partnership is to co-develop next-generation tooling for decision support, synthetic reasoning, and high-integrity automation. From dynamic context engines to verified language pipelines, we are aligning efforts across research, development and client delivery.
Jointly, we’re building systems that understand more than just prompts. And we’re making sure they scale responsibly.
Persona AI: Interface, meet intelligence
What began as a push toward photorealistic digital avatars has grown into something far more dynamic.
Persona AI is evolving into a full-spectrum interactive layer. Behind each face is a modular intelligence system that can adapt tone, intent and non-verbal signaling in response to user type, emotional state or environmental trigger.
Clients are already experimenting with Persona in roles that require speed, subtlety and consistency: onboarding flows, legal pre-screens, multilingual instructional walkthroughs. We’re now testing new interaction states, calm versus urgent, formal versus friendly, reactive versus passive, and refining how the avatar can shift in real time while preserving continuity.
If the surface looks human, the foundation must act like one too.
Lexiana 3.0: Launching september 1st
Lexiana 3.0 will roll out on 1 September 2025.
This update brings full support for both Dutch and Spanish-language legal domains, with jurisdictional specificity and improved clause-level comprehension. The citation engine has been redesigned from the ground up. Expect deeper traceability, better multi-document reasoning and noticeably sharper accuracy across regulatory questions and contract analysis.
We’ve also implemented improved model reasoning around ambiguous or underspecified queries—especially in structured uploads like leases, terms and procurement frameworks.
Meanwhile, the Lexiana iOS app continues to outperform expectations. Now used by 63 percent of our user base, it delivers the same AI capabilities in a fast, encrypted mobile interface. Usage is rising among legal teams that need answers anywhere, not just at their desks.
Inside the lab
Some projects don’t need names yet. Others have them, but shouldn’t be Googled. Here are a few things we’re exploring,without spoiling what they become.
There’s a framework in testing that lets models remember less, but mean more. It helps them make sharper decisions with shorter context and reassemble meaning across silos. We’re calling it a compression strategy, but it may evolve into something more foundational.
Another internal tool is beginning to understand not just what’s visible in a document or image, but what’s implied. The early results are abstract, but promising. Especially in use cases that involve multi-agent coordination and legal gray zones.
And we’re experimenting with a guidance layer that acts more like a map than a model. Not all intelligence needs to speak. Some just needs to point.
These systems are still in the dark. But the lights are on.
AI conversations, every week
Starting October 2025, we’ll be releasing a weekly podcast co-produced with Inmeso Insights. The format will be raw, technical and quietly provocative.
Expect discussions about applied AI, the myths of “alignment”, open-source breakthroughs, edge deployments, false positives, prompt leakage and whatever else deserves a longer sentence than a tweet.
We’re not announcing a name yet. But the first few guests might surprise you.
Coming soon: terlouw.academy
We’re currently preparing the launch of terlouw.academy, a new learning environment focused on applied artificial intelligence.
This platform will offer curated courses, hands-on workshops and certification tracks built for professionals who need more than just AI theory. Whether you’re a legal specialist integrating Lexiana, a developer building with open models or a strategist designing future-proof workflows, the Academy is designed to teach what we actually use in practice.
Expect content that’s sharp, current and grounded in deployment reality. Masterclasses will be taught by our team and invited experts across law, tech and systems design. Certification programs will align with emerging standards in explainability, compliance and model control.
Brief internal notes
Our team has grown. Our infrastructure has matured. Our QA processes are now live-monitored and adversarially tested. We’ve invested heavily in compliance auditing and built new tools to trace model decisions back to input fragments.
The result is more robust, more observable and more explainable AI, delivered with the same minimalist philosophy we started with.
Thank you to the clients who continue to push us, and the engineers who don’t let us settle. This update marks only a small fraction of what’s in motion.
We’ll share more when it’s finished. Or when it breaks.