9 min read
Reasoning Models and Deep Reasoning in LLMs: Chain-of-Thought, Tree of Thoughts, and Test-Time Compute
Language models don't reason. Not in the way humans do.
Design system engineering · POS · Payments · Security
Corebaseit
I’m Vincent Bevia. I work in payments at MultiSafepay (part of Ant Group), and I’ve spent years on design system engineering for POS, payments, and security — architecture, EMV, cryptography, and the infrastructure behind card-present and digital acceptance. I’m the author of Point-of-Sale Systems Architecture and The Obsolescence Paradox.
This site is where I write about payment platforms as systems: channels, orchestration, HSM and tokenisation, acquiring rails, and the ops layer that makes settlement and audit trustworthy. Also AI in payments, and the EE / telecom foundations underneath.
Less corporate, more personal. Opinions, reflections, and the kind of things I’d talk about over coffee.
Topics I care about
- Design system engineering — POS, payments, and security as one stack (overview)
- POS & EMV — terminals, SoftPOS, chips, contactless, certification
- Cryptography & security — DUKPT, HSMs, PIN, and key management
- AI in payments — where it helps, where it doesn’t, and what’s coming next
If you’re into payment platforms, POS, or security architecture — or just curious how the pieces fit — welcome.
Latest field notes
Writing on payment platforms, security, and system design
7 min read
Why PIN Still Matters in Card-Present Payments
EMV authenticates the card. PIN still helps authenticate the cardholder.
7 min read
Why EMV Chip Cards Resist Cloning: The ARQC Mechanism Explained
Every chip card generates a unique cryptographic proof each time you tap or insert it. That proof is why cloning a chip card’s transaction capability is effectively …
5 min read
Scheme-Level Signaling: How Networks Identify Unattended Transactions
PCI PTS classifies devices at the hardware level. But Visa and Mastercard classify transactions at the message level. No scheme inspects the physical terminal to determine whether …
7 min read
Transformers vs. Diffusion Models: Not All AI Is the Same
We casually say “AI can write, AI can draw, AI can code” as if it’s one thing. It’s not. Two of the most talked-about model families in AI today solve …
8 min read
Stochastic, Entropy & AI: From Thermodynamics to Information Theory to Modern Machine Learning
I was listening to a podcast the other day about AI and the mathematics behind it — especially stochastic processes, entropy, and probability — and it immediately drew me in. With …
5 min read
Offline EMV vs Store-and-Forward: Two Different Mechanisms, One Confusing Name
When a POS has no connectivity, merchants still need to accept payments. The industry uses terms like “offline transaction” and “offline processing” loosely …
8 min read
L3 Certification Paths Are Not Created Equal: C-TAP, SmartPOS, and SoftPOS
When people talk about “L3 certification,” they often treat it as a single, uniform process. It isn’t. EMV Level 3 focuses on validating integration of the …
7 min read
BeviaLLM: Building a Language Model From Scratch With Python and NumPy
There’s a gap between using language models and understanding them. You can call an API, get a response, and build a product on top of it — without ever knowing what happens …
5 min read
AI as an Amplifier, Not a Replacement: Why Domain Expertise Matters More Than Ever
There’s a narrative floating around that AI will replace domain experts. That if the model can generate code, write architecture docs, and explain EMV flows, then maybe you …