Articles

Articles.

What we learn building and proving software for organisations: legacy modernisation, cybersecurity and enterprise software.

Sovereign

Legacy modernisation

Understand old software. Build a replacement. Check its behaviour against the original.

Showing all 8 articles

Legacy modernisation

How to prove a rebuilt system does exactly what the old one did

How to tell whether a rebuilt system has been proved to behave exactly like the old one: five questions to ask any supplier, three ways organisations prove a move today, and a checklist.

Article 11 October 2026 4 min readRead

Legacy modernisation

We rebuilt a public mainframe credit card system and proved it works the same

We rebuilt the 13 batch programs of CardDemo, a public mainframe credit card application, and compared every printed line and every written record with the originals. All thirteen match.

Article 11 October 2026 4 min readRead

Cybersecurity

Proving what AI agents can reach, instead of watching for trouble

Most security tools for AI agents flag behaviour that looks unusual; Kern proves in advance what a set of agents can and cannot reach, and issues a certificate others can re-check.

Engineering note June 2026 4 min readRead

How we work

A team cannot reliably judge whether its own idea is new

Twelve ideas that had passed our internal checks went to an outside reviewer told to disprove them; all twelve were rejected or downgraded, mostly because someone had already published them.

Field report June 2026 4 min readRead

Enterprise software

The most important software is usually the part nobody sees

From bookkeeping to AI agents, the systems that matter most work out of sight; this note explains why all six Syntaxe products are built for that hidden layer.

Engineering note May 2026 4 min readRead

Enterprise software

AI agents change the main question in software design

Software used to wait for a person's click; now AI agents act on their own, so their limits must be set and proven before they are released.

Note May 2026 2 min readRead

How we work

Good ideas are rarely unique

When an outside reviewer checked twelve of our own ideas, almost all had been thought of before; lasting value comes from data, methods and engineering that others cannot easily copy.

Note May 2026 2 min readRead

Explore Syntaxe.