Writing
Field notes from the IrenicTech team on building AI agents, AI automation, and custom software, and on shipping production AI products that actually reach real users.
Cybersecurity
In October 2026, ASOS customers got a push notification from the ASOS app saying the company had been hacked. ASOS says it started with one employee tricked into handing over a login. No malware, no software flaw. Here is how that kind of attack works, and the handful of controls that stop it at a company of any size.
IrenicTech · Oct 27, 2026 · 6 min read
Cybersecurity
In one month, Railway, Zeabur and TiDB Cloud all disclosed that attackers reached data their customers trusted them with, including the passwords and API keys apps run on. If a developer built your app on a platform like these, here is what to check, what to rotate, and how to make the next leak cost you less.
IrenicTech · Oct 27, 2026 · 7 min read
Cybersecurity
Two critical PHP vulnerabilities, known together as PHP2Shell, let an attacker run code on a web server without logging in. Attacks started before the patch shipped, and most servers are still unpatched. Here is what a business owner needs to check, in plain language.
IrenicTech · Oct 27, 2026 · 7 min read
Cybersecurity
Ransomware had its busiest quarter in over a year, and the typical victim is not a bank or a hospital chain. It is a manufacturer, an IT firm or a professional services company with fewer than 200 people. Here is what the Q3 2026 numbers show, why backups alone no longer save you, and the short list of controls that matter most.
IrenicTech · Oct 27, 2026 · 6 min read
AI Automation
Almost nobody publishes the architecture of the automations they deliver, so almost nobody can say what one actually looks like. We publish nine, which means we can count them. Across 139 nodes and 148 edges in nine sectors, the systems fan out rather than funnel: 15 ways in against 31 ways out. And of those 31 endings, only 15 are the thing the system was built to do. The rest are refusals, holds, escalations and self-maintenance.
IrenicTech · Sep 23, 2026 · 7 min read
SaaS Architecture
For twenty years the question was whether a person could use your product. The new question is whether an agent can, because agents now shortlist, evaluate, operate and increasingly buy software on someone's behalf. Agent-readiness is not a feature you add later; it is a property of your data model, your API and your permission system. This is what it actually requires.
IrenicTech · Sep 21, 2026 · 8 min read
Agentic AI
Most writing about AI agents is either a demo or a warning. This is the working guide: what an agent is made of, how it reaches your systems through MCP, when one agent should become several, how much the model choice actually matters, why the majority never reach production, and what governance has to be in place before any of it is load-bearing. Every section links to the deeper piece on that question.
IrenicTech · Sep 21, 2026 · 16 min read
AI Security
AI created three security surfaces at once: code written faster than it can be reviewed, agents holding real credentials, and attackers using the same tools you are. Most security advice addresses one and ignores the other two. This is the working guide to all three, plus what the compliance picture actually requires rather than what the vendor deck claims.
IrenicTech · Sep 21, 2026 · 9 min read
AI Automation
TypeSafe AI released a model that cannot hold a conversation, which is the point of it. Jev takes unstructured state in and returns a typed decision with a calibrated probability, in 70ms to 500ms. We went through the 21 decision nodes in our published automation teardowns to see where a model like this actually belongs. The answer is six of them, and the reasoning matters more than the number.
IrenicTech · Sep 21, 2026 · 7 min read
Read us on Google
Tell Google you want our work first. It takes one tap, and IrenicTech is favoured in your Top Stories, AI Mode, and AI Overviews whenever you search the topics we write about.
Add to preferred sourcesStart a conversation
Share the essentials and we’ll reply within 4 hours with a real next step, not an auto-responder.
What happens next