Posts

Can you honestly say…?

That you would deploy a mission critical system made using agentic tools controlling a hazardous resource without examining every line of code by an human expert? Agentic software development promises faster coding, automated testing, and increased productivity, but it comes with important tradeoffs. AI agents can introduce subtle bugs, security vulnerabilities, and technical debt while confidently producing incorrect solutions. They often lack the architectural understanding and business context needed for long-term maintainability, making human oversight essential.  As organizations adopt autonomous development tools, the focus shifts from writing code to verifying it. Success depends on strong governance, thorough reviews, and clear accountability. Agentic AI is a powerful accelerator, but treating it as a replacement for experienced engineers can create more problems than it solves.

The Expanding Attack Surface of Agentic Development

Agentic development is transforming software engineering by enabling AI agents to rapidly design, write, test, and deploy applications. While this dramatically accelerates innovation, it also creates a significant security challenge: an exponentially growing attack surface. As organizations use AI to generate hundreds or even thousands of applications, maintaining consistent security becomes increasingly difficult. Applications may be built using different frameworks, third-party APIs, open-source libraries, and cloud services, each introducing new dependencies and potential vulnerabilities. Without strong governance, security reviews can quickly become a bottleneck—or be skipped entirely. The speed of agentic development also means that vulnerabilities can be introduced and deployed faster than traditional security teams can detect them. Misconfigured permissions, exposed secrets, insecure APIs, and outdated components can all become entry points for attackers. When multiplied across ...

Agentic software quality

I recently attended an agentic developer conference that posed a few philosophical questions: Humans use a lot of software; humans have traditionally been associated with the software development process. Can agents produce better software for humans than humans themselves make with older tools? Is agentic software development dazzled by the promise of software improvement even though the real measure of quality is arguably a human judgement rather than a machine one?   Are human developers in the strongest position with hand crafting tools like classic IDEs or does having agents improve software quality, is the answer subjective and will it depend on the client’s preference?   Time will tell, keep in mind that the final preference is likely to come from users.

Massive scale automation for environmental cleaning.

At one time of day if you said to a hardware expert you are going to be able to fit billions of transistors onto a single silicon chip they would say it’s not possible at the moment and it might be possible in ten years.   This is the kind of feeling I get when I consider this idea: Massive scale automation, humanoids, robots cleaning the entire environment down to every leaf on every tree.   These robotic automated systems work silently throughout the night and return to their charging hangar before sunrise.   These robots repair the roads, unblock drains and gutters, clean windows, clean vehicles. Customer services might get a call like ‘The overnight automated cleaning didn’t clean my Bentley very well and I’d like to complain’

The internet killed the TV star

Amongst my varied and interesting academic life I have studied psychology and social science. When I do a comparison of the way people appear on tv and the internet I see that tv seems stale and the rising internet stars seem energised and motivated.   On a technical basis the tv is old tech and I believe that one of two things will happen to tv long term. 1. A reinvention 2. Devolution into a news service only for less technically able groups.   As my elderly mother told me recently the same thing happened to radio many decades ago.   Seems like the truer interactivity the internet provides leaves fans feeling closer to their celebrities more than tv ever could.

The smartphone as the centre of a user’s digital world

In 2026 the smartphone is an indispensable device supporting the users digital life, messaging, payments for goods and services, entertainment, business, location and directions, some limited productivity, shooting movies for services like YouTube. Battery life used to be very limiting but no longer, screen resolution used to be limited but is now excellent, the touch interfaces work superbly with high reliability in all conditions. Even the audio is pretty strong on modern smartphones, easily producing music and sound for movies. The apps ecosystem provides extensibility for the device, along with nearfield protocols to add hardware easily. Voice recognition is quite strong as is the language translation offering. In 1990 my mobile phone could make voice calls for 30 minutes per battery charge and that was it, no other functionality. Excellent progress in 35 years!

Flow of data

Image