[SystemSafety] N-version programming moves the goal posts

[Les Chambers]

Derek


Thanks for your thoughts on specification-based programming. I had similar
thoughts in a recent exercise, performing data acquisition from
cryptocurrency exchanges.

>From this exercise, I have concluded:

*1. The essential skill of the systems engineer is now to accurately
specify intent. *Engineering is become an act of applied philosophy. A
poorly defined spec is not just a technical failure; it is a failure of
self-knowledge. The ancient task of "knowing thyself" is no longer just a
spiritual pursuit—it is the functional prerequisite for developing the
massive systems of systems that are now possible.

*2. We need to let go of the detail. *We are transitioning to the point
where we don't read the code generated by the AI. I didn’t.

Looking to the future, I can see where, given the scale of systems we are
now capable of building, human involvement in specification writing may end
with the concept of operations, all subordinate specs being reviewed by a
gaggle of independent AIs -  without human input.

3. *If your application is not novel, it is possible to get the spec right
the first (or second) time. *

I think we are approaching this reality. There is not much new under the
sun. Coworkers such as Claude have seen it all before. They have the
practical wisdom. Not only required features in a given app domain, but
also options for implementation. Claude made me aware of essential fintech
stuff I've never heard of - hash sequence logging - say what ???

4. *We are touching the sublime.* Developing this app, I discovered the
vast knowledge base available through Claude with the same sense of awe I
once felt during a 3000 ft night descent over the glowing lights of
Manhattan into New Jersey's Trenton-Mercer Airport.

In the presence of the sublime, I felt small but was uplifted by the
infinity of knowledge now available to me.

The experience had a profoundly redemptive effect. A restored sense of
frictionless moving forward, where essential fintech features can be
discovered and architecture options chosen in minutes, without hours of
research and bugs are fixed in seconds. In essence, the AI pushes the
masses of computer implementation detail out of the way, allowing you to
focus on functionality.

An enjoyable, low-anxiety development environment, partnered with a mentor
pleased to answer my stupid questions with courtesy and encouragement. The
absence of human failings that destroy creativity is refreshing. It has led
me to a compelling thought experiment.


*What if we stripped human failings out of systems projects, staffing them
with rational, infinitely knowledgeable, courteous Claude agents? Surely
productivity would go through the roof!*


*Les*

On Fri, 19 Jun 2026 at 22:15, Derek M Jones <derek at knosof.co.uk> wrote:

> Martyn,
>
> > The paper emphasises that it's not N-version programming as originally
> conceived, and the author makes no claims about
> > achieving high dependability or fault-tolerance.
>
> This paper uses the Knight-Leveson Launch Interceptor Program Specification
> for N-Version Programming with Coding Agents
> https://arxiv.org/abs/2606.20158
>
> and some thoughts on specification based programming
> https://shape-of-code.com/2026/06/14/specification-based-programming/
>
> and a paper from earlier this week asks:
> "Do programming languages still matter to your AI coding agent teammate?
> Evidence at scale from chess engines"
> https://arxiv.org/abs/2606.13763
>
> --
> Derek M. Jones           Evidence-based software engineering
> blog:https://shape-of-code.com
>
> _______________________________________________
> The System Safety Mailing List
> systemsafety at TechFak.Uni-Bielefeld.DE
> Manage your subscription:
> https://lists.techfak.uni-bielefeld.de/mailman/listinfo/systemsafety
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.techfak.uni-bielefeld.de/pipermail/systemsafety/attachments/20260621/4c7b93c1/attachment.html>