Luis, great question – and you’re actually closer to understanding this than you think.

You said I’m adding complexity because the process is possible, and what humans can do is
only the practical. That’s exactly right. That’s the whole point.

But here’s what I’m not doing: I’m not adding complexity to TOC’s logic. The logic stays the
same. Find the constraint. Subordinate everything to it. Exploit it. Don’t touch the
non-constraints. Goldratt’s reasoning is fully intact in VSS. The Velocitizer runs TOC rules,
not optimization equations.

What’s different is the execution layer. With humans, you had to simplify the execution
because people can only hold so many variables at once. Schedulers work with five or six
factors when the real schedule requires fifteen. So you dropped the others – not because they
didn’t matter, but because humans couldn’t manage them under pressure. That’s the gap
between practical and possible.

AI doesn’t have that gap. So you can put the steps back without asking people to hold more in
their heads.

On your APS question – this is important. APS doesn’t work with TOC because APS tries to
optimize locally without honoring the constraint. It calculates the “best possible” schedule
for every work center simultaneously, which violates everything Goldratt taught about
subordination. You end up with a locally optimized mess that starves the constraint or floods
it.

VSS+iVSS is not APS logic. The AI is executing TOC rules – it’s just executing them faster,
more consistently, and across more variables than a human scheduler can track. The constraint
is still the organizing principle. Subordination is still the method.

Are we killing TOC? No. We’re finishing it.

Goldratt had to stop at “keep it simple enough for humans to run.” That was never the
destination. It was the limitation. The destination was always: understand the system deeply,
find the leverage points, manage them precisely. AI just removes the ceiling that forced us
to simplify the execution.

But you run TOC, the weakest link that’s the basic model. You are saying that TOC needs more complexity in order to be applied. But it’s not the APS way trying to get every factor in order to produce the best possible schedule ?

We know APS doens’t work as TOC does, because it doesn’t consider constraints the way TOC does. But with AI are you saying the APS logic can result ? To get every variable that humans can’t handle in the process – and so get the optimum?

Are we not killing TOC reasoning anf logic ?

Best Regards
Luis CRistovao

It is a renovation of the Goldratt’s basic, original Change Matrix, into a more flexible, and more powerful, “pro” and “max” versions, the Change Matrix +

Hi Noah,
I can’t find a better version.
Here’s a summary of drum buffer rope you may find helpful: https://www.velocityschedulingsystem.com/blog/drum-buffer-rope/
Lisa

Hi Karol,
As stated in the articke, Kanban does not use time buffers but space buffers. I’m not familiar with POLCA but any approach that puts buffers at every process step will have slower flow than DBR as more buffers = more WIP = longer flow time.
Best Wishes,
Dr Lisa