One of the most interesting articles I’ve read recently wasn’t written for learning leaders at all. It was written for educators and explored a deceptively simple idea from computer science and engineering: not everything can be parallelized.
The author applied this insight to human cognition, arguing that learning has hard bottlenecks. You can’t just pile on more activities, more modalities, more content, and expect learning to scale. Some mental work must happen sequentially: understanding before application, integration before transfer. When we ignore that, we get overload and the illusion of learning.
What struck me is how cleanly that idea transfers, not just to learners, but to learning organizations themselves.
If learning has limits, then learning systems do too.
Learning systems are not infinitely scalable
Many instructional design, development, and delivery management systems are built on an unspoken assumption: if something isn’t working, the answer is to add more.
More content.
More tools.
More activities.
More modalities.
More assessments.
From a systems perspective, this is an attempt to increase parallel processing. Work is distributed across teams, platforms, and timelines in hopes that learning outcomes will improve through sheer volume and velocity.
But just like cognition, learning systems have non-parallelizable work.
You cannot parallelize:
- Sense-making
- Conceptual dependency
- Feedback interpretation
- Judgment and decision-making
- Reflection and adjustment
These are sequential processes. They create natural choke points in the system and no amount of additional content or tooling can bypass them.
When systems move faster than understanding
This mismatch shows up in familiar ways:
- Courses completed, but performance unchanged
- High satisfaction scores with low retention
- Learners “checking boxes” without building capability
- Instructors racing the clock to cover material
- Designers optimizing for efficiency instead of flow
From the outside, the system looks productive. Inside, learning is fragile.
In engineering terms, the system is optimized for throughput, not stability. It moves people through the pipeline faster, but it does not improve what comes out the other end.
Designing for critical paths, not content volume
If we take the limits of parallelism seriously, learning systems need to be designed less like content factories and more like engineered processes with critical paths.
The key question shifts from:
“How much can we deliver at once?”
to:
“What must happen before something else can work?”
That reframing changes everything.
Instructional design becomes an exercise in mapping conceptual dependencies, not just sequencing topics. Development focuses on reducing friction at known bottlenecks rather than maximizing asset production. Delivery prioritizes pacing, pause points, and feedback over coverage.
This is not about slowing learning down for its own sake. It’s about preventing downstream failure by respecting upstream reality.
Manufacturers learned this lesson decades ago: fixing defects at the end of the line is far more expensive than stabilizing the process earlier.
Learning systems are no different.
Feedback Loops as Quality Gates
One of the most dangerous assumptions in learning systems is that feedback automatically exists because data exists.
Surveys, quizzes, analytics dashboards, and completion reports are often mistaken for feedback loops. But real feedback requires interpretation, judgment, and action. Those steps are inherently sequential.
In well-designed systems, feedback loops act as quality gates:
- They slow the system down intentionally
- They test for variance and drift
- They force decisions before proceeding
When feedback loops are rushed or treated as optional, errors compound. Misunderstandings propagate. Confidence grows faster than competence.
Fast systems with weak feedback don’t just fail quietly; they fail at scale.
A systems role hiding in plain sight
This is where many learning organizations feel tension but lack language.
They sense that:
- Designers are stretched between creativity and control
- Facilitators are compensating for system gaps in real time
- Quality assurance happens too late
- Leaders feel pressure to “do more” instead of “design better”
What’s missing is not effort or expertise. It’s systems stewardship.
Someone needs to protect the non-parallelizable parts of the system. Someone needs to design flow, identify critical paths, and balance speed against integrity. That work looks less like content creation and more like engineering.
Not mechanical engineering… but learning process engineering.
A respectful conclusion
None of this is a critique of people. It’s a recognition of reality.
Learning professionals are deeply committed to their craft. When systems struggle, it’s rarely because individuals don’t care or don’t know enough. It’s because the system quietly ignores the limits of how learning (and organizations) actually work.
Cognitive science reminds us that understanding has a pace. Systems thinking reminds us that ignoring that pace doesn’t make learning faster, it makes it weaker.
The opportunity for learning organizations is not to do more in parallel, but to design systems that honor both human cognition and system constraints.
That’s not a step backward.
It’s how durable learning is built.
Blog Index Summary
Learning systems, like human cognition, have hard limits on what can be parallelized. This article explores how ignoring those limits leads to fragile learning outcomes and how instructional design, development, and delivery systems can be strengthened by respecting critical paths, feedback loops, and system constraints.
Newsletter Blurb
What if the reason learning systems struggle isn’t effort or expertise but a hidden assumption that learning can be endlessly sped up? This article explores why some parts of learning simply can’t be parallelized, and what that means for learning organizations.
X (Twitter) Posts
- Learning systems fail quietly when they assume learning can be endlessly sped up. Some work (e.g., sense-making, feedback, judgment, etc.) can’t be parallelized. Ignoring that doesn’t make learning faster. It makes it fragile.
- More content, more tools, more activities often look like progress. In reality, they’re attempts to bypass the non-parallelizable parts of learning. Understanding doesn’t scale that way.
- Fast learning systems with weak feedback loops don’t just fail, they fail at scale. Feedback isn’t data collection. It’s interpretation, judgment, and action.
- Instructional design isn’t just creative work. It’s systems work. Mapping dependencies, protecting critical paths, and designing for flow matters as much as content quality.
- You can’t parallelize understanding. But you can design learning systems that respect how learning actually works.
LinkedIn Posts
- Many learning systems are optimized for throughput, not stability. When organizations try to speed learning up by adding more content and tools, they often bypass the very processes that make learning durable. Respecting critical paths and feedback loops isn’t inefficiency, it’s quality assurance.
- Cognitive science tells us learning has bottlenecks. Systems thinking tells us organizations do too. When instructional design and delivery systems ignore those constraints, performance suffers even when participation looks strong.
- Learning organizations don’t need to do more in parallel. They need to design better systems; ones that honor human cognition, protect feedback loops, and prioritize understanding over speed.
