When the learning wasn't the problem, the publishing workflow was
After six months of Weekly Learnings, I realised reflection worked but public communication didn't. I added an explicit editorial reasoning layer between reflection and publication.
This week’s focus
This week I paused the CAS Product Discovery work to revisit another experiment that had quietly been running for almost six months: publishing Weekly Learnings.
The original objective was simple—capture what I was learning while building products and exploring Enterprise Architecture in the age of AI. Looking back, the learnings themselves were valuable, but the public outcome wasn’t achieving what I had hoped. The posts helped me think more clearly, yet they rarely generated meaningful professional conversations or communicated the broader architectural thinking behind the work.
Rather than changing the writing style, I wanted to understand whether the workflow itself was the real constraint.
What actually happened
I reviewed the entire publishing process from first principles.
Instead of treating LinkedIn as the final formatting step, I reframed it as an editorial problem. Weekly Learning and public communication serve different purposes: one captures reflection; the other requires deliberate editorial judgement.
This led to designing and validating a new Editorial Pipeline consisting of two distinct capabilities:
- Editorial Selection, responsible for deciding what is worth communicating publicly by combining Weekly Learning with professional context, active experiments and editorial history.
- Editorial Articulation, responsible only for communicating that selected position faithfully for a specific audience.
The workflow was implemented and validated across multiple historical Weekly Learnings using both Claude Code and Google Antigravity, then refined through several PDCA cycles before becoming stable.
Key trade-offs
I deliberately chose not to continue expanding the Editorial Pipeline once the architecture proved effective.
Instead of exploring larger ideas—such as an Editorial Knowledge Base, Professional Knowledge Graph or automated portfolio synthesis,I captured them for the Ideas to Life roadmap and returned focus to validating the original experiment.
Similarly, I kept the articulation workflow intentionally simple. Rather than introducing additional writing agents, I invested effort in making the editorial reasoning explicit and inspectable before the writing even begins.
What changed in my thinking
The most important insight was that the Weekly Learning process wasn’t underperforming.
It had always been optimised for reflection.
The weakness was assuming that reflective writing could be transformed directly into effective public communication.
The real missing capability was editorial judgement.
Once that judgement became an explicit, inspectable workflow, the writing itself became dramatically simpler. Instead of asking an LLM to simultaneously decide what was worth saying and how to say it, those responsibilities became separate architectural concerns.
This also reinforced a broader pattern emerging across my work: many problems improve not by adding more AI reasoning, but by making expert reasoning explicit through well-defined contracts and workflows.
Architecture signals
- Separate editorial reasoning from editorial articulation.
- Reflection and publication optimise for different outcomes.
- Explicit contracts reduce dependence on model behaviour.
- Evidence should drive professional positioning, not personal opinion.
- Reader relevance should precede implementation details.
- Small architectural refinements outperform increasingly complex prompts.
Key takeaways
- Reflection and communication are complementary capabilities, not interchangeable ones.
- Better public communication starts with better editorial reasoning, not better writing prompts.
- Separating what to say from how to say it produces more consistent outcomes.
- Explicit workflows make AI-assisted reasoning easier to inspect, evaluate and improve.
- Product thinking applies as naturally to publishing workflows as it does to software systems.
Assumptions invalidated
- Publishing more consistently would naturally improve professional engagement.
- The quality of the LinkedIn posts depended primarily on better writing prompts.
- Weekly Learning should be optimised for both personal reflection and public communication simultaneously.
- Improving the writing layer alone would solve the underlying problem.
System evolution
- The publishing workflow evolved from a two-stage process (Weekly Learning → LinkedIn) into a three-stage system (Weekly Learning → Editorial Selection → Editorial Articulation).
- Editorial reasoning became an explicit capability rather than an implicit responsibility of the writing prompt.
- The publishing system now uses structured contracts and iterative validation in the same way as the architecture workflows it documents.
Looking ahead
With the first phase of the Editorial Pipeline complete, it becomes the default publishing workflow for future posts.
The next step is to return to the paused CAS Product Discovery work with a stronger communication capability in place. The broader vision—an Editorial Knowledge Base and Professional Knowledge Graph—has been captured as a separate Ideas to Life experiment and will be prioritised alongside the rest of the roadmap based on value, effort and strategic fit.
Related Experiments
Note:
- This Weekly Learning was produced using the Ideas to Life Weekly Learning system map
- This Weekly Learning marks the conclusion of the first Editorial Pipeline experiment. Future publications will follow the new editorial reasoning workflow while continuing to evolve through observation and iterative refinement.