What Machine Learning Actually Does, Without the Jargon
A plain-language walkthrough of how machine learning works, written for someone who has never studied computer science but wants to genuinely understand the concept.
read case study →
"
Case study breakdowns
AI in practice — what actually happened
These are documented accounts of real implementations. Each breakdown covers what problem was being solved, which approach was taken, and where the results diverged from expectations. No simplified narratives — just the specifics.
5
documented cases
6
AI domains covered
3
industries analysed
All case studies
Each entry is self-contained — read any in any order.
Neural Networks: What the Brain Analogy Gets Right and Wrong
A grounded look at what neural networks actually are, using a concrete scenario to separate the useful parts of the brain metaphor from the parts that mislead beginners.
read case study →
Supervised vs. Unsupervised Learning: A Decision That Shapes Everything
An analytical breakdown of two foundational AI approaches, explained through a real-world scenario that shows why the choice between them is not just technical but strategic.
read case study →
Training Data: The Part of AI That Quietly Controls Everything
A clear explanation of what training data is, where it comes from, and why its quality determines the ceiling of any AI system, written for readers new to the field.
read case study →
Overfitting: When an AI Learns Too Much From Too Little
A narrative explanation of overfitting, one of the most common failure modes in AI, using a concrete study scenario to show what goes wrong and how practitioners catch it.
read case study →How each case study is structured
Each entry follows a consistent format so you can compare across cases without re-orienting yourself. The structure separates the setup from the decision-making and the decision-making from the outcome.
Where a project produced mixed results, that tension is documented rather than resolved into a tidy conclusion. AI implementations rarely produce clean before-and-after stories, and the breakdowns here reflect that.
01
Problem definition
The specific constraint or gap that prompted the project, stated without framing it as a solved problem from the start.
02
Technical decisions
Which models, pipelines, or methods were chosen, and the reasoning behind those choices given the constraints at the time.
03
Observed outcomes
What the implementation produced in practice — including edge cases, failure modes, and areas that required manual correction.
04
Transferable notes
Observations that may apply to different contexts — not prescriptions, just patterns worth considering.
Written by practitioners
The cases on this platform are documented by Tobias Wrenfield, an analyst who has worked directly on AI integration projects across logistics, content classification, and decision-support tooling. The breakdowns draw on firsthand project records, not secondary reporting.
Tobias Wrenfield — lead analyst, Field Notes DailySubmit a case for review
If you have worked on an AI project and want to have it documented here, get in touch. Cases are selected based on specificity — generic overviews are not useful to readers, but detailed accounts of constrained problems often are.
Submissions are reviewed within two weeks. You will hear back regardless of whether the case is accepted. Confidentiality requirements are accommodated on request.