Build your AI skills with evidence
TrendsWhat is an independent English-language publication for people who want to use AI thoughtfully in real work. We explain the task, show the method, keep the evidence available, and name what remains uncertain.
Our promise is practical learning: a question you can investigate, a result you can inspect, and a checklist or workflow you can take into your own project.
The person behind the publication
TrendsWhat is operated by Jay Jung (Jihyuk Jung), a software developer based in the Republic of Korea. His work includes Imgobot, a learning platform; StockNote, an expense and asset-recording project; Send, a quiz app; MaybeTomorrow, a planner; and Review-in-Korea, a multilingual publication.
Our project case studies use inspected implementation details from these projects. They distinguish the operator’s project experience from new synthetic experiments conducted for this publication. We do not present code inspection as a customer study or claim production results that were not measured.
What you can learn here
Start with task design and context preparation, then explore research verification, prompt evaluation, structured outputs, document processing, and reliable tool actions. Publishing guides cover multilingual releases, original visuals, and the cost of accepted work.
The experiments include complete synthetic fixtures, executable scripts, recorded outputs, original figures, and limitations. Guides include worksheets; case studies explain a concrete implementation and the boundaries of the evidence.
How the articles are made
AI assists research organization, drafting, code preparation, and visual production. Articles are attributed to TrendsWhat as a publication. We disclose this assistance instead of presenting generated prose as an unaided personal diary or claiming independent human review that has not occurred.
An Experiment reports an executed procedure and its observed result. A Case Study identifies the project material inspected and says whether any part was executed. A Workflow or Visual Guide offers a reusable method; hypothetical examples and proposed tests are labeled. A source link supports a specific claim, not every conclusion in an article.
Evidence and original visual material
Result charts are generated from recorded data. Architecture diagrams are original redrawings of inspected mechanisms, and workflow diagrams are explanatory guides. Captions identify their role. Synthetic records are fictional; they are not customer data or evidence of real customer incidents.
A small test can reveal a useful failure without establishing a general success rate. Each article names important exclusions, such as untested runtime behavior, a small fixture set, or the absence of a language-model benchmark.
Corrections and updates
Send the article URL, disputed passage, and supporting evidence through our contact page. For reproduction problems, include the script version, runtime, and observed output without credentials or private records. We assess reports against the evidence and correct material errors in the affected text and artifacts.
Publication dates identify the first release. A modification date is updated for a substantive revision. Material corrections should explain what changed; dates are not refreshed merely to make an unchanged article appear new.
Independence and advertising
TrendsWhat may display Google-served advertising to support the publication. An advertisement is separate from our editorial recommendations and is not an endorsement. Any sponsored or affiliate relationship in editorial material must be disclosed where it appears.
Our articles are educational material, not a promise of professional outcomes, model accuracy, income, or platform approval. Readers should verify current tool behavior and adapt each method to their own constraints.