Working with AI: practical guides
Task-focused guides with checkable results, examples and sources. Each language edition has its own collection, not a promise that every article is translated.
Model guides
A separate workflow, limits and result checks for each model. Language editions cover the same model through their own practical examples.
Dikaya Flash: a practical image-review workflow
Use Dikaya Flash for evidence-led screenshot and image review. Learn the exact model ID, native vision boundaries, follow-up prompts and API precautions.
Dikaya Pro: compare long documents with traceable findings
A practical Dikaya Pro guide for comparing policy revisions and long document sets: stable references, contradiction checks, context boundaries and API handling.
Dikaya Pro 2: code review built around reproducible tests
Review code with Dikaya Pro 2 using explicit contracts, small diffs and reproducible tests. Understand the modified GLM-5.3 attribution and demo boundaries.
GLM 5.3 in Dikaya: review multi-file API contracts
Use seller-glm-5-3 for a traceable multi-file contract review. Cover webhook schemas, declared effort settings, bounded output and verification of proposed tests.
GLM 5.3 Flash: repeatable editing with meaning checks
A GLM 5.3 Flash workflow for normalizing event listings without changing meaning: source IDs, style rules, semantic checks, effort controls and request recovery.
Qwen3.8 Flash Next: structured extraction and tariff checks
Plan evidence-linked extraction with seller-qwen-3-8-flash-next. Define null rules, validate records and distinguish catalogue identity from verified availability.
DeepSeek V4.1 Flash: design an API state machine
Use the modified DeepSeek V4.1 Flash identity for explicit API state design: transitions, uncertain outcomes, acceptance tests and bounded implementation review.
DeepSeek 1017 Safeguards: summaries with a coverage audit
Create traceable summaries with DeepSeek V4.1 Flash 1017 Safeguards. Preserve decisions, dissent and unresolved actions, and understand its original-model identity.
More practical articles
LLM API timeouts: avoid paying for blind retries
A practical timeout workflow for LLM APIs: distinguish HTTP rejection from a lost response, disable hidden SDK retries, and reconcile uncertain requests.