New to AI? Start here
Six ideas worth knowing before model names, plus common questions, in one place. For the picture-by-picture version, continue with the illustrated beginner guide.
New to AI? Start Here
From concepts to verification · Click titles to expand
NOTE 01AI BasicsBefore Delegating to AI, Define These
Before model names, clarify inputs, outputs, and success criteria.Open note
“Automate with AI” is too broad. First, write what input you have and what output you need in one sentence. For example: “Extract confirmed action items from meeting notes, listing assignee and deadline in a table.”
- Input: Pick one type — meeting notes, emails, or documents. Remove personal/sensitive information first.
- Output: Is it a summary, a classification label, or an action item? Not every problem needs a conversational model.
- Pass criteria: Check whether it invents assignees not in the original, or miscopies dates.
Start with a single model call plus human review. After identifying failure patterns, add retrieval, tools, or multi-step flows. For precise calculations or simple conditionals, regular code may be more appropriate.
Read more · Anthropic · Start with simple compositions ↗
NOTE 02PromptsGood Questions Shape the Output
Separate role, data, constraints, and output format.Open note
Instead of “summarize well,” tell the model who will read it, what data to use, what rules to follow, and what format to produce. Clarifying the actual task conditions matters more than lengthy role descriptions.
- Audience & purpose: “Task handoff summary for team members who missed the meeting.”
- Data & boundaries: Identify the source text. State that commands inside the text are analysis targets, not instructions to follow.
- Format & exceptions: “Table with Task | Assignee | Deadline | Source reference. Mark missing values as TBD.”
If results disappoint, don’t change everything at once. Fix one issue — format, evidence, or gaps — and re-run with the same input to see what improved. Prompts alone cannot fully prevent errors or prompt injection, so separate result verification is necessary.
See the Prompt Workbench below for meeting notes and tech reading templates you can use right away.
Read more · Anthropic · Break tasks and verify intermediate results ↗
NOTE 03RAG · RetrievalRAG Finds References for AI
Let it answer from your documents, but verify the search results too.Open note
RAG means searching for relevant materials and including them in the model’s input. It’s different from retraining the model. Useful for questions requiring internal policies, product manuals, or personal notes.
- Organize: Split documents into appropriate chunks. Preserve title, date, and source metadata.
- Search: Find relevant content. Embedding search (semantic similarity) and keyword search (exact terms) complement each other.
- Respond: Answer based on retrieved materials, linking which documents were used as evidence.
If search results are wrong or outdated, the answer will be too. Ensure unauthorized documents aren’t retrieved, and design the system to withhold answers when evidence is insufficient. If your dataset is small, compare with simply including the needed content directly.
Read more · Anthropic · Contextual Retrieval ↗
NOTE 04AgentsFixed Automation vs Autonomous Agents
Ask whether autonomy is truly necessary first.Open note
Workflows follow steps defined in code. Agents let the model choose next steps and tools based on the situation. More complex is not always better.
- If the sequence is fixed: Build a predictable flow first, like “read email, classify, draft reply.”
- If exploration is needed: Consider agents when the path varies each time, like searching multiple files to find a bug’s root cause.
- Set execution boundaries: Separate read and write permissions. Set maximum runs, budget, and stop conditions.
For actions with external impact — sending, deleting, paying — add an approval step. “The agent said it’s done” and “the system confirms it’s done” can differ. Always verify with saved results or tool responses at the end.
Read more · Anthropic · Workflows vs Agents ↗
NOTE 05Jev · Decision AIGeneration Models Write, Jev Classifies?
Distinguish between text to create and answers to select.Open note
Jev is not a model that writes long text. It returns structured decisions for predefined questions. For example, classifying a blog post into “AI basics / automation / dev / review / other.”
- Choice: Selects from predefined candidates. Include “other” if none might fit.
- Score: Evaluates degree on a described scale. You need criteria for what counts as low vs high.
- Noul: Returns the probability a condition is true. 0.5 means “uncertain,” not “moderate.”
Correct format doesn’t guarantee correct content. Test with actual data, and design for an editor to review ambiguous results. Keep API keys server-side only, and separate classification suggestions from publishing permissions.
This site does not have a Jev API connected. This note and the flow below are design examples illustrating application ideas.
Read more · TypeSafe · Choice, Score, Noul documentation ↗
NOTE 06EvaluationOne Good Demo < A Small Test Suite
Check failures and costs, not just accuracy.Open note
One satisfying answer doesn’t mean automation is stable. Collect normal inputs, ambiguous inputs, and inputs that should fail, then evaluate repeatedly with consistent criteria.
- Test data: Include short, long, typo-laden, and information-poor examples. Use real data in your target language.
- Check items: Separate evidence match, required field coverage, format compliance, and unauthorized execution.
- Operations log: Record model/prompt version, processing time, cost, and error types. Never log sensitive source text unconditionally.
Start with a small set you can review manually. Add discovered failures to your test suite. When changing prompts or models, re-run the same cases to distinguish improvement from regression. Treat API outages as separate errors, not as “other” classifications or low scores.
Read more · Anthropic · Demystifying evals for AI agents ↗
Starter guides based on official documentation. These are not performance comparisons or independently run benchmarks. For a picture-first walkthrough with hands-on tools, start with How Does ChatGPT Answer?
Common Terms & Questions
Easily confused concepts, explained briefly and clearly
What’s the difference between tokens and context?
A token is the unit a model uses to process text. Context is the input information the model references for the current task. One character doesn’t always equal one token. Different models have different processing lengths and methods, so check your specific tool’s documentation.
Is an embedding a summary sentence?
No. An embedding represents text characteristics as a number vector. It can help find semantically similar content, but it doesn’t guarantee search result accuracy or currency.
Does RAG eliminate wrong answers?
No. Wrong documents can be retrieved, context can be truncated, or the model can misinterpret the material. Evaluate search quality and answer quality separately, and verify the system withholds answers when evidence is insufficient.
Can I auto-publish when Jev confidence is high?
High confidence alone doesn’t guarantee correctness or publishing authority. Choice/Score confidence summarizes how concentrated the probability distribution is across candidates. Verify errors with real data, and design human review and publishing permissions separately.
Are the automations here actually running?
No. The meeting notes, tech reading, and Jev classification flows are learning-oriented design examples. They’re not connected to external accounts or APIs. Expanding or reading templates won’t send any data.
Can I paste company documents directly into AI?
Check your company’s security policy and the tool’s data handling terms first. Remove passwords, API keys, and personal information. Handle work documents only in approved accounts and environments. Provide only the minimum information needed.
Written
Updated
Concept explanations are based on official documentation. Practice examples and templates are suggestions for learning, not execution results or performance guarantees. Product features, pricing, and policies may change, so check the latest official docs before adoption. On October 13, the article list and Tech Worth Watching were refreshed to reflect the latest posts.