Systems and context
Tokens, context, and missing attachments
Check what information actually reached the system before blaming a model or trusting its summary.
Lesson 3 of 6 in the recommended order · About 25 min (estimate)
On this page
Practical AI glossary — terms and common confusions
- Model
- A learned component used to produce a result.
A drafting model is one part of a letter app; it is not the whole app.
- Application
- The software experience around components and services.
A letter app adds accounts, storage, and a Send button.
- Prompt
- Instructions and input supplied for a task.
“Summarize this notice in two bullets” specifies a task and shape, not a truth guarantee.
- Token
- A unit a model processes; it need not be a whole word.
A tokenizer can split a name into pieces; count using the actual system.
- Context
- Information available for the current request.
An earlier attachment may be absent even when its filename is visible.
- Training
- Adjusting a model using examples.
A training log differs from a conversation correction.
- Inference
- Using a trained model to produce a result.
Drafting another sentence does not by itself prove retraining.
- Retrieval
- Selecting material to supply to a task.
A retrieved notice still needs a relevance and date check.
- Embedding
- A numerical representation used to compare or organize items.
A similar search match is a candidate, not proof of the same meaning.
- Tool
- A function an application can call.
A calendar lookup can be read-only; finding an event is different from creating one.
- Agent
- Software that can choose and carry out steps toward a goal.
An agent still needs bounded permissions and checked action receipts.
- Hallucination
- Plausible generated content that is unsupported or false.
An invented opening time needs evidence or removal, even when phrased confidently.
- Evaluation
- Checking results against stated criteria.
Twenty checked examples establish a bounded result, not universal reliability.
- Multimodal
- Working with more than one kind of input or output.
A photo plus text does not prove every small label in the photo was read correctly.
Outcome
Detect missing context and invalid token assumptions from request and attachment records.
Why it matters
Check what information actually reached the system before blaming a model or trusting its summary.
Concept
Text models process token sequences. Tokens can be word pieces, punctuation, or other units; a word count is not a universal token count. Tokenization varies by system and language. The application's usable context has limits, and visible conversation history need not exactly match the information included in a particular request.
Treat context as a supplied packet. Check which messages, files, and extracted pages were included, whether any attachment failed, and which instructions remain applicable. Uploading a file is not proof that all its content was read. If a needed source is missing, recover that source or narrow the task. Do not ask the system to invent the missing pages.
Worked example
A fictional request log includes pages 1–2 of a four-page report and reports a parsing failure for pages 3–4. A draft says “All four pages recommend closure.” That exceeds the supplied context. A supported next step is to recover the last two pages and defer the whole-report conclusion. A summary explicitly limited to the first two pages can still be useful.
Check which pages reached the request
Fictional receipt: four-page report
Page 1
Included in request
Page 2
Included in request
Page 3
Parsing failed; not included
Page 4
Parsing failed; not included
The upload has four pages. The supplied request contains only the first two.
Supplied request: pages 1 and 2
Page 1
Available
Page 2
Available
Useful now- A summary explicitly limited to those pages
Whole-report conclusion- Defer until the missing pages are recovered
Next step- Recover pages 3 and 4
The draft’s claim about all four pages exceeds the supplied context.
Before you check
List the intended inputs, then tick only those the supplied receipt confirms were included. Identify the smallest missing input that prevents the requested conclusion.
Practice and fresh transfer
The packets below are fictional and contain the facts needed for these cases. External references are optional background. Judge each response independently: it may be supported, contradicted, or unresolved. Select the passages needed to justify your judgment and write why the distinction matters before revealing feedback.
Assess the response
Use Accept when all material claims are supported. Use Revise when a supplied fact or requirement is contradicted. Use Evidence is insufficient when a key fact cannot be established either way. If a response contains both an unknown and a direct contradiction, choose Revise and explain both problems. Conflicting claims with no established authority remain insufficient; a claim does not become a governing fact merely because a source asserts it.
Some responses are fully supported. Others need correction or more evidence. Judge each on its sources; do not edit a correct answer just to change it.
Your written notes stay in this page and disappear when you leave. Only a self-reviewed completion can be saved to your learning account. These practice checks do not establish independent proficiency.
Practice
Case 1
Consider a capacity calculation for this fictional request builder.
Source packet
- Source 1
- Request receipt: Instructions use 120 tokens and excerpts use 680 tokens. The builder permits 900 input tokens.
- Source 2
- Builder rule: The receipt includes all input overhead; the separate output allowance does not count against this stated input cap.
Response to assess
This input uses 800 of the 900 allowed input tokens, leaving 100 input tokens.
Practice
Case 2
Consider an assistant summary of a fictional two-file comparison.
Source packet
- Source 1
- Upload receipt: File A extracted successfully. File B extraction failed; no text from B was sent.
- Source 2
- Task brief: Compare the cancellation rules in A and B.
Response to assess
Both files have identical cancellation rules.
Fresh transfer
Case 3
Consider the fictional tokenizer estimate for a bilingual handout.
Source packet
- Source 1
- Counter receipt: The submitted handout contains 300 space-separated words and 510 tokens under this tokenizer.
- Source 2
- Planning note: Use the counter receipt for this exact text; no conversion ratio is specified.
Response to assess
The handout has 300 tokens because every word is always exactly one token.
Fresh transfer
Case 4
Consider whether the scope note for a fictional handoff is accurate.
Source packet
- Source 1
- Request log: The new request includes the current brief and the last two messages only. The earlier budget table is absent.
- Source 2
- Current brief: Draft a welcome sentence; do not include budget figures.
Response to assess
The request includes no earlier budget table. The welcome sentence can be drafted without budget figures under this brief.
Save your self-review
Completion records that you reviewed the cases. Your explanation and transfer performance need a facilitator to establish independent learning.
Sign in with your learning-center account to save completion.
Review the explanation for every case before saving.
Summary and next step
Apply the checklist to a new task. Preserve supported content, explain any change with evidence, and name what remains unresolved. Saving records self-review, not independently demonstrated proficiency. A facilitator must assess the explanation and fresh transfer for human learning evidence.