Run the same job brief on grok.com and through the Responses API
Judge both outputs with the same rubric
Keep the key out of the artifact
The API does not make the brief better. A sloppy job card is still sloppy over HTTP.
Why this matters
People expect “API Grok” to be smarter than grok.com. It is the same family of models with fewer UI guardrails and a bill. If Applied taught you a job card, use that card here. If you never did Applied, write a six-line brief for a job you actually have this week.
You will run it twice: once in grok.com, once via POST https://api.x.ai/v1/responses. You will say which output you would actually use, and why.
Teach
Keep the brief identical. Same constraints. Same “must not invent.” Same stop.
On grok.com you can attach a file. On a first API call you probably cannot, unless you add the Files API later (docs.x.ai). For this lesson, paste the brief as input text. If the job needs a PDF, summarize the PDF yourself first or wait until you read the Files guide. Do not fake a file upload.
Judge both outputs with the Applied habit: HOLD / WRONG / UNVERIFIED on any fact you did not give it.
Worked example
A realtor’s job card: “Rewrite this listing paragraph. 80 words. Do not invent square footage. Do not name a school.” grok.com returns 92 words and a school. The API return is shorter and still invents a porch. She rejects both, tightens the brief, runs the API again. The artifact is the brief, both outputs, and “I would send neither until the porch is gone.”
Your turn
Same brief, two surfaces.
1. Paste my job brief (I will provide it).
2. I will also paste Grok.com’s answer and the API’s output_text.
3. Score both: usable / needs edit / reject. Name one invented fact if it exists.
4. Stamp: “Both surfaces used the same brief. No API key is in this paste.”
Artifact to paste
Brief + both outputs (or API error) + scores + stamp.
Rubric
Pass: Same brief. Two surfaces. A judgment. No key.
Fail: Only grok.com. Only “the API is cooler.” Invented endpoint.
Common failure
Changing the brief between surfaces so the API “wins.” You measured nothing.
What’s next
A tiny script you will run again next week — the capstone.
Your turn
grok.com
Your turn
Copy-ready prompt
Same brief, two surfaces.
1. Paste my job brief (I will provide it).
2. I will also paste Grok.com’s answer and the API’s output_text.
3. Score both: usable / needs edit / reject. Name one invented fact if it exists.
4. Stamp: “Both surfaces used the same brief. No API key is in this paste.”
Artifact
Brief + both outputs (or API error) + scores + stamp.
Read the lesson if you want. Submitting an artifact and marking complete needs a seat.