xAI Grok Users Manual

As of …

A practical, step-by-step guide to xAI's Grok lineup — every current model, every product surface (Grok app, X integration, API, Imagine), and copy-paste prompt templates for the most common goals.

🎈 ELI5

Grok is xAI's AI helper — and its secret superpower is that it lives on X (Twitter) and can read what's happening on X right now. Ask it "what are people saying about [thing] today?" and it actually knows.

You can chat with Grok in the Grok app or on X.com, generate pictures and short videos with Grok Imagine, or call the API for your own apps. Pricing on attap.ai: Grok 4.1 Fast (1 credit, super cheap), Grok 4.3 (6 credits, best brain), Imagine Image (10 credits), Imagine Video (100 credits).

Getting started in 60 seconds

Pick your door: grok.com for the standalone app, x.com for chat inside X, docs.x.ai for the API.
Sign in with your X account or email. Grok is free to use with limits; subscriptions remove them.
Pick a model: Grok 4.3 (flagship, slow + smart), Grok 4.20 (fast frontier), Grok 4.1 Fast (cheap + huge context), Grok Imagine (image/video).
Tell Grok what good looks like. Goal, audience, format. Like every model — but Grok also responds well to direct, opinionated prompts; soft hedging gets soft answers.

Which xAI surface should I use?

Grok app / X.com

grok.com or X chat

Free chat with limits
Real-time X data access
Voice mode + Companions
Grok Imagine for images/video

Subscriptions

$8 → $300/month

X Premium / Premium+
SuperGrok Lite / Standard / Heavy
Heavy = multi-agent reasoning + biggest models

API (docs.x.ai)

Developer access

OpenAI-compatible endpoint
1M–2M token context windows
Live Search for X grounding
Imagine API for image & video

The five prompt fundamentals

Part	Purpose	Example phrase
Role	Frame the model's perspective	"You are a senior staff engineer reviewing a junior PR."
Goal	What "done" looks like	"Produce a 1-page exec summary I can paste into Notion."
Context	Background & constraints	"Audience: non-technical execs. Tone: confident, plain English."
Inputs	The raw material	Pasted text, attached file, URL, image, video, X posts.
Format	Shape of the output	"5 bullets, ≤15 words each, no preamble."

Grok-specific tip Grok's personality is more direct and less hedging than Claude or Gemini by default. If you want measured, balanced answers — say so explicitly ("Be balanced; surface the strongest counter-argument"). If you want spicy opinions — Grok is the right tool and you don't need to ask twice.

🎈 ELI5

Grok comes in three text sizes right now: Grok 4.3 (newest, smartest, slowest), Grok 4.20 (fast frontier with the biggest 2-million-word memory), and Grok 4.1 Fast (cheap and fast for simple stuff).

For pictures and videos, use Grok Imagine — it makes both, and the videos come with sound (footsteps, dialogue, ambient noise) without you asking. Grok's special move: it can read X (Twitter) right now, so it knows what's trending today.

The current Grok lineup

As of 2026-05-04, Grok 4.3 is xAI's flagship. The lineup splits into three text tiers (4.3 / 4.20 / 4.1 Fast) plus Grok Imagine for image and video.

Latest releases Grok 4.3 rolled out fully 2026-04-30 with a 40% input price cut, native video input, and 1M-token context. Grok 4.20 shipped 2026-03-31 with a 2M-token context. Grok Imagine API launched 2026-01-28 and added Extend from Frame on 2026-03-02. See What's new in Grok 4.3 below.

About these dates Dates pulled from xAI's announcements and docs.x.ai release notes. Pricing in USD per million tokens unless noted. Always confirm against docs.x.ai/developers/models before billing-sensitive decisions.

Frontier text/multimodal models

Model	API ID	Released	Best for	Pricing (in / out)
Grok 4.3 flagship	`grok-4.3`	2026-04-30	Hardest reasoning, deep research, agentic work. Native video input. Reasoning always-on (no effort dial).	$1.25 / $2.50
Grok 4.20	`grok-4.20`	2026-03-31	Fast frontier — speed-of-thought tool calling. Largest context (2M tokens).	$2.00 / $6.00
Grok 4.1 Fast	`grok-4.1-fast`	2026-Q1	Cheapest tier in the family. 2M context. High-volume, low-latency.	$0.20 / $0.50
Grok 4 (legacy)	`grok-4`	2025-Q3	Predecessor flagship. Still available; migrate when convenient.	See pricing page

Image & video — Grok Imagine

Capability	Notes	Pricing
Grok Imagine — Image	Text-to-image and image edits. 9+ styles (anime, cyberpunk, futuristic, kawaii, minimal art, etc.).	~1 credit/image (third-party); included in SuperGrok
Grok Imagine — Video	Text-to-video, image-to-video, Reference, Extend, Modify modes. Native synchronized audio — dialogue, SFX, ambient. Up to 15 sec per clip; chain via Extend from Frame for longer sequences. 480p / 720p output.	$0.05/sec for 720p w/ audio (~$0.50 for 10-sec clip)

Release timeline (chronological)

Date	Release	What changed
2023-07	xAI founded	Elon Musk launches xAI; mission: "understand the universe."
2023-11-04	Grok 1	First public release. Limited beta to X Premium+ subscribers.
2024-03-17	Grok 1 open-weight	314B-parameter Mixture-of-Experts model open-sourced.
2024-08	Grok 2 + Grok 2 mini	Major capability lift; image gen via partner (FLUX) integrated.
2024-12	Aurora image model	xAI's first in-house image generator (precursor to Imagine).
2025-02	Grok 3 + Grok 3 mini	"DeepSearch" mode debuts. SuperGrok subscription tier launches.
2025-07	Grok 4 + Grok Imagine (initial)	Multi-agent "Heavy" mode. Imagine launches with 6-second text-to-video clips with audio.
2025-10	Grok Imagine speed lift	Generation time slashed to under 15 sec; instant image generation added.
2026-Q1	Grok 4.1 Fast	Cheap-and-fast tier with 2M-token context.
2026-01-28	Grok Imagine API	Programmatic access — text-to-video, image-to-video, video editing.
2026-02-03	Grok Imagine v1.0	Video length extended to 10 seconds at 720p.
2026-03-02	Extend from Frame	Chain clips by reusing the last frame as next clip's start; longer sequences.
2026-03-25	SuperGrok Lite ($10/mo)	New cheap subscription tier with basic Imagine access.
2026-03-31	Grok 4.20	Fast frontier model. 2M context. Strong agentic tool calling.
2026-04-30	Grok 4.3	Current flagship. 40% input price cut, 1M context, native video input, always-on reasoning.

What's new in Grok 4.3 (2026-04-30)

Grok 4.3 is xAI's latest flagship and a meaningful rebalancing of the cost model for agentic applications. Highlights:

Area	What changed
Pricing	Input price cut ~40% from Grok 4. New rate: $1.25 / $2.50 per million tokens — competitive with GPT-5 and meaningfully cheaper than Opus 4.7.
Context window	1,000,000 tokens. (Note: 4.20 has 2M, but 4.3 prioritises depth-per-token over raw window.)
Native video input new	First Grok model to take video as direct input — no transcript pre-step required.
Always-on reasoning	Reasoning is built in and cannot be disabled or configured by effort level. (Different design choice from Opus 4.7's `xhigh` dial or GPT-5.4's five-level effort.)
No output cap	No max output token limit — well-suited to long-document analysis and long-form generation.
Agentic tool calling	Improved over 4.20 on multi-step tool use; Live Search can ground answers in real-time X data.

Pricing comparison

	Input	Output	Context
Grok 4.3	$1.25 /M	$2.50 /M	1M
Grok 4.20	$2.00 /M	$6.00 /M	2M
Grok 4.1 Fast	$0.20 /M	$0.50 /M	2M
For comparison at the frontier tier: Opus 4.7 = $5/$25, GPT-5.4 = $2.50/$15, Gemini 3.1 Pro = $2/$12. Grok 4.3 is the cheapest frontier-class model on USD-per-million-tokens.

Migration note Always-on reasoning means latency is higher than non-reasoning peers at small inputs. Do not treat 4.3 as a drop-in replacement for fast cheap models — for short, simple tasks, route to Grok 4.1 Fast instead.

How to pick a Grok model

Pick Grok 4.3 when…

The task is genuinely hard reasoning or research.
You need video as direct input.
You want frontier-class quality at the cheapest frontier price.
Long-horizon agentic work — multi-step planning and tool use.

Pick Grok 4.20 when…

You need a true 2M-token context window.
Speed matters — fastest agentic tool calling in the lineup.
You want frontier-class quality without always-on reasoning latency.

Pick Grok 4.1 Fast when…

Tasks are repetitive, unambiguous, high-volume.
Cost dominates ($0.20/$0.50 per M tokens — among the cheapest anywhere).
You still want a 2M context window for big inputs.
Real-time UX where latency matters more than depth.

Pick Grok Imagine when…

You want generated images or short videos.
Video must include synchronized audio out of the box.
You're chaining clips into longer sequences (Extend from Frame).

The "match the model to the task" rule Don't default to 4.3 for everything — it's smart but not fast. For triage, routing, classification, simple rewrites: Grok 4.1 Fast wins on $/quality. Reserve 4.3 for tasks where reasoning depth matters.

Grok Imagine — image & video deeper dive

Image generation

Modes: Text-to-image, image edit, reference-based, style transfer.
Styles: Photorealism, anime, cyberpunk, futuristic, whimsical, kawaii, minimal art, abstract, more.
Strengths: Fast generation; image-edit instructions work well.
Watch-outs: Text rendering inside images is not at Imagen 4 / gpt-image-2 level; for typography-heavy work, consider those.

Video generation

Modes: Text-to-video, image-to-video, Reference (style anchor), Extend (continue from a frame), Modify (edit existing).
Audio: Native synchronized audio generated alongside video — dialogue, sound effects, ambient. No separate TTS pass needed.
Length: Up to 15 seconds per clip; chain via Extend from Frame for longer sequences.
Resolution: 480p (basic plans) / 720p (paid).
Pricing (API): $0.05/sec for 720p w/ audio. ~$4.20/minute generated.

Deprecated & legacy models

Model	Status	Migrate to
Grok 1 / 1.5	Retired	Grok 4.3 / 4.1 Fast
Grok 2 / 2 mini	Legacy	Grok 4.20 / 4.1 Fast
Grok 3 / 3 mini	Legacy	Grok 4.3 / 4.1 Fast
Grok 4	Active but superseded	Grok 4.3
Aurora (image)	Superseded by Grok Imagine	Grok Imagine

🎈 ELI5

The Grok app is where you chat with Grok. You can use it on grok.com (standalone) or X.com (right inside Twitter). Grok's secret superpower is that it sees X in real time — ask "what are people saying about [thing] right now?" and it knows. No other AI does this as well because no other AI owns X.

You can also talk to Grok with voice mode and even create AI Companions (animated characters) that you chat with.

Setup & surfaces

Pick where to chat: grok.com (clean standalone), x.com sidebar (chat while browsing X), iOS app, or Android app.
Sign in with X account, email, or Apple/Google.
Pick your model from the picker — Grok 4.3 / 4.20 / 4.1 Fast. Free users get limited access; subscriptions unlock higher tiers and bigger limits.
Drop files / images / X posts. Grok reads attached PDFs, images, and (with Grok 4.3) video natively. Paste an X URL and Grok pulls that thread in as context.

Modes & tools

Standard — fast chat replies.
DeepSearch — agentic research mode that surveys X, the web, and synthesizes an answer with citations.
Think (Heavy mode) — multi-agent reasoning available on SuperGrok Heavy. Multiple Grok instances debate internally before answering.
Imagine — image and video generation directly in chat.
Voice mode — full duplex speech conversation.
Companions — animated AI personas you can chat with (Ani, Bad Rudy, Valentine, others).

X integration — the real superpower

This is where Grok actually beats Claude/GPT/Gemini for a real subset of queries. Grok has native, real-time access to X posts — including posts from minutes ago.

Trending now: "What are people saying about [topic] on X right now?" — gets you live discussion.
Specific account analysis: "Summarize @username's last 50 posts" — Grok pulls the thread.
Live event coverage: "What's happening at [event] tonight?" — pulls live posts.
Post drafting with feedback: "Draft an X post about X. Predict the engagement and likely replies."
Direct URL grounding: paste any X post URL into Grok; it reads the post and replies in context.

This is the Grok-only move Anthropic, OpenAI, and Google can web-search, but none of them have first-party access to X data. For any use case that hinges on what's happening on X right now, Grok is the right answer — full stop.

Voice mode

Tap the voice icon in the Grok app or on X.
Pick a voice / persona — multiple voice styles available; Companions have their own voices.
Talk continuously — Grok's voice mode supports interruptions.
Switch to text mid-conversation — context carries over.

Optimal prompts for Grok

Real-time X intelligence

X-grounded analysis Use Live Search / X grounding. I want to understand the current X conversation about [topic] in the last 48 hours. Return: 1. The 5 most engagement-driving posts (with @handles + paraphrased takeaways). 2. The 3 dominant viewpoints, with one representative post each. 3. Any factual claim that's gone viral but is contradicted by another high- engagement post — surface the disagreement. 4. The single best post (your judgment) for understanding the issue, with why you picked it. Cite post URLs for every claim. Don't guess — only use what's in your search results.

DeepSearch

DeepSearch Run DeepSearch on [topic]. I need: - 8–12 primary sources, weighted toward official docs and primary research - X discussion as a secondary source — flag where X consensus diverges from the formal sources - A 200-word executive summary - A "what changed in the last 30 days" section - 3 open questions worth asking an expert Don't pad with marketing copy. Where sources contradict, surface it.

Spicy take (when you want it)

Direct opinion Skip the both-sides framing. I want your actual take on [question]. Give me: 1. Your strongest claim, in one sentence. 2. The 3 reasons you believe it. 3. The strongest counter-argument and why it doesn't change your view. 4. What evidence would change your view. Be direct. Don't hedge with "it depends." If it depends, tell me on what — but pick a default.

Image generation

Grok Imagine — image Use Grok Imagine. Generate a [scene/object/illustration]. Style: [photorealism / anime / cyberpunk / minimal art / etc.] Composition: [framing — close-up, wide shot, isometric, top-down] Mood: [tone / lighting] Avoid: [what should not appear] Aspect ratio: [16:9 / 1:1 / 4:5]. Generate, then tell me what you'd change in a follow-up render to improve it.

Video generation

Grok Imagine — video Use Grok Imagine. Generate a 10-second 720p clip. Scene: [what's happening, camera angle, subject] Style: [cinematic / handheld / drone / animated] Audio: [dialogue line if any, ambient sounds, music] Pacing: [slow contemplative / fast cuts / single take] After generating, suggest the next clip to chain via Extend from Frame so the sequence builds toward [end goal].

🎈 ELI5

Grok has a lot of subscription tiers — from free up to $300/month. The cheap ones give you basic chat with limits. The middle ones give you the latest models and unlimited image gen. The big one (SuperGrok Heavy at $300/month) gives you "Heavy" mode, where multiple Grok instances think together before answering — for the very hardest problems.

Or, if you use attap.ai, you skip the Grok subscriptions and pay per call in credits — much more flexible if you're trying multiple models.

Subscription tiers

Plan	Price	Best for	Notes
Free	$0	Trying Grok	Limited messages, smaller model.
X Premium	$8/mo	Casual X users + light Grok	Bundled with X subscription.
SuperGrok Lite	$10/mo	Cheap creator pack	Basic image/video gen at 480p, 6-sec clips.
SuperGrok	$30/mo	Heavy daily users	Full Grok 4 access (128K context), ~100 prompts every 2 hours, unlimited Imagine images, daily video renders.
X Premium+	$40/mo	X power users	Combines X premium features with extended Grok access.
SuperGrok Heavy Heavy	$300/mo	Researchers, power users	Multi-agent Heavy mode; full access to Grok 4.3; biggest contexts; first to receive new releases.

SuperGrok Heavy — what makes it different

Heavy mode is the only consumer plan that runs multi-agent reasoning — multiple Grok instances tackle a problem in parallel, debate, and converge on an answer. Result: meaningfully better quality on the hardest problems, but slower and pricier.

Best for: research synthesis, hard reasoning, complex agentic tasks.
First to receive new model releases (e.g. Grok 4.3 GA was Heavy-only initially).
Comparable in spirit to ChatGPT Pro's Pro mode or Gemini 3.1 Pro's Deep Think.

Using Grok via attap.ai instead

If you're already an attap.ai user, you can skip xAI subscriptions and pay per call in credits. Currently:

Model on attap	Credits per call	Equivalent use
Grok 4.1 Fast	1 credit	Volume work, classification, drafts
Grok 4.3	6 credits	Hard reasoning, deep research
Grok Imagine — Image	10 credits	Single generated image
Grok Imagine — Video	100 credits	Short generated clip with audio

When attap is the better deal If you use Grok occasionally alongside Claude / GPT / Gemini, attap's per-call credit model is more flexible than committing to a $30/mo SuperGrok subscription. If you use Grok intensively every day (especially Heavy mode), the direct subscription is cheaper.

🎈 ELI5

The xAI API at docs.x.ai works almost exactly like the OpenAI API — same Python and Node SDK shape — so if you've used OpenAI before, you can plug Grok in by changing one URL. The unique features are Live Search (Grok grounds answers in live X data and the web) and the Imagine API (image and video generation).

Account & API keys

Sign in at console.x.ai. Free trial credits available; usage tiers unlock with spend.
Create an API key — use named, scoped keys per project.
Set a usage limit in the console before you start.
The base URL is https://api.x.ai/v1 — OpenAI-compatible.

First API call

Python — OpenAI SDK pointed at xAI pip install openai export XAI_API_KEY="xai-..." python - <<'PY' from openai import OpenAI client = OpenAI( api_key=os.environ["XAI_API_KEY"], base_url="https://api.x.ai/v1", ) resp = client.chat.completions.create( model="grok-4.3", messages=[{"role": "user", "content": "Write a haiku about debugging at 2am."}], ) print(resp.choices[0].message.content) PY

Node — OpenAI SDK npm install openai export XAI_API_KEY="xai-..." node -e " import('openai').then(async ({ default: OpenAI }) => { const client = new OpenAI({ apiKey: process.env.XAI_API_KEY, baseURL: 'https://api.x.ai/v1', }); const resp = await client.chat.completions.create({ model: 'grok-4.3', messages: [{ role: 'user', content: 'Write a haiku about debugging at 2am.' }], }); console.log(resp.choices[0].message.content); }); "

Live Search — grounding in X & web

xAI's signature API capability. Pass search_parameters to ground the answer in live X posts, the web, news, or specific X accounts.

Live Search resp = client.chat.completions.create( model="grok-4.3", messages=[ {"role": "user", "content": "What's the current sentiment on X about [topic]?"} ], extra_body={ "search_parameters": { "mode": "on", # "auto" or "on" "sources": [ {"type": "x"}, {"type": "web"}, ], "max_search_results": 20, "from_date": "2026-05-01", "to_date": "2026-05-04", } }, ) print(resp.choices[0].message.content) # Citations are returned in resp.choices[0].message.citations

Tool use & structured output

Grok supports OpenAI-compatible function calling and structured JSON output via response_format.

Function calling tools = [{ "type": "function", "function": { "name": "get_weather", "description": "Get current weather for a city.", "parameters": { "type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"], }, }, }] resp = client.chat.completions.create( model="grok-4.3", messages=[{"role": "user", "content": "What's the weather in Lagos?"}], tools=tools, tool_choice="auto", ) # resp.choices[0].message.tool_calls contains the function call.

Structured JSON output resp = client.chat.completions.create( model="grok-4.1-fast", messages=[{"role": "user", "content": f"Extract structured fields from this resume:\n\n{resume_text}"}], response_format={ "type": "json_schema", "json_schema": { "name": "Resume", "schema": { "type": "object", "properties": { "name": {"type": "string"}, "email": {"type": "string"}, "years_experience": {"type": "integer"}, "skills": {"type": "array", "items": {"type": "string"}}, }, "required": ["name", "skills"], }, }, }, )

Imagine API — image and video generation

Image generation POST https://api.x.ai/v1/images/generations { "model": "grok-imagine", "prompt": "A futuristic Tokyo street at night, cyberpunk style, neon reflections on wet pavement, ultrawide shot", "n": 1, "size": "1024x1024", "response_format": "url" }

Video generation (10-sec, 720p, w/ audio) POST https://api.x.ai/v1/videos/generations { "model": "grok-imagine-video", "prompt": "A barista carefully pouring latte art in a cozy cafe, soft morning light, ambient cafe sounds, brief 'thank you' from customer at the end", "duration_seconds": 10, "resolution": "720p", "audio": true, "aspect_ratio": "16:9" } # Cost: ~$0.50 for this 10-sec 720p clip # For longer sequences, chain via "extend_from_frame" referencing the prior clip's last frame.

Optimal prompts for the xAI API

Cost-aware routing (3-tier)

Three-tier triage # Tier 1 — Grok 4.1 Fast for triage / classification triage = client.chat.completions.create( model="grok-4.1-fast", messages=[{"role": "user", "content": f"Classify intent: 'simple' / 'medium' / 'hard' for:\n{query}"}], ).choices[0].message.content.strip() # Tier 2 — Grok 4.20 for medium tasks (fast frontier) # Tier 3 — Grok 4.3 for hard reasoning (always-on reasoning) model_map = {"simple": "grok-4.1-fast", "medium": "grok-4.20", "hard": "grok-4.3"} final = client.chat.completions.create( model=model_map.get(triage, "grok-4.20"), messages=[{"role": "user", "content": query}], )

X-grounded research at scale

X grounding production resp = client.chat.completions.create( model="grok-4.3", messages=[ {"role": "system", "content": "You are a market intelligence analyst. Always cite X post URLs " "for sentiment claims. Distinguish high-engagement from low-engagement " "posts. If signal is sparse, say so — never extrapolate."}, {"role": "user", "content": f"Synthesize the X discourse on {topic}."} ], extra_body={"search_parameters": {"mode": "on", "sources": [{"type": "x"}]}}, )

Long-context analysis (1M / 2M)

Big input resp = client.chat.completions.create( model="grok-4.20", # 2M context, faster than 4.3 for long inputs messages=[ {"role": "system", "content": "Read the entire bundle. Identify every place [pattern] appears. " "Report file:line for each. Don't summarize — be exhaustive."}, {"role": "user", "content": full_repo_or_corpus_bundle}, ], )

🎈 ELI5

A prompt is directions for Grok. Same as every model — tell it WHO it is, WHAT you want, WHO reads it, HOW to format. The Grok-specific twist: be direct. Grok responds to opinionated prompts. If you ask wishy-washy, you get wishy-washy.

Use-case prompt library

Copy-paste-ready prompts for the most common goals. Edit the bracketed parts to fit your situation. Click Copy on any card.

Writing & email

Email reply Reply to this email. Tone: warm, concise, no apology-for-apologies'-sake. Length: 4–6 sentences. End with one clear next step. If the thread mentions a specific commitment I made, restate it back so they know I read carefully. Output only the reply text — no preamble, no signature. Email: """ [paste email] """

Slack / DM rewrite Rewrite this Slack message: - 30% shorter - friendlier tone - keep the load-bearing facts - no apology-for-apologies'-sake Output only the rewritten version, no commentary. Original: """ [paste] """

Difficult message — 3 versions Help me write a [decline / push-back / disagree] message. Recipient: [who they are, relationship] What I need to communicate: [the message] Constraints: [keep relationship intact / be firm / be diplomatic] Draft three versions: gentle, neutral, firm. End each with one open question that invites them to engage further.

Meeting follow-up Turn these meeting notes into: 1. A 3-bullet summary I can paste into Slack 2. An action items list (owner — action — by when) 3. One open question that wasn't resolved Notes: """ [paste] """

LinkedIn post Draft a LinkedIn post about [topic]. Tone: confident, not braggy. Hook in the first line. 5–8 short paragraphs. End with a question that invites comments. No emojis, no hashtags. Avoid "thrilled to share" and "humbled".

Editing & feedback

Sharp editor Act as a sharp editor. Don't rewrite — diagnose and prescribe. For the draft below, return: 1. The single biggest weakness, in one sentence. 2. Three specific edits with before → after. 3. One sentence I should consider cutting entirely, and why. Audience: [who reads this]. Tone target: [tone]. Draft: """ [paste] """

Tighten by N% Cut this text by 40% without losing meaning. Keep the voice. Return only the tightened version, no commentary. """ [paste] """

Stress-test my argument Find the holes in this argument as a tough but fair critic would. Return: 1. The strongest counter-argument I haven't addressed. 2. The weakest claim I'm making, and why it's weak. 3. What evidence would change your mind. 4. One thing I should NOT change — what I'm getting right. Argument: """ [paste] """

Learning & research

Teach me [topic] Teach me [topic]. I already know [X, Y]. I don't know [Z]. Structure: 1. The one-sentence elevator definition. 2. The mental model (an analogy that maps to something I already know). 3. Three concrete examples of increasing complexity. 4. The two most common misconceptions and why they're wrong. 5. A 5-question self-test (no answers — I'll check myself). Skip filler. Be willing to be wrong if I push back.

Concept to analogy Find me the best analogy for explaining [concept] to [audience]. Return 3 candidates ranging from safe to creative. For each, name what the analogy gets right AND what it breaks. Pick your favorite and defend the choice.

Compare options Compare [option A] vs [option B] for [use case]. Return a 1-page brief: - The decision in one sentence - Top 3 dimensions where they differ - Best fit for each (when would you pick A, when B) - One scenario where neither is right and what is Skip the generic "depends on your needs" framing. Pick.

Decision-making

1-page decision memo Help me decide between [option A] and [option B]. Context: [situation, who's affected, deadline]. Constraints: [budget, time, reversibility]. What I value most: [criteria, in order]. Produce a 1-page decision memo with: - Recommendation in the first line - The 3 reasons that drove it - The strongest counter-argument and your response - What would have to be true for the other option to win

Pre-mortem Run a pre-mortem on this plan: [paste plan]. Imagine it's [N months] from now and the project failed. Write the failure post-mortem. List the top 5 root causes in order of likelihood, and what we could do today to prevent each. Be specific. "Insufficient communication" is a non-answer; "the design team and engineering used different definitions of done" is useful.

Devil's advocate Take the position that [my proposal] is wrong. Be the strongest advocate for the opposite view. Give me: 1. Three reasons my proposal is the wrong call. 2. A version of the opposite plan that you'd actually defend. 3. One question that, if answered honestly, would tell us who's right. Don't soften — I want the spicy version. I'll evaluate it on the merits.

Brainstorming

Diverge then converge Help me brainstorm [topic]. Round 1 — diverge: 12 ideas covering safe, weird, ambitious, contrarian. Round 2 — converge: pick the top 3 by [criterion] and explain trade-offs. Round 3 — sharpen: turn the winner into a one-paragraph pitch. Don't hedge. I want opinions.

Opposite of obvious What's the obvious answer to [question]? State it in one sentence. Now: what if the OPPOSITE is true? Make the strongest case for the counterintuitive answer. Don't strawman; defend it as if you believed it. End with: which one do you actually think is right, and why.

Coding

Bug repro & fix Bug: [describe the bug + what should happen]. Reproduce with a failing test first, then fix the smallest amount of code that turns it green. Constraints: - Don't refactor unrelated code. - No new dependencies. - Test goes in [path] next to similar cases. When done: show me the diff, the test output, and one sentence on the root cause.

Code review of a snippet Review this code as a senior reviewer would. Flag: 1. Anything broken or buggy 2. Anything that looks like dead code or leftover debug 3. Tests that test the implementation rather than behavior 4. Risky assumptions that aren't documented 5. One thing that needs a comment explaining the WHY Skip stylistic nits the formatter would catch. Code: """ [paste] """

Refactor for [property] Refactor [file/function] to [be smaller / be testable / remove duplication]. Rules: - No behavior changes — every existing test must still pass without edits. - Keep the public API identical. - Split by responsibility, not by line count. Start by reading the code and tests, then propose the split before editing.

Generate test cases For the function below, generate test cases covering: - Happy path - Edge cases (empty input, null, max size, off-by-one) - Error cases (what should throw, what should return safely) Use [framework]. Output the tests only, no commentary. Function: """ [paste] """

Regex / SQL from English Generate a [regex / SQL query] that [describes what it should do]. Then explain it line-by-line in plain English. Then give me 3 test cases — 2 that should match/return rows and 1 that shouldn't.

Data & analysis

Clean a CSV Attached CSV is messy. Clean it for analysis. Steps: 1. Detect and list every quality issue (nulls, mixed types, dupes, weird dates). 2. Propose a fix for each, with the assumption you're making. 3. Apply the fixes; return a clean version. 4. End with a "trust this output if…" caveat list. Never silently drop rows — quarantine them in a separate sheet.

Extract structured fields Extract structured fields from this text. Required fields: - [field 1] (type) - [field 2] (type) Output as JSON. For any field that's not stated in the text, return null. Never guess. Text: """ [paste] """

Find anomalies Look at this dataset. Identify the 5 most surprising rows or values. For each, explain why it's surprising and what plausible explanations exist (data error, real outlier, definition issue). Don't moralize about which explanation is right — just lay them out.

Documents

TL;DR a long doc TL;DR this document. Return: 1. The thesis in one sentence. 2. The 5 claims it actually defends. 3. Any claim that's asserted but not supported. 4. The single most surprising fact. 5. Three questions worth asking the author. If something is unclear, say "not stated" — never guess.

Find every claim Pull every factual claim from this document into a numbered list. For each, note: - The page/section it appears in - Whether the doc supports it (with evidence) or just asserts it - Confidence level (high / medium / low) Useful for fact-checking before sharing.

Outline → draft Turn this outline into a [first draft / one-pager / blog post]. Tone: [tone]. Length: [target]. Audience: [who]. Don't add new claims I didn't include in the outline. If a section is underspecified, leave a note like [needs example here] instead of inventing.

Creative & personal

Name brainstorm Brainstorm 12 names for [thing]. Cover these vibes: - 3 plain/descriptive (does what it says) - 3 evocative/metaphorical - 3 weird/punchy/short - 3 contrarian (intentionally counterintuitive) For each: a one-line rationale and one risk. Pick your top 3 and rank them.

Cold outreach email Draft a cold outreach email to [recipient role at company type]. Goal: [what I want them to do] What I bring: [my context, what's interesting] Constraint: ≤120 words, no buzzwords ("synergy," "leverage"), one specific reference to their work. End with one easy ask (e.g., 15-min call) — not "let me know if interested".

Travel itinerary Plan a [N-day] trip to [destination] for [N people, ages, interests]. Constraints: [budget, mobility, dietary, packed-vs-slow style]. Day-by-day: - Morning / afternoon / evening for each day - One must-do, one easy fallback - Estimated cost per day - One thing locals do that tourists miss Skip generic recommendations everyone already knows.

Interactive prompt builder

What's your goal? Who is Grok (role)? Who's the audience? Topic / subject / paste-in Output format Constraints / things to avoid

Fill in the fields above…

Patterns library

The "X-grounded" pattern (Grok-only superpower)

X-grounded Use Live Search on X for the last [N] hours about [topic]. Don't summarize the whole conversation — surface only: 1. The single most-engaged post 2. The 3 dominant viewpoints (one representative post per view) 3. Any factual disagreement between high-engagement posts 4. What's notably absent (any obvious angle that nobody's discussing) Cite post URLs for every claim. If signal is weak, say "low signal" — don't pad.

The "direct opinion" pattern

Direct opinion Skip both-sides framing. I want your actual take on [question]. 1. Your strongest claim, in one sentence. 2. The 3 reasons you believe it. 3. The strongest counter-argument and why it doesn't change your view. 4. What evidence would change your view. Be direct. Don't hedge with "it depends" — pick a default.

The "stop and ask" pattern

Stop & ask Before doing anything, list the 3 most important things you'd need to know to do this well that I haven't told you. Number them. Then propose the answer you'd assume by default for each. Don't start the work until I confirm.

The "two passes" pattern

Two passes Pass 1: draft [the deliverable]. Don't optimize — just get it out. Pass 2: critique your own draft as a tough editor would. List the 3 weakest points specifically. Pass 3: rewrite, fixing those 3 things. Show only the final.

Anti-patterns

Anti-pattern	Do this instead
Asking Grok for "balanced" answers and getting a hedge-fest	Either ask explicitly: "balanced — name the strongest case for each side" — or skip the qualifier entirely and ask Grok to pick.
Using Grok 4.3 for every task	Triage: 4.1 Fast for simple, 4.20 for medium-fast, 4.3 for hard. Latency & cost stay sane.
Forgetting Live Search exists	For anything time-sensitive or X-relevant, turn on Live Search — Grok's biggest moat is wasted otherwise.
Treating Grok like ChatGPT for non-current tasks	Grok's edge is real-time + opinionated. For pure reasoning with no time component, Claude Opus or Gemini 3.1 Pro often do better.
Asking for typography-heavy images	Imagen 4 / gpt-image-2 currently render text better in images. Use Grok Imagine for non-text-heavy visuals.

Universal rescue prompts

Diagnose & reset Pause. Tell me in 3 bullets why your last response missed what I wanted. Then propose a tighter prompt I could give you to fix it. Wait — don't try again yet.

Cut the fluff Redo your last response with no preamble, no caveats, no closing summary. Just the substance.

Show your work Walk me through your reasoning before the answer this time. I want to see the trade-offs you considered.

Spicier please That answer was too soft. Give me your actual opinion — you can be wrong, that's fine. Pick a side and defend it. I'll push back if I disagree.