Pack 01 · Bundle
Build with AI — Websites, Apps, Code & AI Media
Idea → live app this weekend. Build with AI is the system for shipping things people can actually open: websites and landing pages, app architecture and production code, motion and animation, and the AI image-to-video pipeline — every step driven by a filled-in prompt rather than a blank template.
One-time €18. No subscription, ever. 54 ready-to-run prompts + 10 installable skills. Instant download, unlimited use, yours to keep — backed by a 30-day money-back guarantee.
What’s inside Build with AI
- Websites, landing pages & full builds — go from blank page to a live, styled site with copy, sections and calls to action already in place.
- App architecture, components & code — scope the data model, generate components and get production-ready code instead of disconnected snippets.
- Scroll animation, parallax & motion — the movement layer that makes a page feel designed rather than templated.
- AI image-to-video (Higgsfield) pipeline — turn stills into motion for heroes, ads and product shots.
- AI engineering: RAG, agents & evals — decision memos, agent scoping and evaluation harnesses for people shipping real LLM features.
What you actually download
A complete folder: the full system guide (PDF), All Prompts (PDF) and drop-in Prompts (Markdown), All Skills (PDF) and drop-in Skills (Markdown), the How It Works system map, a Quick Start & Cheat Sheet, Tips, Hooks & Tactics, a worked end-to-end project walkthrough, a Bonus folder, a README and a cover.
Is €18 worth it for one system?
At €18 for 54 prompts and 10 skills, a single system is about €0.28 per item. If you want more than one area covered, the maths changes fast: The Vault is the lowest cost per item: €30 for 323 prompts and 81 installable skills works out to roughly €0.07 per item, versus €18 for a single system. Buying the three systems separately costs €54 — the Vault is €30 and adds the UGC & Lip-Sync Kit, the plan-mode mega-prompts and the brand kit on top of them.
What a Build with AI prompt actually looks like
These are two real, unedited prompts from inside the system — not a sample, not a summary. Note that nothing is left as a blank [bracket] to guess at: the role is specific, the constraints are enforced, and the output format is pinned down. That is the whole difference between a prompt list and a system. Copy either one straight into ChatGPT, Claude or Gemini.
Role: Senior AI engineer who has shipped both RAG-based and fine-tuned LLM products at production scale. You believe most teams pick the wrong path because they ask the wrong question first.
Context: A team is deciding between RAG, fine-tuning, prompt engineering, or a mix for: [DESCRIBE_USE_CASE]. Constraints — corpus [N docs], freshness [hourly/daily/static], budget per query [$X], expected QPS [Y], data sensitivity [public/PII/regulated].
Task: Produce a decision memo — 1) quality requirement (recent facts → RAG, domain vocabulary → fine-tune, structured reasoning → prompts); 2) cost per query per path; 3) latency (cold start, warm, p95); 4) operational complexity; 5) risk (hallucination tolerance, regulatory); 6) ONE recommendation + the test that proves it.
Constraints: Show your math ("$0.04 vs $0.12 per query, 100k/day = $2.9k vs $8.7k/mo"). Name the runner-up and what would tip it. Refuse to recommend without a success metric.
Output: Markdown memo, 6 sections, ≤900 words, exec-ready.
Role: AI engineer who has built evals suites that caught 30+ production regressions before they shipped. Vibes-based "this looks fine" testing is the #1 cause of LLM products that silently degrade.
Context: A team wants to ship an LLM feature in [DOMAIN] with no structured evals. Current "testing" is the PM eyeballing 5 outputs in a sprint review.
Task: Design a complete evals harness — 1) ground-truth construction; 2) eval dimensions, each operationalized ("% of outputs citing ≥1 verifiable source", not "quality"); 3) per-dimension grader (string match / LLM-as-judge / human) + calibration; 4) CI gating (per PR, nightly, pre-release); 5) drift detection on model upgrades; 6) cost per run; 7) regression budget per dimension.
Constraints: LLM-as-judge graders MUST be calibrated to humans (κ ≥ 0.7 or invalid). Every threshold justified. Include a "what we won't test" list.
Output: Markdown spec ≤1200 words + a sample eval YAML.
Guides that use this system
- How to build a website with AI — the full walkthrough.
- How to build an app with AI — no-code and code paths.
- AI prompts for landing pages — high-converting page copy and structure.
Frequently asked questions
What do I get in Build with AI?
54 ready-to-run prompts and 10 installable AI skills covering websites and landing pages, app architecture and code, scroll animation and motion, the AI image-to-video pipeline, and AI engineering (RAG, agents, evals), plus the system guide, cheat sheet, system map and a worked project.
Do I need to know how to code?
No. The prompts are written in plain language and every one ships with a filled-in example rather than blank brackets. If you can chat with an AI you can use it, and you can ask the AI to explain any code it writes.
Which AI tools does it work with?
It is model-agnostic. The prompts and skills work in ChatGPT, Claude, Gemini and most assistants, with no special setup.
Is Build with AI a subscription?
No. It is a one-time 18 euro payment with instant download, unlimited use and lifetime access, backed by a 30-day money-back guarantee.
Should I buy this or The Vault?
Buy Build with AI if you only need websites, apps and code. If you also want content, sales or the UGC video kit, The Vault is 30 euro for all three systems plus the extras, which is cheaper than buying two systems separately.
Hazel is an educational toolkit; the AI tools mentioned are third-party and not affiliated with Hazel. Your results depend on your own effort.
Compare the bundles
Build covers what you make. Create & Grow covers content and audience, Operate & Scale covers selling and running the business, and The Vault is all three plus the extras for €30.