Every scene is a path you can run: which skills to use, how to install, how to prompt, and what traps to avoid.
Pick a scene to start
The clone is fluent in Dalio. That is the product, and that is the ceiling. His book argues the opposite of what the clone does: a principle only fires if you wrote it. So don't ask an AI to recite his — make it interview you until the borrowed ones fall out.
The book is static, the AI is not. The book answers "which advice is worth doing"; a skill answers "does this one hold for the mess I'm in right now". Pack it, and the next time you're stuck you just ask — but don't expect it to decide for you.
Personal efficiency is not "busier" — it's offloading fragments to a reliable system: capture → consolidate → plan → review. Let AI do note cleanup, knowledge merge, planning breakdown; you only judge "keep or not, what's next", leaving your head for what truly needs thinking.
Deep Research compresses hours-to-days analysis into tens of minutes: you give the question, the agent searches in steps, cross-checks, and returns a sourced structured report. Community method stresses: break the question fine first, cross-verify from multiple sources, every claim traceable — fast
The biggest lever in social content is the topic — communities widely agree "topic decides 60%+ of a post's success". Have AI mine last-7-day hits for direction first, then run topic → script → image/video → multi-platform → review. Far better than staring at a blank page.
A workflow that actually runs is not "one bot sells for you". It splits sourcing → copy → video → listing → review into repeatable steps, AI drafts each, you make the key calls. Community consensus: build the minimum viable chain first, then automate deeper.
AI writes fast, but the quality gate matters more. Community consensus in three: reverse TDD (AI writes tests, then minimal impl), zero-warning gate (`--max-warnings=0`, warnings silently become the new baseline), review as a gate (every change through the same gate). Make AI output verifiable and g
Ecommerce pain is "copy never ends, video never shoots, listing is all manual". The point of AI: give your product and angle, let it draft copy/video/listing, you only check "does it sound like buyers talk", not stare at a blank page.
Fastest way to regret AI code: "it writes fast but you dare not merge". Core: give it coding standards, then review and refactor; you make the merge call, not hand over the keyboard.
Paper pain is "read for ages, still don't know what it proved". The point of AI: give the source, let it split problem-method-conclusion-limit, you judge usability — not let it declare "this is great" for you.
Video's pain is "shoot for hours, edit longer". The point of AI: give script and angle, let it do shot structure and captions; you only check "does it sound like me", not stare at an empty timeline.
Browser-automation pain is "brittle scripts, break on any change". The point of AI: give a clear goal and a done-criteria, let it walk and self-verify, not hardcode XPath and maintain forever.
The worst tech blog is "copy the docs" or "fluff padding". The point of AI: give it your real experience/pitfalls (the unique value), let it handle structure, examples, polish — not invent stuff you never did.
The pain of PRD isn't "wrong format" — it's "goal and acceptance unclear, dev guesses". The point: give background, target user, problem; let AI output "feature list + acceptance + edges", not decide "why build" for you.
AI SQL's scare is "runs but wrong口径" — looks right, aggregation dimension wrong. The point: give schema and clear口径 (by day / by user, dedup?, exclude refunds), have it explain then write, not just "count sales".
The worst cold email is "mass-send feel" — binned as spam at a glance. The point: give the recipient profile and one real value, make it point-to-point like handwritten, not templated blast. Reply rate comes from relevance, not length.
The worst text-to-image is "one-line prompt, pure luck". The point: set style ref + subject + composition + ratio as structured prompt, then iterate small — not throw "draw something nice" and hope.
The pain of dashboards isn't "can't draw charts" — it's "metrics undefined, charts piled, nobody looks". The point: define the north-star metric and the question first, then let AI pick chart types and write queries, not pile fancy charts upfront.
The biggest value of testing isn't "high coverage" — it's "catching the paths that actually break". The point of AI-generated tests: give it the implemented code or a clear behavior contract, then have it add edge cases. Don't let it invent the thing under test; a test that passes on a fake object p
Machine translation's scare is "fluent but distorted" — wrong terms, wrong tone, stiff local phrasing. The point: give a glossary and target-reader tone, have it "translate + keep proper nouns", then you verify key sentences — not trust the whole blindly.
The pain of illustration isn't "can't draw" — it's "inconsistent style, blurry when scaled, can't edit". The point of AI vector illustration: set style and size first, want vector not bitmap, so it scales sharp and the source stays editable.
The pain of podcast/talking-head isn't "no voice" — it's "script reads like a paper, voice sounds robotic, nobody finishes". The point of AI: write the script as spoken human language, pick a natural voice, then decide on visuals — don't pile effects upfront.
The pain of weekly reports isn't "can't write" — it's "turning a log into outcomes". The point: give AI the raw material (what you did, what's stuck), have it close into "progress / risk / next", not invent content for you.
The worst brainstorming is "diverge and stop". The point of AI help: first force enough candidates (quantity before quality), then converge on 3 doable ones by "feasibility / impact" — not stop at "many ideas".