← reposition.dev

Blog

Original research and practical guides on AI workflows, prompting, and strategy. Based on real data.

Watch How You Actually Use AI: A Walkthrough of Claude Code /insights
The /insights command in Claude Code reads your whole history and tells you how you actually work, then suggests improvements to your CLAUDE.md. It turns iteration from a manual chore into a...
One Skeleton, Many Reviews: Scale One Prompt Across Code, Proposals, and Reports
The same prompt skeleton can review code Monday, proposals Tuesday, and business reports Wednesday. Different standards, same shape. One judgment system becomes many, without starting from scratch.
Why Your AI Output Drifts (and the 15-Minute Fix Every 3 Weeks)
Your standards shift, the work changes, and AI output drifts without you noticing until it misses something obvious. Fifteen minutes every three weeks keeps it sharp. Here is the cadence that...
Build a Judgment System: Make AI Review Work the Way You Do
The hardest work to automate is the review only you can sign off on. A judgment system encodes your standards into a prompt so AI evaluates work the way you would, worst problems first, with a...
From Asker to Builder: The Shift That Unlocks Real AI Leverage
Most people use AI as an asker, typing one request at a time. The leverage move is to become a builder, turning AI into systems that run your judgment. The proof was built in under 15 minutes.
Context Is Not Background: The Prompt Change That Makes AI Think Like You
Everyone fills the Context slot of their prompt with generic background. The unlock is treating Context as your standards, the checklist in your head. That list is your judgment, encoded.
You Are the Bottleneck: Why AI Left You the Hard 20%
AI sped up the easy 80 percent of your work. The hard 20 percent, the reviews and sign-offs only you can make, still has your name on it. Here is why you are the bottleneck, and the one move that...
Why the Loom Video Closes Deals (and Long Emails Don't)
People do not read long outreach emails. They click a twenty-second screen recording. Here is why the Loom video is the asset that closes AI service deals.
One Person, Ten Clients: The Leverage Math of an AI Service Business
Why one person running an AI content system can serve ten clients in two days a week, and what the revenue math actually looks like once delivery compresses.
Setup Fee Plus Retainer: The Only Pricing Model That Works for AI Services
A one-time price is either too expensive for the client or too much unpaid work for you. Setup fee plus monthly retainer is the model that actually works.
AI Is Bad at Generating Options: Why You Should Feed It Your Own
Most people ask AI 'what should I do?' and get generic advice. The fix is simple: give AI the options you have already considered. It is excellent at evaluating. Terrible at inventing.
Red Flags and Green Flags: How to Spot a Client Worth Pitching
A two-minute framework for scoring prospects. Red flags mean opportunity. Green flags mean they can pay. The sweet spot is a business with both.
Revenue Over Client Count: Why Your Goals Should Be in Dollars, Not Numbers
Most founders and freelancers set goals in client numbers. That is a mistake. Goals in revenue force better strategy, better pricing, and better decisions. Here is why.
Sell the Truth, Not the Brochure: Why a Client's Reviews Are Your Best Sales Hook
When a business claims to be the best but its reviews tell a different story, the gap between the two is your entire pitch. Here is how to use it.
The Pre-Mortem: How to Stress-Test Any Idea with AI Before You Build It
Most people fall in love with their idea and then look for evidence it will work. The pre-mortem does the opposite: it tries to kill the idea first. Here is how to use AI to stress-test anything...
The Priya Case Study: How One Agency Owner Used AI to Decide Which Clients to Drop
A marketing agency owner at 60 hours a week, earning $5500/month, unsure whether to drop her three worst clients. Here is how the 4-part strategic AI framework gave her clarity.
Which AI Model for Which Task: A Practical Guide
Opus for strategy. Sonnet for daily work. Haiku for simple tasks. Fable for complex reasoning. Here is when to use each model and what effort level to set.
Build Something That Earns: Why Delivering a Service Beats Building a Product
Three ways to earn with AI: build a product, get a better job, or deliver a service. The service path is the fastest to real money, and AI compresses the work.
The AI Content System: A Month of Content in 10 Minutes
A repeatable AI workflow that produces a full month of on-brand content, 12 posts, 4 newsletters, 8 blog outlines, for a local business in about 10 minutes.
Operational AI vs Strategic AI: Same Tool, Different Question
Most people use AI to do work faster. That saves minutes. A smaller group uses it to figure out what work is worth doing. That saves weeks. Here is the difference and when to use each.
The 4-Part Framework for Any Decision: Situation, Options, Criteria, Blind Spots
A repeatable framework for using AI to make better decisions. Four parts that work for any situation: career, business, product, or life. No theory, just structure.
Build AI Workflows for Yourself Before Building for Others
The first person you should automate is yourself. Building AI workflows for other people without doing it first leads to the wrong solution.
Role-Based Prompting: Why Your AI Output Depends on Who You Say Is Reading It
The same prompt produces different output depending on who you tell the AI is reading it. Here is why audience awareness changes everything about your workflow output.
How to Build an AI Video Production Pipeline: From Green Screen to Auto-Post
A 6-step pipeline that takes a talking-head video, generates AI backgrounds, assembles the final cut, and posts across social media. Built with Claude, Remotion, and FAL.AI.
The Predictability Principle: Why Every Repeating Task Is Automatable
Most professionals overlook automatable tasks because they confuse content variation with structural unpredictability. Here's how to identify hidden workflow candidates in your week.
Prompt Templates for Code Reviews: A Practical Guide
Copy these prompt templates for code reviews, PR summaries, and architecture decisions. Structured for Claude, ChatGPT, or any LLM.
Using Claude Sonnet for Quick Tasks: When to Skip Extended Thinking
Extended thinking is powerful but expensive and slow. Here is when to use Claude Sonnet with effort low for faster, cheaper results without quality loss.
The 3-Step System That Keeps People Using AI Workflows (When Most Quit)
Most people who try AI workflows quit within a week. The reason isn't complexity or lack of time. It's that they never built a system. Here's a 3-step framework that makes AI workflows stick.
From 20 Minutes to 10 Hours: How One PM Scaled His AI Workflows
A case study from our workshops. One PM started with a single workflow saving 20 minutes. After a full audit, he found 10 hours of weekly savings. Here's the exact process.
The Trap of Automating Everything at Once
Most people who try AI workflows burn out by trying to automate everything simultaneously. The fix: pick three tasks, master them, then expand.
The Workflow Audit: Find 10 Hours of Wasted Time in Your Week
Most professionals have 10+ hours of repetitive work hiding in their calendar. Here's a 15-minute audit process to find every workflow candidate in your week.
How to Use AI as a Learning Tool, Not a Copy Machine
Most people paste AI responses without understanding them. Learn the sparring partner approach: get the output plus the reasoning, so you actually get better at the work itself.
One Role Per AI Prompt. Two at Most. Never Three.
Cramming multiple roles into one AI prompt produces mediocre output. Learn when to split roles into separate prompts and when combining actually works.
Why Simple AI Prompts Beat Complex Ones
Most people write massive AI prompts and wonder why the output is mediocre. The fix is the same principle behind good product development: start with an MVP prompt, test it, then grow.
Why Planning Before Building with AI Saves Hours
The biggest waste of AI output isn't bad prompts. It's skipping the planning step. Why planning mode in Claude Code changes everything, and a 5-minute checklist before any AI task.
Experience Level Calibration
A 30-second setting that changes how Claude responds to every prompt. Set your experience level for better output.
The Solo Dev Startup Stack
How one developer uses AI as PM, UX writer, and engineer. Turn a messy feature brief into spec, copy, and code in 10 seconds.
The Workflow Chain Pattern
Take one input and run it through multiple AI roles to produce different professional outputs. This separates chatting with AI from having a system.
The Pragmatic AI Workflow for Developers
Move beyond basic AI coding tips. Learn how to build a practical AI workflow with prompt chaining, reusable templates, and layered code review.
Claude Code Tutorial for Beginners
A practical beginner guide to Claude Code, Anthropic's CLI tool. Learn to install, configure, and use agentic AI coding directly in your terminal.
The Complete Faceless YouTube Channel Workflow with AI
A stage-by-stage AI production pipeline for faceless YouTube channels, from niche research through upload, built for operators who think in systems.
The Five Failure Modes
Data teams produce solid analysis and nothing happens. Here are the five reasons stakeholders ignore your insights, with numbers from real teams and a fix for each one.
AI Code Review Is Broken. Here Is the Fix
Most AI code review tools flag style issues and miss real bugs. Here is why AI code review fails at scale, and the 5-step fix that actually works.
AI-Powered Job Hunt Systems That Actually Work
Stop spray-and-pray applying. Build a real system with AI that handles sourcing, tailoring, and tracking your job search. Concrete tools, prompts, and workflows included.
EU AI Act Compliance for Small Teams: A Practical Field Guide
How small engineering teams can navigate EU AI Act requirements without drowning in legal overhead. Risk classification, technical documentation, and a compliance checklist you can actually use.
How to Stop AI Detection False Positives Without Degrading Your Writing
AI detection tools flag human writing as AI-generated up to 30% of the time. Here is exactly why that happens and what you can actually do about it.
Why AI Prompts Give Inconsistent Results
The same prompt, different answers every time. Here is what is actually happening inside language models and how to write prompts that produce reliable, repeatable output.
5 AI Pain Points Nobody Talks About (From 10,000 Reddit Posts)
We analyzed 10,000 Reddit posts about AI across 10 subreddits. Here are the 5 pain points experienced professionals face ·and they're not what you'd expect.