- AI Business Insights
- Posts
- I almost paid $100 to fix a $20 problem
I almost paid $100 to fix a $20 problem
I had the upgrade button under my thumb last week
Sponsored by
My Claude usage kept hitting the wall by lunchtime, and that shiny $100 plan looked like the obvious fix. One click, problem solved, back to work. Then I stumbled on a post that stopped me cold.
The original poster laid out 21 hacks that make the $20 a month plan more than enough for most people. The core idea from this AI professional is simple. You are probably not out of power. You are burning tokens on sloppy habits.
I read the whole thing twice, and honestly, a few of these changed how I work inside Claude. Here is the breakdown of what the creator shared, cleaned up and explained so you can start using them today.
The best voice models now listen, adapt, and resolve too.
Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one.
ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient.
You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change.
More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute.
*Ad
The waste starts before you even type the first word
Skip raw PDF uploads. One PDF page eats around 3,000 tokens. The author suggests pasting the text into a Google Doc, downloading it as .md, and feeding that in. Same content, under 200 tokens.
Plan in Chat, build in Cowork later. Jumping into Cowork too early wastes cycles. Figure out exactly what you want in a normal chat first, then move over.
Write short prompts, not essays. Long 500-word prompts reload constantly. The creator's fix is a 29-word template: "I want to [task] to [goal]. Ask me questions using AskUserQuestion." Let Claude pull the details out of you.
Fix one section, not the whole thing. Instead of "redo everything," try: "Only redo section 3. Keep everything else. No commentary. Just the output."
Batch your tasks. Three messages for three jobs burns three loads. Combine them: "Summarize this, list the points, suggest a headline."
Manage the conversation like a pro
Edit instead of correcting. Typing "No, I meant..." stacks more history onto the chat. Click Edit on your original message, fix it, and regenerate clean.
Keep a prompt library. Rewriting prompts from scratch every time is wasted effort. The expert keeps a reusable structure and just swaps the variable.
Match the model to the job. Use Sonnet for quick tasks. Save the heavier Fable and Effort modes for deep work that actually needs them.
Trim your about-me file. A 22,000-word profile is overkill. The author recommends cutting it under 2,000 words and turning it into a reusable Claude Skill.
Restart when things go sideways. Do not keep stacking a broken chat. Click "Restart the conversation from here" on an earlier message and get a clean slate.
Summarize before it gets long. Every 15 to 20 messages, summarize, copy the brief, and start fresh. Old context is dead weight.
PRDs by voice. Bug reports by voice. Ship faster.
Dictate acceptance criteria and reproductions inside Cursor or Warp. Wispr Flow auto-tags file names, preserves syntax, and gives you paste-ready text in seconds. 4x faster than typing.
*Ad
Set up your workspace once
Use Projects for recurring files. Upload a file once into a Project, and every chat inside references it without re-burning tokens.
Only load what the task needs. Dumping 50 files into Cowork just in case is expensive. For quick jobs like an email draft, use zero folders.
One topic, one chat. Three topics in a single chat means Claude re-reads all of them. New topic gets a new chat, always.
Turn features off by default. Search and connectors left on by default drain resources. The creator flips everything off and turns tools on per task.
Work smarter with automation and the right tool
Schedule recurring reports. Running the same weekly report by hand is work a scheduler should own. Use /schedule: "Every Monday at 7am, create my weekly briefing."
Be specific with Claude Code. Do not let it explore your whole repo. Try: "Build a bar chart from this CSV. Save as chart.png."
Set Personal Preferences. Go to Settings, then Personal Preferences, and set your tone and style once so you stop re-explaining it.
Speak your prompts. Lazy prompts like "make it better" get lazy results. This industry pro dictates prompts with a voice tool like wispr.ai for richer context in one shot.
Respect the rolling window. Claude runs on a rolling 5-hour window. Do not blow your whole limit in one morning. Split the work across the day.
Know what Claude cannot do. Do not force it. Use ChatGPT for images and Grok for real-time search.
Not just another AI newsletter. MavSource aggregates updates from all major AI newsletters, podcasts, company news, AI labs, and hundreds of other sources — then summarizes what matters, analyzes emerging trends, and adds founder commentary. One 5-minute daily email. Free.
*Ad
Where I landed after sitting with this
The part that stuck with me is not the money. It is realizing that most of the friction I blamed on the tool was friction I introduced myself, one careless upload and one bloated prompt at a time.
Twenty-one habits is too many to install at once, so do not try. Tonight, pick the three that match how you actually work. Heavy on documents, start with number 1. Living in long chats, run 10 and 11. Running dry before lunch, respect number 20.
The upgrade button is the lazy fix. The workflow is the real one.