It’s been a thing on the socials for a couple of weeks. Funny to see some people absolutely destroying their little bot.
Im just to polite I always am friendly untill I get pissed off! ![]()
haha I love that.
Check this out. If you love Stormlight Archive from Sanderson, must watch:
That was really well done! Good interpretation of the scene in the book.
AI must hurry up and create the whole thing!!!
Cheap holiday!
I’ve been playing and perfecting my prompt to help give me oversight into my finances and thought this could help anyone who is also curious, prompt here:
Role: Act as my personal certified financial analyst. I will upload my bank statements for the last [Insert Number, e.g., 3 or 6] months.
Objective: Conduct a deep-dive forensic analysis of my spending. I want you to identify my habits, find inefficiencies, and build a realistic, livable budget for me going forward.
Instructions: Please structure your response in the following 5 sections. Do not skip any sections.
Section 1: The "Big Picture" Summary (Executive Summary)
Calculate my total income (average monthly).
Calculate my total expenses (average monthly).
Calculate my current burn rate (savings rate or deficit).
Categorize my top 3 largest spending categories (e.g., Housing, Food, Transport).
Section 2: Spending Habit & Trend Analysis (The "Health Check")
Identify any "Subscription Creep" (recurring charges I might have forgotten about).
Identify any "Lifestyle Inflation" (are my discretionary expenses increasing month over month?).
Flag any specific transactions that seem unusually high for my income level (e.g., frequent dining out, excessive coffee runs, high grocery bills for one person).
Identify the average cost per transaction for restaurants/food delivery.
Section 3: The "Low-Hanging Fruit" (Savings Opportunities)
Give me the Top 3 actionable things I can cut or reduce immediately without drastically changing my lifestyle.
Suggest specific alternatives (e.g., "Instead of buying lunch 5 days a week, try 3 days and meal prep for 2; saving approx $X").
Section 4: The Budget Proposal (The 50/30/20 or Custom Rule)
Propose a new monthly budget for me based on the data.
Break it down into these specific buckets (adjust the percentages as you see fit based on my data):
Fixed Necessities (Rent/Mortgage, Utilities, Insurance, Minimum Debt Payments).
Discretionary/Variable (Groceries, Dining, Entertainment, Shopping, Gas).
Savings/Debt Payoff (What I should aim to save/invest).
Important: The total budgeted amount must be less than my average monthly income. Tell me exactly how much cash flow I will have left over each month if I follow this.
Section 5: The Action Plan (Next 30 Days)
Give me a 3-step plan to implement this budget starting tomorrow.
Provide 2-3 specific questions I should ask myself before making a purchase in my "danger categories."
Final Rule: Be brutally honest, but constructive. Do not judge me; just give me the hard data and practical solutions. If I am overspending, tell me clearly.
Myself and Codex have hit a rythm. It has access to all my personally written test code and it is making new test creation and test maintenance a breeze, as long as you use a variety of tricks to keep the credit usage down.
Its singular magical ability to spot patterns and apply them with variation make writing code to test website operations a breeze.
GPT 5.4-Mini on Medium reasoning is my sweet spot. It’s the perfect intern/junior dev that I can order around. And yes, tricks like starting new chats often, small ruleset to remind it how to work upfront, not letting it sprawl through whole codebases and only targeted based on what I tell it to do, not letting it run builds and tests on its own, etc.
By doing all that I seldom hit the rate limit. I’ve only hit it twice after extended sessions, and even then the rate limit resets within around 20 minutes I hit it.
I run Claudette on sonnet 4.6 medium and markdown all my pdfs to make life easier - tokens is a thing
I think you were looking into a local deployment at some point. Did you abandon that idea for the newer models that keep appearing? Or have you found a middle ground? I’ve gotten Ollama integrated in my Home Assistant config, so that some notifications are not hard coded, but nothing more fancy than that yet.
I’ve given up on local LLMs in recent times. I don’t have the time to fiddle with them anymore. Work has been consuming my life over the past year and grade 2 is a very difficult emotional and developmental year for our son (not the academic part though, he flourishes in that). So I’m just falling back to the options that are plug-and-play inside my IDE to get stuff done.
Exactly what I do! It works.


