5 New Prompts & Today's Actionable News
By Robert MacKelfresh — August 10, 2026Turn a forty-message thread into a decision
EASYHere's a thread that has gone on too long. I need it to end. 1. WHAT'S ACTUALLY BEING DECIDED. One sentence. Threads sprawl…Open full prompt →for ChatGPT or ClaudeWrite a one-pager they'll actually read
EASYI need a one-pager on [THE SUBJECT] for [WHO]. 1. THE ASK, in the first two sentences. What I want and what it costs. Not…Open full prompt →for Claude or ChatGPTWork out what only you can do
MEDIUMHelp me work out what I should stop doing. Sort everything I listed into four buckets, and be blunt: 1. ONLY ME. Genuinely requires…Open full prompt →for ChatGPT or ClaudeSplit a big job across parallel workers
ADVANCEDI have a big job I want to split across several parallel AI workers. 1. IS THIS ACTUALLY PARALLELISABLE? Say no if the answer is no…Open full prompt →for Claude or ChatGPTGet what's in your best person's head out
EXPERTWe have someone who is very good at [THE THING]. Nobody else can do it to the same standard, and none of it is written down. Ask…Open full prompt →for Claude or ChatGPT
The Context Coda
One more thing: use anything you already know about me from our past chats and anything you're connected to. If you can find it yourself, don't make me look it up.
Paste this at the end of any prompt above. It tells your AI to use what it already knows about you instead of asking you to repeat it — which is the difference between a generic answer and a personal one.
Actionable News
What just happened, and the prompt, skill file, or setup that turns it into something you can use.
Anthropic cancelled the 50% Claude Sonnet 5 price rise scheduled for September 1
here's the current price table, plus the tokenizer note on the same page that quietly eats part of the saving
SETUP GUIDEMeta released Muse Glimmer, a 30B multimodal agentic model, under a genuinely permissive Apache 2.0 licence
here's the setup guide and the VRAM envelope, because under-20GB at 4-bit is not the same as fitting on a 20GB card
An unreleased Claude moved a Riemann bound from 41.6% to 67.2% using ~60 subagents and 31 million output tokens
→here's the orchestration recipe — the first public figure for what heavy parallel agent work actually costs
GET THE RECIPEChatGPT Business added a $125 Premium seat with 5× the usage and no five-hour limit
→here's the comparison table — and the mixing rule that means you upgrade four people, not forty
GET THE COMPARISON TABLEOpenAI's GPT-5.6-Cyber completes 95% of advanced security requests where the stock model completes 1.5%
→here's the refusal table, and why a 63-point gap produced by verification rather than capability is the thing to notice
GET THE REFUSAL TABLE
The AI Five is published by Automation Squad. Every prompt is written from scratch and every claim links to a primary source. Corrections: [email protected].