You finally get access to a shiny new AI model, fire it up, and the answers feel… mid. Not because the model is weak. Because nobody told you how to actually turn it on the right way. I ran into this exact wall, and then I found a LinkedIn post that fixed it in about two minutes.
This whole breakdown comes from a sharp AI professional who read the OpenAI docs so the rest of us don’t have to. The original poster laid out the precise setup for the new ChatGPT-Astra-6 (GPT-6 Astra), plus a set of prompt tricks that squeeze way more out of it. I was genuinely impressed by how simple the creator made it, so let me walk you through what they shared.
Step one: set it up the right way
The expert is blunt about this part. Most people open ChatGPT, start typing, and never realize they’re using a weaker path. Here’s the exact sequence the author recommends:
- Skip the browser version of ChatGPT completely, it’s not the play here
- Download the ChatGPT app first, that’s your home base
- Stick to Work or Codex modes only, nothing else
- Pick “Medium” as your effort level, the author calls it the clear winner
- Follow the path: ChatGPT Work, then GPT-6 Astra, then Medium
Simple as that. The creator boiled a pile of docs down to one clean route.
Why “Medium” is the sweet spot
This is the part I keep coming back to. You’d think the highest setting always wins. The original poster says no. According to them, Medium sits right where intelligence, cost, and speed all meet. You get smart answers, you don’t burn extra money, and it stays fast. High effort isn’t automatically better, it’s just heavier. That’s a small insight with a big payoff, especially if you run a lot of prompts in a day.
The free-tier catch
Quick heads-up the author flags twice, so it clearly matters. You cannot touch GPT-6 Astra on a free ChatGPT account. None of it. OpenAI is serving Pro users first, then rolling out to Plus users after that. So if you’re on the free plan and wondering why you can’t find Astra, that’s your answer. No workaround, just the current rollout order.
Thinking about jumping from Claude?
The contributor also weighed GPT-6 against Claude’s Fable-5.1, and the pitch is pretty direct. Here’s how they framed the comparison:
- GPT-6 is smarter than Fable-5.1, in their view
- GPT-6 is about 2x cheaper to run
- GPT-6 is also roughly 2x faster
Smarter, cheaper, faster. That’s the case the author makes for switching. Worth a test if you’re already paying for both and want to trim your stack.
Five prompt tricks that actually move the needle
This is where the post really earns its spot. The expert shared five ways to prompt Astra so it stops guessing and starts delivering. Each one fixes a specific gap:
- Give it a finish line. “Help me analyze this” leaves too much open. Instead, state the exact thing you want by the end: a ranked list, a spreadsheet, a finished deck. Name the destination.
- Decide which questions it may ask. Astra asks for clarification more than older models, and the author says that’s a good thing. Put the key decisions right in your input, and let it handle the rest without a back-and-forth.
- Tell it what each file is for. Is this evidence? A template? A style reference? One quick line per file saves a whole round of clarifying chat.
- Tell it what “readable” means. This is the author’s favorite move, and it’s the one exact prompt line worth stealing. They add this to their prompts to get plain, uncomplicated answers:
Use ASD-STE100.
That single instruction points the model at a simplified-English standard, so the output comes back clean instead of tangled in jargon.
- Interrupt mid-task with a new instruction. You don’t have to wait for Astra to finish. Drop a new direction while it’s working, and the creator says it updates the plan from that point and keeps rolling.
Why it matters: none of these tricks need a fancy setup. They’re just clearer inputs. The person who shared this basically proved that most “bad AI answers” are really bad prompts wearing a disguise.
My quick take
What I love here is how little effort it takes to apply. Download the app, pick Work mode, choose GPT-6 Astra, set it to Medium, then prompt with a finish line in mind. That’s the entire system. I learned more from this one post than from a week of poking at settings myself.
The original poster packed a lot into a short post, including the reasoning behind every step, which is exactly why it’s worth the read. Go check out the full LinkedIn post for the complete walkthrough and the finer details straight from the author.