I posted "AI is still slow af" yesterday and watched the replies roll in.
Everyone had a horror story. The 20-minute wait. The agent that spun in circles. The answer that came back confident and wrong.
But here's the thing i kept noticing in the replies: almost nobody's slowness was the model thinking. It was everything around the thinking.
I've been running my whole operation on AI for a few months now, content, music art, my site, client work. The single biggest speed-up i ever got had nothing to do with a faster model. It came from cutting the three places the time actually leaks.
1. Re-explaining yourself every single time
Most "slow AI" is you, typing context the model already should have known.
Every conversation that opens with "i'm a content strategist, my brand voice is warm and lowercase, i hate em-dashes, here's how i format", that's not the model being slow. That's you re-uploading yourself by hand.
The fix isn't a smarter model. It's a file the model reads first.
i keep my voice, my rules, and my receipts in Obsidian
Claude reads what's relevant before it answers
the prompt becomes the request, not the re-introduction
My prompts got shorter and the first draft got closer, because the part that used to take ten lines of setup now takes zero.
2. Making one model do everything
A lot of slowness is a single model grinding on a job it was never the fastest at.
i don't run one model for everything. i route the task to whatever's actually best at it, and i keep the costs near nothing by splitting work across what each one does well, a heavy reasoning pass here, a fast cheap pass there, a local model for the boring bulk.
The wall most people hit is treating one subscription like one brain that has to do all of it. It doesn't. The speed is in the routing.
3. Running everything in a line
This is the big one, and it's the one the new agent tooling finally fixes.
If your "AI workflow" is one agent doing step one, then step two, then step three, you're not waiting on intelligence, you're waiting on a queue. A line.
My output didn't change the day i had more agents. It changed the day they stopped waiting for each other.
one reads
one drafts
one cuts
All at once. The time a sequential run spends waiting is just gone.
So is AI slow?
Sometimes, yeah. Models still stall. But before you blame the model, check the three things you actually control:
does it have to re-learn you every time, or does it take too long to read a file?
is one model doing a job another would finish faster?
are your steps running in a line when they could run side by side?
Fix those three and "AI is slow af" quietly turns into "wait, that was already done?"
The model was never the limiter. The setup around it might be.
What's the slowest part of your AI right now, the thinking, or everything you make it do before it can think? 🪽
BTW… THANK YOU FOR BEING AN INTEGRAL PART OF MY COMMUNITY
🧡
Here’s a quick gift for you.
Grab this guide for free to make your profile better
Use code FREE to get it for FREE!
FOLLOW WHAT I DO:
Everything here is documented as it runs.
X: @AunySillyMe
Newsletter: Explained Without Fluff
Website: aunysillyme.com
IF YOU WANT HELP BUILDING A CONTENT SYSTEM THAT ACTUALLY SOUNDS LIKE YOU
Book a 1-hour AI Workflow call with me.
We'll look at what you're currently doing, figure out where the friction is, and map out a system that works with how you actually think and create.
Check out some of the other services I offer:
Don’t forget to reply and let me know what you thought of this newsletter.
See you on the next one
- AUNY 🧡