
Have you ever run an AI model on your own computer?
Everyone on Twitter has been talking about owning your model lately, especially after the Fable fiasco in June.
And now that open-weight models like Qwen and Kimi are catching up, it’s only gotten louder.
It makes a great headline but most people cheering it have never run one of these models, even though doing so takes about 15 minutes.
So instead of explaining what "open" means this week, let's just go and use one.
You won't need to write code or pay for anything. If you can install an app, you can do this.
But first, some catchup:
AI Open Source 101
The big open model everyone's posting about, Kimi K3, can build a whole game from a single sentence, and this pack shows you how, step by step.
Every build from the walkthrough, a physics roulette wheel, a GTA-style open world, an exploding interactive guitar, each with the exact prompt behind it.
One prompt in, a working website out, published to a link you can share the same day.
Setup
1. Get the app.
Head to lmstudio.ai and download LM Studio, which is free and runs on Windows, Mac and Linux.
You install it like any other app, and most beginners find it the gentlest way in, which is why we're using it here.

2. Pick a small model.
Open the app and look at the models it suggests, then pick a small one, since the app helpfully marks which ones will fit your computer.

You're after something around 3 to 4 billion in size, a few gigabytes to download, and you'll see familiar names like Gemma, Llama and Phi, any of which is fine for a first go.
Click to download, and know that this is the one step that needs internet.

3. Talk to it.
Once the download finishes, open a fresh chat and type something the way you would with ChatGPT.
The first reply takes a few seconds while the model wakes up, and then it simply answers, with no account, sign-in or bill anywhere in sight.

4. Now the good bit.
Turn off your wifi, and then ask it another question.
It still answers.
Nothing is going out to a server or coming back, because it is running on the computer in front of you, and it would carry on working just the same anywhere with no signal at all.
What you're holding
That model is now a file on your own computer. Every number in it belongs to you, and no company or government can reach in and take it back.
In June the US government ordered Anthropic to pull Fable offline for most of the world, and overnight they were gone.
Nobody can do that to the model on your machine.
Try one more thing.
Ask the model what it was trained on, then look up its page online. You get a vague description and never the real answer.
You own the whole thing and still can't see what it learned from.
This is where open-weight and open source split.
You can run this model and change it however you like, and it's yours to keep. What you never got is the data or the method that made it, and that is the part open source would have handed over.

Nearly every model calling itself "open" this week is the middle type.
You just ran one, so now you know what it gives you and what it keeps back.
What it's good for (and what it isn't)
A small model on a laptop is not GPT-5. It will miss things the big cloud models get right, and it won't touch your hardest work.
The strong open models, like that 2.8-trillion Kimi, need a room full of machines and won't run on your laptop at all.
For everyday work, the small model is plenty.
So the next time you're about to paste something private into a cloud chatbot, there's a free option already running on your own machine.
My take
Run one model on your own laptop and the whole "open" argument stops being just talk.
10 minutes of downloading and one flick of the wifi switch will teach you more than any explainer can, this one included.
You find out what you own, and what you don't. Open-weight turns out to be a real gift, and open source the bigger one almost nobody hands over.
So the next time a post claims the open models have caught the giants, you'll know what that word bought them, and what it didn't.
And you'll have one of your own on your machine already, wifi or no wifi.
Hit reply and tell me which model you ran, and how it held up with the wifi off. I read every one.
Until next time,
Vaibhav 🤝🏻
If you read till here, you might find this interesting
#AD 1
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
#AD 2
Moda is the viral AI design agent for polished, on-brand slides, docs, ads, and more. Turn prompts into fully editable designs on a real canvas, then export to PowerPoint, PDF, and more.







