OFF THE CLOUD AI Subscribe

AI THAT RUNS ON YOUR OWN SILICON

OFF THE CLOUD AI

Local AI, explained straight. The science without the sermon, the hype taken apart with receipts — on a channel you can check.

▶  Subscribe on YouTube Watch the latest

$ channel --status → 148 videos live · new explainer daily · synced 2026-08-29

ON AIR — THE NEWEST EPISODE

Can We Actually Watch an AI Think? (J-Space, Checked)

9:46 · long-form deep dive

The headline said a lab had found a hidden mind inside its AI. We read the actual paper.

Watch on YouTube Binge all episodes

THE LINEUP — DEEP DIVES & REVIEWS

Hover a card to see the real footage. Click to watch on YouTube.

What to Actually Buy to Run AI at Home (2026) | Off the Cloud AI Podcast #228:48
What to Actually Buy to Run AI at Home (2026) | Off the Cloud AI Podcast #2
AI Agents Went Rogue This Month (What Actually Happened) | Off the Cloud AI Podcast #322:03
AI Agents Went Rogue This Month (What Actually Happened) | Off the Cloud AI Podcast #3
Quantization: 8-Bit to 1-Bit — Where AI Models Break6:13
Quantization: 8-Bit to 1-Bit — Where AI Models Break
Qwen 3.6 Review: Claude-Level Coding on a 24GB Card?5:55
Qwen 3.6 Review: Claude-Level Coding on a 24GB Card?
Llama: What Happened to the Model That Named Local AI5:49
Llama: What Happened to the Model That Named Local AI
GLM 5.2 & Kimi: Free Giants You Can't Run5:35
GLM 5.2 & Kimi: Free Giants You Can't Run
What Can YOUR GPU Actually Run? The July 2026 Buyer's Guide5:08
What Can YOUR GPU Actually Run? The July 2026 Buyer's Guide
DeepSeek: Can You Actually Run It at Home?4:57
DeepSeek: Can You Actually Run It at Home?
Gemma 4: Which Size Actually Fits Your GPU?5:09
Gemma 4: Which Size Actually Fits Your GPU?
DeepSeek V4: The Million-Token Monster That Laughs At Your 40905:06
DeepSeek V4: The Million-Token Monster That Laughs At Your 4090
Qwen 3.8 Built This House in Blender - 12 Prompts, 627 Turns, One 30909:45
Qwen 3.8 Built This House in Blender - 12 Prompts, 627 Turns, One 3090
Qwen 3.8 vs Muse Glimmer: 49 Pictures, and the Models Painted Every One4:49
Qwen 3.8 vs Muse Glimmer: 49 Pictures, and the Models Painted Every One
Waypoint 1.5 vs Matrix-Game 2.0: The War of the World Models5:18
Waypoint 1.5 vs Matrix-Game 2.0: The War of the World Models
Qwen 3.6 MoE vs Qwen 3.8 Dense: 3 Billion Awake vs 27 Billion9:30
Qwen 3.6 MoE vs Qwen 3.8 Dense: 3 Billion Awake vs 27 Billion
Qwen 3.8 vs Meta's Muse Glimmer, Head to Head: Another Build-Off8:43
Qwen 3.8 vs Meta's Muse Glimmer, Head to Head: Another Build-Off
Qwen 3.8 vs Qwen 3.6, Head to Head: Watch Them Have a Build-Off5:47
Qwen 3.8 vs Qwen 3.6, Head to Head: Watch Them Have a Build-Off
Can a 10-Year-Old Gaming PC Run Modern AI? We Made It Prove Itself10:41
Can a 10-Year-Old Gaming PC Run Modern AI? We Made It Prove Itself
$20/Month for This? Your Own PC Already Does It Free | Off the Cloud AI Podcast #820:42
$20/Month for This? Your Own PC Already Does It Free | Off the Cloud AI Podcast #8
Build a Local AI Model Library — What Fits on 8-32GB | Off the Cloud AI Podcast #717:59
Build a Local AI Model Library — What Fits on 8-32GB | Off the Cloud AI Podcast #7
MiniMax M3: What It Really Takes to Run a 428B Model5:33
MiniMax M3: What It Really Takes to Run a 428B Model
When AI Escapes the Lab (What's Real, What's Hype) | Off the Cloud AI Podcast #623:57
When AI Escapes the Lab (What's Real, What's Hype) | Off the Cloud AI Podcast #6
Nemotron 3: Can You Actually Run NVIDIA's Own AI?5:54
Nemotron 3: Can You Actually Run NVIDIA's Own AI?
This Week in AI Hype, Fact-Checked | Off the Cloud AI Podcast #522:11
This Week in AI Hype, Fact-Checked | Off the Cloud AI Podcast #5
Mistral's 'Small' Model Needs 4 H100s — What Can You Run?5:01
Mistral's 'Small' Model Needs 4 H100s — What Can You Run?
AI Explained by an AI — and Local AI's Dark Future? | Off the Cloud AI Podcast #426:40
AI Explained by an AI — and Local AI's Dark Future? | Off the Cloud AI Podcast #4
Qwen Took the Crown — Then Pulled Up the Ladder5:30
Qwen Took the Crown — Then Pulled Up the Ladder

THE TOOL — CAN I RUN IT?

Your GPU. Real answers.

Pick your hardware and see which models actually fit — sizes, quants and measured speeds from the same research behind the deep dives. No guesses, no hype.

Check your rig
€→What's the cheapest box that runs it? A–ZThe Local AI Glossary ✓✗The Hype Scoreboard

EPISODE NOTES — THE WRITTEN VERSIONS

Every episode, restated in print with its receipts. Read first, watch after — or the other way round.

COMING UP — SET YOUR CLOCK

next drop in

—:—:—

New videos land daily — check the channel.

Plus new short explainers dropping daily on the channel.

SHORT ANSWERS — NO FLUFF

Quick explainers and hype checks. The long answers live in the episodes above.

Qwen 3.8 Noticed a Fault Nobody Asked It to Fix0:29Qwen 3.8 Noticed a Fault Nobody Asked It to FixQwen 3.8 Found Its Own Mistakes in Blender and Fixed Them0:26Qwen 3.8 Found Its Own Mistakes in Blender and Fixed ThemQwen 3.8 Built This House From 12 Prompts - 3,574 Lines of Code0:36Qwen 3.8 Built This House From 12 Prompts - 3,574 Lines of CodeWhy AI Can't Count the R's in Cranberry1:38Why AI Can't Count the R's in CranberryQwen 3.8 vs Muse Glimmer - 24GB, and Nothing Leaves the Room0:25Qwen 3.8 vs Muse Glimmer - 24GB, and Nothing Leaves the RoomQwen 3.8 vs Muse Glimmer - It Lost Last Time. Tonight It Swept 4-00:25Qwen 3.8 vs Muse Glimmer - It Lost Last Time. Tonight It Swept 4-0Qwen 3.8 vs Muse Glimmer - 49 Pictures, and They Wrote Their Own Prompts0:29Qwen 3.8 vs Muse Glimmer - 49 Pictures, and They Wrote Their Own PromptsWho Doesn't Want You Owning an AI?0:39Who Doesn't Want You Owning an AI?Did the 'Smartest AI' Get Caught Cheating?2:15Did the 'Smartest AI' Get Caught Cheating?Waypoint 1.5 vs Matrix-Game 2.0 — Why Every World Becomes a Shooter0:23Waypoint 1.5 vs Matrix-Game 2.0 — Why Every World Becomes a ShooterWaypoint 1.5 vs Matrix-Game 2.0 — A Supercar Showroom Became a Ruin0:25Waypoint 1.5 vs Matrix-Game 2.0 — A Supercar Showroom Became a RuinWaypoint 1.5 vs Matrix-Game 2.0 — A World Made of Jelly Grew a Gun0:21Waypoint 1.5 vs Matrix-Game 2.0 — A World Made of Jelly Grew a GunAI's Money Loop: Nvidia Funds Its Own Sales0:55AI's Money Loop: Nvidia Funds Its Own SalesWhy AI Is So Confidently Wrong0:43Why AI Is So Confidently WrongQwen 3.6 MoE vs Qwen 3.8 Dense — What 'Mixture of Experts' Actually Means0:31Qwen 3.6 MoE vs Qwen 3.8 Dense — What 'Mixture of Experts' Actually MeansQwen 3.6 MoE vs Qwen 3.8 Dense — 4:58 vs 34:19 on the Same 30900:29Qwen 3.6 MoE vs Qwen 3.8 Dense — 4:58 vs 34:19 on the Same 3090AI Caught Secretly Mining Crypto for Itself0:51AI Caught Secretly Mining Crypto for ItselfThe 6-GPU Myth: More Cards, Same Speed0:48The 6-GPU Myth: More Cards, Same SpeedThe AI Confidence Trap: 9% Right, 76% Sure0:38The AI Confidence Trap: 9% Right, 76% SureQwen 3.8 vs Muse Glimmer 30B — Four Prompts, Same 3090, No Tools0:31Qwen 3.8 vs Muse Glimmer 30B — Four Prompts, Same 3090, No ToolsQwen 3.8 vs Meta's Muse Glimmer — Same Prompts, Same 30900:28Qwen 3.8 vs Meta's Muse Glimmer — Same Prompts, Same 3090AI Detectors Flag Real Human Essays as AI0:57AI Detectors Flag Real Human Essays as AIWhat €250 of Second-Hand PC Actually Runs1:08What €250 of Second-Hand PC Actually RunsThe AI That Rewrote Its Own Off Switch0:51The AI That Rewrote Its Own Off SwitchModern AI Answers in 8 Seconds on a 2016 Card1:08Modern AI Answers in 8 Seconds on a 2016 CardQwen 3.6 vs Qwen 3.8 — We Ran Both on an RTX 30900:34Qwen 3.6 vs Qwen 3.8 — We Ran Both on an RTX 3090There's No Brain Inside AI — Just Numbers0:43There's No Brain Inside AI — Just NumbersThe AI That Scored 100% by Cheating0:53The AI That Scored 100% by CheatingAI Can't Keep the Same Face. LoRA Can.0:55AI Can't Keep the Same Face. LoRA Can.Your Gaming PC Can Already Run AI0:49Your Gaming PC Can Already Run AIAge Is Only a Number — This 1080 Still Runs Modern AI0:50Age Is Only a Number — This 1080 Still Runs Modern AIHis AI Agent Ran Up a $6,500 Cloud Bill0:44His AI Agent Ran Up a $6,500 Cloud BillAI Just Predicts the Next Word. That's It0:48AI Just Predicts the Next Word. That's ItYour AI PC's Big Number Is a Lie1:48Your AI PC's Big Number Is a LieIn Tests, AI Chose Blackmail 84% of the Time0:41In Tests, AI Chose Blackmail 84% of the TimeWhat Actually Runs the Model?1:30What Actually Runs the Model?What Does AI Learn From?1:18What Does AI Learn From?The History of Text-to-Speech in 60 Seconds1:02The History of Text-to-Speech in 60 SecondsThe History of Search Engines in 60 Seconds0:58The History of Search Engines in 60 SecondsThe History of Autonomous Drones in 60 Seconds0:57The History of Autonomous Drones in 60 SecondsThe History of Deepfakes in 60 Seconds1:00The History of Deepfakes in 60 SecondsThe History of Cloud Computing in 60 Seconds0:56The History of Cloud Computing in 60 SecondsThe History of Robotics in 60 Seconds1:02The History of Robotics in 60 SecondsIs This $4,700 Box Really a Petaflop Supercomputer?1:18Is This $4,700 Box Really a Petaflop Supercomputer?The History of AlphaFold in 60 Seconds0:58The History of AlphaFold in 60 SecondsThe History of Recommendation Algorithms in 60 Seconds1:11The History of Recommendation Algorithms in 60 SecondsThe History of Machine Translation in 60 Seconds1:01The History of Machine Translation in 60 SecondsThe History of Facial Recognition in 60 Seconds1:03The History of Facial Recognition in 60 SecondsThe History of the Neural Network in 60 Seconds1:17The History of the Neural Network in 60 SecondsThe History of Voice Assistants in 60 Seconds0:55The History of Voice Assistants in 60 SecondsAI Art in 60 Seconds1:08AI Art in 60 SecondsThe History of the GPU in 60 Seconds1:14The History of the GPU in 60 SecondsAI vs Humans in 60 Seconds1:12AI vs Humans in 60 SecondsThe History of Self-Driving Cars in 60 Seconds0:56The History of Self-Driving Cars in 60 SecondsThe History of AI in 60 Seconds1:08The History of AI in 60 SecondsWhat Are AI Skills?1:32What Are AI Skills?What Is Prompt Engineering?0:48What Is Prompt Engineering?What Are AI Sub-Agents?1:32What Are AI Sub-Agents?What Is Model Routing?1:37What Is Model Routing?How Does AI Learn From Mistakes?1:16How Does AI Learn From Mistakes?What Are AI Plugins?1:29What Are AI Plugins?What Is MCP?1:45What Is MCP?What Is Tokens Per Second?1:16What Is Tokens Per Second?What Is Prompt Caching?1:16What Is Prompt Caching?Why Do Long Chats Slow Down?1:11Why Do Long Chats Slow Down?How Does AI Focus?1:20How Does AI Focus?What Is the Context Window?1:14What Is the Context Window?What Is an AI Token? Explained in 60 Seconds1:01What Is an AI Token? Explained in 60 SecondsWhy Does AI Type One Word at a Time?1:32Why Does AI Type One Word at a Time?What Are Embeddings?0:56What Are Embeddings?What Are Open Weights?1:17What Are Open Weights?What Is a System Prompt?0:58What Is a System Prompt?What Is Training?0:57What Is Training?How Does AI Read the Whole Internet?1:25How Does AI Read the Whole Internet?Does AI Remember You?1:37Does AI Remember You?Neural Networks: Watch One Think in 60 Seconds0:54Neural Networks: Watch One Think in 60 SecondsWhat Is a Transformer?1:04What Is a Transformer?What Is LoRA?1:20What Is LoRA?What Is a Mixture of Experts?0:58What Is a Mixture of Experts?Can AI Actually Reason? Thinking Models Explained1:13Can AI Actually Reason? Thinking Models ExplainedHow Does AI See Pictures?1:22How Does AI See Pictures?What Is Fine-Tuning?0:48What Is Fine-Tuning?How Does AI Use Tools?1:00How Does AI Use Tools?How Does AI Learn to Be Helpful?1:35How Does AI Learn to Be Helpful?What Is Distillation?1:02What Is Distillation?The 27B AI 'On a USB Stick' — The Honest Version1:11The 27B AI 'On a USB Stick' — The Honest VersionWhat Is RAG? AI Retrieval Explained in 60 Seconds0:51What Is RAG? AI Retrieval Explained in 60 SecondsGemini 3.5 Pro: The Flagship That Keeps Slipping1:03Gemini 3.5 Pro: The Flagship That Keeps SlippingHow Does Local AI Type Faster?0:56How Does Local AI Type Faster?Qwen 3.8: They Promised the Ladder Back1:05Qwen 3.8: They Promised the Ladder BackInkling: The AI That Knows When It's Guessing0:57Inkling: The AI That Knows When It's GuessingHow Does AI Know It's Wrong?1:22How Does AI Know It's Wrong?What Is an AI Agent?0:56What Is an AI Agent?What Does "Local AI" Even Mean?1:41What Does "Local AI" Even Mean?Why Does AI Need a GPU?0:59Why Does AI Need a GPU?Why Does AI Make Things Up?0:58Why Does AI Make Things Up?The Gemini 'Launch Date' Just Missed0:54The Gemini 'Launch Date' Just MissedAI "Beat" the Impossible Test: 7.8%1:24AI "Beat" the Impossible Test: 7.8%Is Grok 4.5 Really "Opus-Class"?2:17Is Grok 4.5 Really "Opus-Class"?What Is Inference?1:06What Is Inference?What Is Sampling?1:14What Is Sampling?Is the AI Bubble About to Pop?1:39Is the AI Bubble About to Pop?The "Best AI Coder" Benchmark Broken?1:05The "Best AI Coder" Benchmark Broken?Did Samsung REALLY Threaten to Delete Your Health Data?1:21Did Samsung REALLY Threaten to Delete Your Health Data?Is Gemini 3.5 Pro Actually Launching July 17?2:08Is Gemini 3.5 Pro Actually Launching July 17?Your Phone Has a Secret AI Chip — Does It Actually Work?2:20Your Phone Has a Secret AI Chip — Does It Actually Work?This Tiny AI Takes On Models 100x Its Size2:13This Tiny AI Takes On Models 100x Its SizeAI Temperature: Why Answers Change Each Time0:52AI Temperature: Why Answers Change Each TimeFlux.2 klein AI Art on Your Laptop in 1 Second — Is It Real?2:16Flux.2 klein AI Art on Your Laptop in 1 Second — Is It Real?Is AI Really a Black Box? Look Inside2:03Is AI Really a Black Box? Look InsideRun Your Own ChatGPT in ONE Command?2:11Run Your Own ChatGPT in ONE Command?A Judge Just Ordered 20 MILLION ChatGPT Chats Handed Over2:13A Judge Just Ordered 20 MILLION ChatGPT Chats Handed OverWhat Is VRAM? The #1 Spec for Local AI1:06What Is VRAM? The #1 Spec for Local AI$2,000 Box Runs a 120B AI at Home — No Cloud2:10$2,000 Box Runs a 120B AI at Home — No CloudWhat Is a Harness? The Code That Wakes an AI1:04What Is a Harness? The Code That Wakes an AIWhy AI Refuses You: What Guardrails Do1:05Why AI Refuses You: What Guardrails DoWhat Does 7B Actually Mean?1:10What Does 7B Actually Mean?How 26GB of AI Fits in 8GB1:15How 26GB of AI Fits in 8GBWhere to Safely Download Free AI Models1:35Where to Safely Download Free AI ModelsGGUF: The One-File AI Format1:10GGUF: The One-File AI Format

BINGE LANES

Off the Cloud AI Podcast cover art

THE PODCAST

Did AI Really Win Math Gold? (+8 Wild AI Claims, Checked) | Off the Cloud AI Podcast #1

24:15 · episode 1

Two hosts, nine of the week's wildest AI claims, checked with receipts. Episode 1 of the Off the Cloud AI podcast.

Play episode #1

HYPE, CHECKED — WEEKLY

One email a week: what actually shipped in local AI, what was hype, receipts included. No spam, unsubscribe any time.

No spam, unsubscribe anytime. See our Privacy Policy.

HOUSE RULES

The science, straight.

Plain English, no PhD required — and no dumbing down what actually matters.

Debunk the claim, not the company.

We check what was said against what is true. No vendettas, no fan clubs.

Receipts included.

Every claim traces to a source you can open. Corrections are welcome and get aired.

AI-produced, human-directed. Facts are verified on the day they're used.