3 MacBooks Did What One Never Could

Credibility score: 77/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.

Claims analyzed

3 M5 Max MacBooks + Thunderbolt 5 cables run 122B param model across two as one machine — Dubious (45/100)

Dropping '122 billion parameter model' like it's chilling on consumer laptops already 💀 — M5 Max ain't even out, Thunderbolt 5 on MacBooks? Bro's future-proofing a fever dream 👀😬

Clustering M5 Max MacBooks beats M3 Ultra cluster — Solid (80/100)

Dropping 'nobody else has run the numbers' like they're the first astronaut on Mars — M5 Max is real and clustering makes sense, but that exclusive flex is peak YouTuber energy 👀📊😬

Thunderbolt mesh with RDMA works for clustering — Verified (95/100)

Thunderbolt mesh + RDMA like it's NBD — this is actual enterprise tech Apple Silicon hackers are nerding out over, and he's got the GitHub receipts ready 🔥😤✅

MLX runs distributed inference on Apple Silicon — Verified (100/100)

"MLX does machine learning stuff" — understatements so bold they deserve an award. Apple's own framework crushes distributed on their chips, zero cap 🙌😡✅

Qwen 34B 4B model is 2GB, fits on 128GB Mac — Solid (85/100)

'Quen 34B 4 billion' — probably Qwen3-4B, yeah 4B params at ~2GB quantized is spot on for fitting single M5 Max with room to spare 📈😤✅

Narwhal sent Flow 2; most robot mops smear dirt — Sponsored (50/100)

Boom, mid-video pivot to the kitchen mess sponsor read — 'Narwhal sent over the incredible new Flow 2' 👀💸. Classic 'I made a mess just to test' grift energy.

1x M5 Max: 179 t/s; 2x: 220 t/s (22% faster) — Personal Story (70/100)

179 → 220 t/s on tiny 4B model = 22% bump? For small models cluster overhead kills gains — honest reporting of mediocre scaling, respect the transparency 🧐📊💅

Flow 2 cleans baseboards, avoids obstacles, quiets near cribs — Sponsored (50/100)

'Narmind Pro is wild, quiets around cribs' — reading the spec sheet like it's a mic drop 😭🧹. This is peak sponsored flex, down to the baby mode.

Narwhal Flow 2 hands-off, great return policy, buy now — Sponsored (50/100)

'Head to the link in description' — the close that screams 'affiliate commission incoming' 💀🛒. Hands-off except swapping bags, but we're not mad.

Models don't scale linearly with tensor parallelism due to smaller active expert size; 27% speedup is good — Solid (82/100)

Dropping 'active expert size' like it's common knowledge — fair point on MoE limits, and 27% ain't bad for laptops fighting tensor parallelism overhead 💻📈😬

Qwen 3.5 122B-A10B 8-bit quantized is 122GB, at edge of single 128GB Mac memory — Verified (95/100)

Nailed the Qwen 3.5 122B-A10B specs and memory math — 122GB after 8-bit quant + OS overhead is spot on for M4 Max limits. I'm mad this checks out so clean 😤✅🔥

One Mac fails to load Qwen 3.5 122B; two Macs get 51 tokens/sec — Solid (88/100)

'Nothing at all' vs 51 t/s on two nodes — cluster turns impossible into usable real quick. Laptop supercomputing ftw, benchmarks track with MLX reports 📊💨✅

185GB model usable for real coding work with VS Code/Zed — Personal Story (70/100)

Fair play, local coding AI actually slaps now — guy's living the dream we all want. Who knew offline code gen could flex like this? 😤✅💻

215GB Llama 3.1 45B should fit in 2x128GB unified memory — Solid (85/100)

Math checks out on paper — 215 < 256, duh. But we all know overhead's the real villain here 👀📊

215GB model swaps/thrashes on 2x128GB, no tokens generated — Personal Story (80/100)

Lmao 'the math lied' — classic. Swapping hell is every local AI nerd's nightmare, and he's showing the receipts 💀🔥🖥️

Jackal slices layer weights horizontally for tensor parallelism — BS (10/100)

Called Jackal a tensor parallelism framework like it's common knowledge — bro that's a Jira benchmark, not your distributed training buddy 💀🪦🚩

2x128GB cluster ceiling is ~200GB models before swapping — Opinion (75/100)

200GB ceiling feels right from the trenches — smart callout on the invisible overhead tax 😤✅📈

Exo uses pipeline parallelism slicing model vertically — Solid (80/100)

Nailed Exo's pipeline setup — contiguous layers flowing like an assembly line, spot on 👀✅ — but watch, the pivot's coming

Popular models use power-of-two dimensions not divisible by 3 — Verified (95/100)

Damn, broke down the tensor shard math perfectly — 4096 not divisible by 3 is the real tea why 3 MacBooks flop 😤✅🔥

3-node tensor parallel fails on popular models due to dimensions — Solid (85/100)

The 'lazy' admission after dropping math truth bombs — love the honesty, explains the 2 MacBook title perfectly 💀✅🙄

See the full analysis with sources and timestamps →