Google Just Found a Loophole in AI Hardware Limitations
Credibility score: 55/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
Claims analyzed
12B Gemma bridges both size and capability gap — OK (60/100)
Sounds plausible but no benchmarks shown yet — just the speaker's read.
12B unified beats E models "without thinking" — Dubious (45/100)
"Significantly better than thee without thinking" — the phrasing is doing the work here; charts aren't quoted.
Qwen 3.5 9B beats Gemma 4 12B on published benchmarks — Dubious (45/100)
Mentions an unnamed benchmark mashup — zero link, zero numbers.
QAT isn't a new Google invention — Solid (75/100)
QAT predates Google frontier releases by years — checks out.
This is the first frontier lab using QAT at scale — Dubious (45/100)
Meta already applied QAT to original Llama models — not first.
Meta used QAT on the first Llama models — Solid (80/100)
Meta did post-training QAT on original Llama — speaker's own source backs it.
BitNet is a subset of QAT, not the same thing — Opinion (50/100)
Calling BitNet 'maybe a subset' of QAT is his read, not a settled fact.
BitNet trains weights to only 1, 0, and -1 — Solid (75/100)
That matches the published BitNet paper description of ternary weights.
Prism ML models use post-training quantization like QAT — Dubious (45/100)
No public details confirm Prism ML's exact training method — this is speculation.
Google's 12B Gemma QAT is different from 1-bit/ternary models — OK (60/100)
True that Google's approach isn't full 1-bit, but he gives zero specifics on how it differs.
QAT models keep near-full intelligence with far fewer parameters — Dubious (40/100)
'Way, way, way less parameters' while staying 'just as smart' is the usual QAT marketing stretch.
This is similar to BitNet / 1-bit / ternary models that already showed promise — Dubious (45/100)
Calls it "similar but maybe not the same" — the connection stays vague.
See the full analysis with sources and timestamps →