Shadowing-Übung: Meta's new model wants "deep access" to your personal life... - Englisch Sprechen Lernen mit Video
Übungsmodus wählen
Jeder Modus öffnet einen Übungsraum im Vollbild.
57 Sätze
Wenn Sätze zu kurz oder zu lang sind, klicke auf Edit, um sie anzupassen.
1
On Monday, the company that spent the last year starving Llama, poaching researchers with nine-figure Zuck bucks, and locking its new models behind a Facebook auth page,
2
just released a free and open-source model under the Apache 2.0 license.
3
It's called Muse Glimmer, a name I believe was inspired by Zuck's fanfiction My Little Pony character,
4
and it's a 30 billion parameter agentic model that's small enough to run on your poverty spec PC.
5
That means you can now have an always-on mini Zuck agent living rent-free in your gaming PC, reading your email, managing your calendar,
6
and using your data to overthrow local elections in middle America.
7
And according to Meta, an agent like that needs deep access to personal context, which is a sentence they've been dreaming about writing since 2004,
8
except this time, the surveillance can run entirely on your own hardware.
9
In today's video, we'll find out what Muse Glitter Glimmer can actually do, learn how Meta shrunk it down with distillation,
10
quantization, and speculative decoding, and determine whether Zuck's open-source redemption arc is sincere, or if this is just what losing looks like right now.
11
It is August 12th, 2026, and you're watching The Code Report.
12
It wasn't long ago, when Meta was seen as the champion of the open weight class.
13
Their Llama models lopped up over a billion downloads and created an entire ecosystem of custom models built on top of them.
14
That is, until Llama 4 was released, and it became the biggest Facebook bust since Justin Timberlake partied with those underage girls.
15
In the release were two models, Scout and Maverick.
16
Initially, they benchmarked well, but that was only
17
because Meta submitted a secret juiced up version to LMArena that none of us normies could actually download.
18
And the models we could download had the intelligence of farm animals, which was fitting given their names.
19
And while this was happening, comrade Xi Jinping was dropping open models that were better, cheaper, and released in accordance with Mao Zedong's thought.
20
Then Zuck responded to this humiliation the only way a billionaire knows how.
21
He spent $14 billion on a 49% stake in scale AI just to acquire its CEO, Alexander Wang.
22
Then he went on a poaching spree for researchers at OpenAI and Google, and rebranded the whole operation as Meta Super Intelligence Labs.
23
The first thing his new dream team did was rebuild the stack from scratch and abandon open source entirely, shipping MuseSpark earlier this year as a closed API-only model.
24
That created a new problem, though, because no sane person would choose Muse Spark over something like Claude or Gemini.
25
So with Wall Street starting to ask questions about the $145 billion of CapEx being incinerated with nothing to show for it,
26
I assume Zuck flew to a secret compound in Kauai that the locals say he stole from them and formed a plan, which eventually led us here to the launch of Muse Glimmer.
27
It's a dense 30 billion parameter model distilled directly from Muse Spark, their big but closed model using logit distillation, which basically means they had the big model whisper its exact
28
probability distributions into the ear of Glimmer until it learned how to fake it.
29
And in Zuck's manifesto he made after the release, he described distillation as an important principle of how the open-source ecosystem works, which is ironic
30
because every American lab spent the last year accusing the Chinese of doing this exact same thing to their outputs.
31
So how good is it actually?
32
Well, according to the TMBBs, it clearly beats Gemma 4 and goes bar for bar with Quen 3.6.
33
But my favorite number is the prompt injection benchmark, where attacks against Glimmer succeeded 28% of the time, but they're listing that as a win because with Quen, it's 40%.
34
And the way they got it to run on consumer GPUs was interesting.
35
At full precision, the model would need over 55 gigs of memory, so their first move was quantization, where Meta compressed the weights down to about 4 bits,
36
which shrank it to just under 20 gigs.
37
And the second was speculative decoding, which is basically autocomplete for your autocomplete.
38
The way it works is a tiny model called D-Flash blurts out an entire block of tokens, then the big model reads the whole block in one pass, and throws out the bad guesses,
39
which they say led to a 3x speedup on a 5090.
40
But to be honest, the actual model is way less interesting than Zuck's manifesto I mentioned earlier.
41
In it, he argues that the real risk in AI isn't rogue superintelligence, but a small handful of companies owning it,
42
and that he doesn't understand why anyone who believes AI will end humanity would rush to build it, which is a reasonable point coming from a guy whose company
43
was just fined $567 million last week for being a public nuisance in New Mexico.
44
He also wants Frontier Labs to hand the US government mid-training checkpoints of unreleased models, and he announced a billion dollar fund for the towns willing to host his data centers,
45
probably so they'll stop causing all that ruckus.
46
So, is the redemption arc sincere?
47
Probably not, but an Apache license is an Apache license, and they can't put that genie back in the bottle.
48
Zuck and Wang both say open waits for MuseSpark 1.2 are coming soon, which would mean you could self-host the exact model behind Meta's own coding agent, which is pretty cool.
49
And if you want to test out Muse Glimmer for yourself, you can do it with the sponsor of today's video, OpenRouter.
50
It gives you a single API to access every LLM, so you only have one endpoint and one bill to worry about.
51
You can switch between models manually, or use their routers to pick one for you based on a combination of factors, like price and accuracy.
52
My startup still hasn't been able to raise a seed round, so I used Open Router to rewrite the entire codebase in Rust to attract more investors.
53
Because Rust is a notoriously easy language, I used Pareto Router to route every request to cheaper coding models that could still cargo build.
54
But I eventually ended up splurging for some fancier models for the UI,
55
the dedicated image and speech endpoints to generate a few especially attractive profiles to seed the community and get the matches flowing.
56
Open Router is easily the best tool I've seen for discovering new models without getting stuck in 25 different subscriptions, and you can try it out for free at the link below.
57
This has been The Code Report, thanks for watching, and I will see you in the next one.
App herunterladen
Alles, was du zum fließenden Sprechen brauchst
KI-AussprachebewertungJeder Satz bewertet
IPA-TrainingJeden Laut meistern
WortschatzBaue deinen Wortschatz auf
VokabelspielSpielend lernen
Gleicher Kanal
C11
Mind-bending new programming language for GPUs just dropped...
Daily ConversationUniversity Students
C11
PostgreSQL in 100 Seconds
Daily ConversationWorking Adults
C13
A $6.3 billion open-weight model just got embarrassed by the French...
Daily ConversationUniversity StudentsWorking Adults
C13
Every app on Earth is now open source... whether it likes it or not
Daily ConversationWorking Adults
C13
OpenAI’s lead engineer was secretly a spy...
Daily ConversationUniversity StudentsWorking Adults
B26
Intel chips can’t possibly be this bad… 100% crash rate?
American AccentBusiness EnglishUniversity StudentsWorking Adults
C13
Did Google just kickstart the intelligence explosion?
American AccentBusiness EnglishDaily ConversationUniversity Students
C12
Godot in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
C17
The most interesting "hack" in history...
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
Grok 4 pushes humanity closer to AGI… but there’s a problem
American AccentBusiness EnglishUniversity StudentsWorking Adults
C13
Array Map in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
C130
Docker in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
C110
The one OpenAI announcement that can actually make you money...
American AccentDaily ConversationUniversity StudentsWorking Adults
C16
How to NOT Fail a Technical Interview
American AccentInterview EnglishUniversity StudentsWorking Adults
B23
OpenAI’s new browser feels familiar…
American AccentBusiness EnglishUniversity StudentsWorking Adults
C112
10 weird OSS projects you need right now...
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
Firefox and Tor hit with 9.8 critical level exploit
American AccentBusiness EnglishUniversity StudentsWorking Adults
B23
5 open source tools that replaced my $320/mo AI stack...
American AccentBusiness EnglishUniversity StudentsWorking Adults
B210
25 crazy software bugs explained
American AccentDaily ConversationUniversity StudentsWorking Adults
C12
C++ in 100 Seconds
American AccentBusiness EnglishUniversity StudentsWorking Adults
FOR YOU
Empfohlenes Video
C12
43 Sätze
Meta’s Llama 4 is mindblowing… but did it cheat?
American AccentBusiness EnglishUniversity StudentsWorking Adults
B22
54 Sätze
Meta is pivoting again... everything you missed from Connect 2026
American AccentBusiness EnglishUniversity StudentsWorking Adults
C15
270 Sätze
Transformers, the tech behind LLMs | Deep Learning Chapter 5
American AccentBusiness EnglishUniversity StudentsWorking Adults
B21
247 Sätze
Why everyone is quitting social media
American AccentBusiness EnglishUniversity StudentsWorking Adults
C12
144 Sätze
WWDC26: Run local agentic AI on the Mac using MLX | Apple
Business EnglishUniversity StudentsWorking Adults
C11
19 Sätze
Mark Zuckerberg: “Alex, are you sure this is going to work?”
Daily ConversationUniversity StudentsWorking Adults
B21
190 Sätze
This 25-Year-Old Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan
Business EnglishUniversity StudentsWorking Adults
C18
102 Sätze
World Models explained in 10min..
Business EnglishUniversity StudentsWorking Adults
B23
165 Sätze
How I'd Learn AI From Scratch in 2026 (skip the useless 80%)
American AccentBusiness EnglishUniversity StudentsWorking Adults
C11
151 Sätze
New Facebook Ads Updates July 2026 What You Need to Know
Business EnglishWorking Adults
C13
176 Sätze
Is Amazon Too Big?
Business EnglishDaily ConversationUniversity Students
B22
155 Sätze
The perfect technology for creeps?
Business EnglishDaily ConversationUniversity Students
Sprachziele für diesen Clip
Dieser Clip hilft dir, komplexe Sachverhalte in Englisch zu verstehen und zu wiedergeben – ideal für Fortgeschrittene, die ihre Fähigkeit trainieren wollen, schnell gesprochene, sarkastische oder technische Sprache zu verarbeiten. Ein weiteres Ziel: Shadowspeaks-Übungen für flüssigeres Sprechen, insbesondere bei langen Sätzen und komplexen Argumentationssträngen.
Redewendungen für deinen Wortschatz
- to poach researchers – Forscher anwerben (mit hoher Bezahlung)
- behind a Facebook auth page – hinter einer Facebook-Login-Seite
- poverty spec PC – Computer mit schwacher Hardware
- to incinerate CapEx – Investitionen verbrennen (ohne Ertrag)
- logit distillation – Logit-Destillation (Technik zur Komprimierung von KI-Modellen)
- prompt injection benchmark – Test zur Abwehr von Prompt-Angriffen
Verbessere deine Schwächen
Der Clip eignet sich hervorragend, um die Rhythmusänderung in schneller, konversationaler Sprache zu üben. Viele Sätze sind lang und enthalten Sarkasmus (z. B. „an agent like that needs deep access to personal context“), was die richtige Betonung und Pausensetzung erfordert. Übe shadow speech, indem du den Sprecher direkt nachfolgst – achte darauf, die Ironie in der Stimme widerzugeben. Achte auch auf Wörter mit ungewöhnlicher Betonung wie „distillation“ (di-sti-LLA-tion) oder „quantization“ (quan-ti-ZA-tion). Diese Übung hilft dir, Englisch sprechen üben zu können, auch wenn der Text technisch oder sarkastisch ist.
Verwende die Redewendungen in eigenen Sätzen – z. B. „The company poached researchers from its competitors“ oder „His old laptop is a poverty spec PC“. So festigst du nicht nur den Wortschatz, sondern auch die Grammatik in komplexen Kontexten.
Grammatik in diesem Video
Die Strukturen, die der Sprecher am häufigsten verwendet, mit den genauen Worten aus dem Video:
| Struktur | Im Video |
|---|---|
| Relativsätze who / which + Satz – eine Zusatzinformation über eine Person oder Sache | context, which is · animals, which was · works, which is |
| Passiv be + Partizip Perfekt – wichtig ist, was geschieht, nicht wer es tut | was inspired · was seen · was released |
| Present Perfect have/has + Partizip Perfekt – eine vergangene Handlung, die jetzt noch wichtig ist | hasn't been · I've seen · has been |
Was ist die Shadowing-Technik?
Shadowing ist eine wissenschaftlich fundierte Sprachlerntechnik, die ursprünglich für die professionelle Dolmetscherausbildung entwickelt und durch den Polyglotten Dr. Alexander Arguelles populär gemacht wurde. Die Methode ist einfach aber wirkungsvoll: Du hörst englisches Audio von Muttersprachlern und wiederholst es sofort laut — wie ein Schatten, der dem Sprecher mit nur 1–2 Sekunden Verzögerung folgt. Anders als passives Hören oder Grammatikübungen zwingt Shadowing dein Gehirn und deine Mundmuskulatur, gleichzeitig echte Sprachmuster zu verarbeiten und zu reproduzieren. Studien zeigen, dass es Aussprachegenauigkeit, Intonation, Rhythmus, verbundene Sprache, Hörverständnis und Sprechflüssigkeit signifikant verbessert — was es zu einer der effektivsten Methoden für die IELTS Speaking-Vorbereitung und reale englische Kommunikation macht.