跟读练习: Meta's new model wants "deep access" to your personal life... - 通过视频学习英语口语

选择练习模式

每种模式都会打开全屏练习室。

57 句

如果句子过短或过长，请点击 Edit 进行调整。

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,

just released a free and open-source model under the Apache 2.0 license.

It's called Muse Glimmer, a name I believe was inspired by Zuck's fanfiction My Little Pony character,

and it's a 30 billion parameter agentic model that's small enough to run on your poverty spec PC.

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,

and using your data to overthrow local elections in middle America.

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,

except this time, the surveillance can run entirely on your own hardware.

In today's video, we'll find out what Muse Glitter Glimmer can actually do, learn how Meta shrunk it down with distillation,

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.

It is August 12th, 2026, and you're watching The Code Report.

It wasn't long ago, when Meta was seen as the champion of the open weight class.

Their Llama models lopped up over a billion downloads and created an entire ecosystem of custom models built on top of them.

That is, until Llama 4 was released, and it became the biggest Facebook bust since Justin Timberlake partied with those underage girls.

In the release were two models, Scout and Maverick.

Initially, they benchmarked well, but that was only

because Meta submitted a secret juiced up version to LMArena that none of us normies could actually download.

And the models we could download had the intelligence of farm animals, which was fitting given their names.

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.

Then Zuck responded to this humiliation the only way a billionaire knows how.

He spent $14 billion on a 49% stake in scale AI just to acquire its CEO, Alexander Wang.

Then he went on a poaching spree for researchers at OpenAI and Google, and rebranded the whole operation as Meta Super Intelligence Labs.

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.

That created a new problem, though, because no sane person would choose Muse Spark over something like Claude or Gemini.

So with Wall Street starting to ask questions about the $145 billion of CapEx being incinerated with nothing to show for it,

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.

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

probability distributions into the ear of Glimmer until it learned how to fake it.

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

because every American lab spent the last year accusing the Chinese of doing this exact same thing to their outputs.

So how good is it actually?

Well, according to the TMBBs, it clearly beats Gemma 4 and goes bar for bar with Quen 3.6.

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%.

And the way they got it to run on consumer GPUs was interesting.

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,

which shrank it to just under 20 gigs.

And the second was speculative decoding, which is basically autocomplete for your autocomplete.

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,

which they say led to a 3x speedup on a 5090.

But to be honest, the actual model is way less interesting than Zuck's manifesto I mentioned earlier.

In it, he argues that the real risk in AI isn't rogue superintelligence, but a small handful of companies owning it,

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

was just fined $567 million last week for being a public nuisance in New Mexico.

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,

probably so they'll stop causing all that ruckus.

So, is the redemption arc sincere?

Probably not, but an Apache license is an Apache license, and they can't put that genie back in the bottle.

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.

And if you want to test out Muse Glimmer for yourself, you can do it with the sponsor of today's video, OpenRouter.

It gives you a single API to access every LLM, so you only have one endpoint and one bill to worry about.

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.

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.

Because Rust is a notoriously easy language, I used Pareto Router to route every request to cheaper coding models that could still cargo build.

But I eventually ended up splurging for some fancier models for the UI,

the dedicated image and speech endpoints to generate a few especially attractive profiles to seed the community and get the matches flowing.

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.

This has been The Code Report, thanks for watching, and I will see you in the next one.

下载应用

流利口语所需的一切

AI 发音评分逐句打分

音标练习掌握每个音

词汇学习建立你的词库

词汇游戏边玩边学

扫码

Download on theApp Store Get it onGoogle Play

同频道

FOR YOU

本节课学习内容

本节课将通过分析一段关于Meta新AI模型的视频，帮助学习者练习听懂科技类话题的英语表达，掌握专业术语的发音和用法，同时提升影子跟读（shadowspeak）的技巧。视频内容涉及AI模型、科技公司动态等，语速中等偏快，适合训练听力理解和口语流畅度。

核心词汇与短语

open-source：开源的。例：The model is released under an open-source license.
parameter：参数。例：It's a 30 billion parameter model.
distillation：蒸馏（此处指AI模型压缩技术）。例：Meta used distillation to shrink the model.
quantization：量化（AI领域的压缩方法）。例：Quantization reduces the model's memory usage.
speculative decoding：推测解码。例：Speculative decoding speeds up the model's performance.

影子跟读练习技巧

影子跟读（shadow speak）是提升口语的有效方法，针对本视频的特点，可采用以下技巧：

1. 先慢速跟读：视频中包含较多专业术语，语速较快时容易混淆。建议先将视频速度调至0.75倍，逐句跟读，重点关注distillation、quantization等长单词的发音，确保每个音节都清晰。

2. 抓逻辑重音：视频讲解逻辑严密，如“Meta shrunk it down with distillation, quantization, and speculative decoding”一句中，三个技术名词是重点，跟读时需加重语气，突出信息层级。

3. 模仿语气变化：讲解者在提到“Zuck's fanfiction My Little Pony character”时带有调侃语气，跟读时需注意语调的轻微上扬，体现语境中的幽默感。

通过在影子跟读网站（shadowing site）反复练习，可逐渐适应科技类英语的表达节奏，提升听力和口语的同步反应能力。坚持使用shadowspeaks技巧，能有效改善口语流畅度和发音准确性。

视频中的语法

说话人最常用的结构，并附上视频中的原话：

结构	视频中的用法
定语从句 who / which + 从句 — 补充说明人或事物	context, which is · animals, which was · works, which is
被动语态 be + 过去分词 — 强调发生了什么，而不是谁做的	was inspired · was seen · was released
现在完成时 have/has + 过去分词 — 过去发生但与现在仍有关联的事	hasn't been · I've seen · has been

什么是跟读法？

跟读法 (Shadowing) 是一种有科学依据的语言学习技巧，最初开发用于专业口译员的培训，并由多语言者Alexander Arguelles博士普及。这个方法简单而强大：您在听英语母语原声的同时立即大声重复——就像是一个延迟1-2秒紧跟说话者的影子。与被动听力或语法练习不同，跟读法强迫您的大脑和口腔肌肉同时处理并模仿真实的讲话模式。研究表明它能显着提高发音准确性，语调，节奏，连读，听力理解和口语流利度——使其成为雅思口语备考和真实英语交流最有效的方法之一。

影子跟读法: 阅读完整分步指南 →

跟读练习: Meta's new model wants "deep access" to your personal life... - 通过视频学习英语口语

选择练习模式

下载应用

Did the AI bubble just pop?

PostgreSQL in 100 Seconds

React Native in 100 Seconds

Did Google just dethrone OpenAI and Anthropic?

A $6.3 billion open-weight model just got embarrassed by the French...

Every app on Earth is now open source... whether it likes it or not

OpenAI’s lead engineer was secretly a spy...

Mind-bending new programming language for GPUs just dropped...

China just dropped a new frontier model... and nobody can explain how

The last real programmers are dying out...

Intel chips can’t possibly be this bad… 100% crash rate?

The most interesting "hack" in history...

Docker in 100 Seconds

OpenAI’s new browser feels familiar…

Firefox and Tor hit with 9.8 critical level exploit

5 open source tools that replaced my $320/mo AI stack...

C++ in 100 Seconds

DHH has gone completely off the rails...

Rust in 100 Seconds

Linux in 100 Seconds

Meta’s Llama 4 is mindblowing… but did it cheat?

Meta is pivoting again... everything you missed from Connect 2026

Transformers, the tech behind LLMs | Deep Learning Chapter 5

Why everyone is quitting social media

WWDC26: Run local agentic AI on the Mac using MLX | Apple

Mark Zuckerberg: “Alex, are you sure this is going to work?”

This 25-Year-Old Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan

How I'd Learn AI From Scratch in 2026 (skip the useless 80%)

The perfect technology for creeps?

Codex Is Winning… So Why Am I Still Paying for Claude?

The AI Bubble Survives on $1.65 Trillion in Hidden Debt

The Rise Of Open-Source Software

本节课学习内容

核心词汇与短语

影子跟读练习技巧

视频中的语法

什么是跟读法？