RussGoPackGo

Green Bay Packers 24/7/365!!!!!

[Green.Bay.Packers @GBPTrevor]

As a #Packers fan, I will never forget this moment 🥹

[Zach Jacobson @zacobson]

he’s the best to ever throw that ball

[淘沙者(TheSandPicker) @Etudecn]

这段马斯克 2003 年在斯坦福的 45 分钟闭门演讲，真的有点东西。
不是那种空话连篇的创业鸡汤。
而是他亲自把“怎么从 0 搞出一家公司”一层层拆给你看。
更狠的是，
他没在讲概念，直接把自己当时手里那三家公司，到底怎么活下来的，整个过程都摊开了。
这种内容放到今天看都不过时。

很多人听完的反应都很统一：
这 45 分钟，顶得上你翻十本创业书。
那种真正能用的干货，往往就藏在这种老素材里，不吵不闹，但句句都是实战。

[Lord Bebo @MyLordBebo]

🇺🇸🚨‼️ SUPER CUT: Secretary of War Hegseth reads a US Air Force version of the fictional Bible verse from Pulp Fiction.

Whether this is awesome or stupid is up to you.

[Eric Daugherty @EricLDaugh]

🚨 HOLY SMOKES. Every single Democrat in the Minnesota House Rules Committee votes to BLOCK investigations into fraud under Tim Walz

With a straight face, Democrats endorsed $9B+ in stolen tax dollars

They're panicked and don't want anyone finding out how this was allowed to happen.

REMOVE WALZ FROM OFFICE and start the criminal proceedings for complicity!

[Felipe Demartini @namcios]

Dois engenheiros da Anthropic acabaram de mudar a forma como devs pensam sobre IA.

Barry Zhang e Mahesh Murag subiram no palco do AI Engineer Code Summit e disseram uma frase que incomodou muita gente:

"Parem de construir agentes. Construam Skills."

Em 16 minutos eles provam que a indústria inteira está resolvendo o problema errado.

Aqui está o que a maioria não entendeu:

→ Skills são pastas. Literalmente pastas com arquivos markdown.
→ Elas ensinam ao Claude o SEU fluxo de trabalho, a SUA expertise, o SEU domínio.
→ Um único agente genérico + biblioteca de Skills específicas supera dezenas de agentes especializados.
→ Fortune 100s já estão deployando Skills em escala pra ensinar agentes sobre processos internos.
→ Times de produtividade com 10.000+ devs usam Skills pra padronizar como código é escrito.

A analogia que eles usaram é perfeita:

Quem você quer fazendo seu imposto de renda? O gênio com QI 300 que nunca viu legislação tributária, ou o contador experiente que faz isso há 20 anos?

Inteligência sem expertise é entretenimento.
Expertise empacotada é produtividade.

O que mudou: a Anthropic parou de tentar criar agentes diferentes pra cada domínio.

Perceberam que com Claude Code, o padrão é sempre o mesmo. Um modelo acoplado a um runtime com filesystem.

A diferença entre um agente medíocre e um extraordinário não é o modelo. É o conhecimento de domínio que você alimenta.

Skills resolvem isso com progressive disclosure. O agente só carrega o nome e descrição da skill. Quando relevante, puxa o SKILL.md. Quando precisa de mais, navega os arquivos de referência. Zero desperdício de contexto.

Isso não é uma feature. É uma mudança de paradigma.

Quem entender isso agora vai operar em outro nível daqui a 90 dias.

Quem ignorar vai continuar escrevendo prompts de mil palavras toda vez que abrir o chat. E ainda explicar de novo e de novo o que “realmente” quer.

[Roan @RohOnChain]

This 2 hour Stanford lecture will teach you more about how LLMs like ChatGPT & Claude are built than most people working at top AI companies learn in their entire careers.

Bookmark this & give 2 hours today, no matter what. It'll be the most productive thing you do this week.

[Dustin Grage @GrageDustin]

🚨 NEW DOCUMENTARY: Tim Walz is Minnesota Mao.

Chilling.

[Nick Sortor @nicksortor]

🚨 WATCH: A LIFELONG DEMOCRAT just testified that Tim Walz's admin TARGETED and ATTACKED her for exposing fraud in Minnesota

Tim Walz should be in JAIL.

Faye Bernstein, a state employee for 20 years, said she first started noticing fraud 7 years ago, and sounded the alarm internally, so the admin began SMEARING her.

"The retaliation in the smear campaign is strong. I was called everything from incompetent to racist. People who speak about fraud at state agencies, the first thing that leadership goes to is "RACISM"

"It is career-killing, obviously. Who is going to hire someone who is considered racist?!"

God bless this brave woman for her bravery 🙏🏻

[Jaynit @jaynitx]

MrBeast: "If you knew what I knew, you could get 10 million subscribers in six months"

"Your videos suck. You think your videos are good, but they suck. They just do. And the sooner you learn how to make good, great videos that people actually want to watch, the sooner you'll get views."

MrBeast shares his early reality:

"When I was 14, I thought my videos were the best in the world. They weren't, they were terrible. To be successful, you kind of have to have a little bit of that ego where you think your content's great. But also, if you have sub-1,000 subscribers, there's a good probability your videos just suck. They just do."

He explains what to do about it:

"You need to make hundreds of videos. Improve something every time. And just get to the point where they don't suck. When you make good content, you'll blow up. It's not the algorithm. It's not anything. Most people who are in my position just made terrible videos, and that's okay. Because you've got to make a bunch of videos and improve over time to be great."

MrBeast uses an analogy:

"You don't just pick up a baseball and become an MLB-level athlete within a year. It takes many, many, many years. YouTube's kind of the same way."

On analysis paralysis:

"A lot of people get analysis paralysis. They'll sit there and plan their first video for three months. If you have zero videos on your channel, your first video is not gonna get views. Period. Your first 10 are not gonna get views. I can very confidently say that. So stop sitting there and thinking for months and months on end. Just get to work and start uploading."

He gives the formula:

"All you need to do is make 100 videos and improve something every time. Do that, and then on your 101st video, we'll start talking. Maybe you can get some views. But your first 100 are gonna suck."

How to improve something each time:

"The second video: put more effort into the script. The third one: learn a new editing trick. The fourth one: figure out a way to have better inflections in your voice. The fifth one: study a new thumbnail tip and implement it. The sixth one: figure out a new title. There's infinite ways. The coloring, the frame rate, the editing, the filming, the production, the jokes, the pacing, every little thing can be improved. There's literally no such thing as a perfect video."

On the algorithm:

"What YouTube wants is for people to click on a video and watch it. That's what it is at its core. By studying the algorithm, you'll learn that you're more studying human psychology. What do humans want to watch?"

MrBeast shares a simple reframe:

"Anytime you say the word 'algorithm,' just replace it with 'audience' and it works perfectly. 'The algorithm didn't like that video?' No, the audience didn't like that video. Literally, that's it. If people are clicking and watching, it gets promoted more. The algorithm just reflects what the people want."

On titles:

"Short, simple, and just so freaking interesting that you have to click. If someone reads it, are they like, do they have to watch it? Is it just so intrinsically interesting that it's gonna haunt them if they don't click?"

He adds nuance:

"Keep it below 50 characters. Above 50 characters, on certain devices it goes dot, dot, dot, and that's the worst thing because then people don't even know what they're clicking on."

MrBeast shares the extremity principle:

"The more extreme the opinion, typically the higher the click-through rate. 'Fiji water sucks', that'd do fine. But 'Fiji water is the worst water I've ever drank in my life', way more extreme, would do way better. But then you have to deliver. The more extreme you are, the more extreme you have to be in the video."

On the first 5 seconds:

"Before you film a video, what is the thumbnail? What is the title? Then what's the first 5 seconds? Then what's the first 30 seconds?"

He explains why autoplay changed everything:

"On YouTube now, videos automatically play. So many people don't even see the thumbnail because it autoplays so quickly. The thumbnail is irrelevant for them. I have to visually convince you to click on the video in the first 5 seconds. Before, the hook was important because you had to convince people to watch. Now you have to convince people to click and watch at the same time, with the first 5 seconds."

On matching expectations:

"Your title and thumbnail set expectations. At the very beginning of the video, to minimize drop-off, you want to assure them that those expectations are being met. If you click on a video called 'Tether is a scam' and at the very beginning, he starts talking about literally anything else, you're like, 'Oh, this is BS. This isn't what I clicked on.' But if at the very start you go, 'Tether is a scam and I'm gonna teach you why,' then it's like, okay, you match the expectations. Then you want to exceed them."

He emphasizes the importance:

"The thing people undervalue the most is literally the first 10 seconds of the video. That 15% difference in viewership between losing 35% of viewers in the first 30 seconds versus losing 20%, that really does make the difference between 2 million views and 10 million views. You just had a more strategic intro that hooked them."

On removing dull moments:

"You basically want to remove every dull moment. Find the 10 most critical people you know, make them watch the video, and just roast it. If I talk to a camera for 10 seconds without a cut, a lot of people will get bored. Having a B-cam and C-cam three seconds in, cutting to a different angle, now it's more interesting even though it's essentially the same thing."

On keeping viewers watching:

"Give them why they clicked. Tell them why they should watch. Then just stick on topic. That right there isn't even super complex, but I would already put you in the upper echelon of YouTube. A lot of people drag it out. It's like, 'I'm going to eat $100 ice cream, but first...' and then it's them birthday shopping for their mom. That's not why I came here."

On quality over quantity:

"It's much easier to get 5 million views on one video than 50,000 views on 100 videos. A lot of small YouTubers just post videos that aren't bad but aren't great, and none of them ever pop off, so they never get an audience. It might be better to upload half or a third or even a fifth of the videos, but make the videos you upload so freaking good that the algorithm has to promote it."

He warns against the consistency trap:

"When you set a consistent schedule and you're constantly having to upload videos that aren't as good as you'd like because you gotta hit 'Oh, this Monday I said I'd upload', that's a dangerous trap. The viewers notice the quality isn't as good and it makes them less likely to watch. I think it hurts your longevity."

On the real metric that matters:

"A big thing that everyone underestimates, what was your experience with your last video? If people loved the last video of yours that they watched, they're more likely to watch your next one. When people watch your video, you don't want them to go, 'Okay, that was good, but that's enough of you for the day.' What you want is them to go, 'Holy crap, that was crazy! Oh my god, what's that?' and they watch 10 videos. That's how you get high view counts. People watch 10 videos, not one."

On thumbnails:

"You want it to be simple. When they're scrolling, you want them to instantly understand what you're conveying and feel some type of emotion. Make it so interesting, or spike their curiosity so much, that if they don't click it, they'll wonder before they go to bed what happened?"

He gives an example:

"If you uploaded 'I rode a skateboard with 1,000 other people on it', and people are falling off the side, it's about to go off a big ramp if you don't click that, you're gonna be so curious. Later in the day, when you're daydreaming, you'll think, 'What happened to those 1,000 people on that skateboard?' That's the mindset you should have when making thumbnails."

On knowledge being the only barrier:

"It's all knowledge. It really is. I could start a new channel tomorrow without using my face or my voice, without ever promoting it, and in six months have 20 million subscribers. I just could. It's purely knowledge. If you knew what I knew, you could get 10 million subscribers no matter where you are right now within six months."

He addresses the skeptics:

"90% of the people watching don't agree with that. Everyone has excuses. 'Nah, YouTube just doesn't work like that, Jimmy.' But I mentor a lot of people. I see it all the time. It is possible. It is simply knowledge. The second you accept that it is knowledge and you start your journey of learning figuring out what makes a good video, what does my audience want, how can I elevate and then you take that knowledge and just assume 'I will never understand what the perfect video is' and every single day be devoted to learning and improving as much as possible there you go."

On money not being the barrier:

"There are tons of viral ideas that don't require money. It does not require money to go viral. One of my most-viewed videos was spending 24 hours in a desert, we just grabbed a tent and some stuff and went to the desert. It got 60-70 million views. People say, 'I could be MrBeast if I had money.' A, I didn't start off with money; I was poor, I had no money. It took me seven years just to buy a camera saving up from YouTube. And B, some of our most-viewed videos literally anyone can do."

On why no one will outwork him:

"No one's ever gonna do what I do better than me. It's just not humanly possible. I reinvest every penny I make. I work every hour I'm awake. I devote every atom in my brain to solving this. I hire the best people on the planet. I've been doing this for 14 years. And I think in decades, not years. I'm gonna be doing this for another 20-30 years. If I thought someone was doing better than me, I'd just start sleeping less so I could work even more."

But he doesn't recommend it:

"I don't have a life. I don't have work-life balance. My personality, my soul, my being is making the best videos possible. That is why I exist on this planet. And I don't recommend it. You should have work-life balance. You should not devote your entire life to this one thing. I have a mental breakdown every other week because I push myself so hard. I don't recommend it."

The only question that matters:

"Subscribers don't matter. Views don't matter. I mean, they do. But everything you want as a creator comes from making the best videos possible and thumbnails. The video part's the hard part. Ask: 'How can I make my videos better?' Do that every single day for years. And then you'll probably get views."

[GREG ISENBERG @gregisenberg]

In case you missed it...

This 58 min video is the clearest introduction to AI agents, agent skills, md files, building AI employees on the internet and it's 100% free

[Corey Ganim @coreyganim]

The clearest explanation of Claude Managed Agents you'll find.

Everyone's talking about it. Nobody's explaining it well.

In 12 minutes we cover:

- What it actually is (platform as a service for AI agents)
- Who it's for and who should avoid it (4 personas)
- Live console walkthrough (sessions, analytics, costs)
- Real cost breakdown ($2.58 to fulfill a $1,000 service)

We also built a free Google Doc that deploys your first managed agent when you hand it to Claude Code.

You can grab it in the podcast show notes (Build With AI podcast) or YouTube description (video link below).

[0xMarioNawfal @RoundtableSpace]

LEARN HOW TO SET UP AND RUN CLAUDE CODE + OBSIDIAN IN ONLY 22 MINUTES

[Avid @Av1dlive]

This 16-minute talk by two Anthropic engineers who built Claude Skills will teach

you more about building them right than most developers figure out on their own in months.

Bookmark this & watch, no matter what.

Then read the guide below by @eng_khairallah1

[Movez @0xMovez]

This 1 hour podcast with the Head of Claude Code will teach you more about vibe-coding than 100 paid courses.

Bookmark this & give it 1 hour today. It's the best video about AI you'll watch this week. Then read the article below.