Manuel Magalhães

“Passa a bola caralho” Adoro ver o meu PL a querer a bola. PL é isto. E gosto de o ver com fome de marcar mais, de querer mais mesmo ganhando 0-2. Pavlidis já sabe o essencial do Português ahahah

926

53K

Manuel Magalhães @MM221923

6 days ago

@SLBoni_ Ouvir Marco Silva no fim do jogo é música para os nossos ouvidos. Que treinador incrível temos este ano.

401

MM221923 retweeted

Central Benfica

@central_benfica

8 days ago

Vídeo simplesmente incrível. São estes os valores que definem o Benfica enquanto clube e instituição❤️🦅

182

MM221923 retweeted

Scof @_scof1904

9 days ago

segue a banda! 🤡

572

267

539K

Manuel Magalhães @MM221923

13 days ago

@jbizarro O que faz aquele Paulo Costa na tribuna ao lado do Rui Costa?

MM221923 retweeted

Bag

@Bag_Investing

13 days ago

$META I have been thinking (again) about Meta and its AI strategy, especially after the launch of Muse. The initial reaction from some people was almost predictable. Why would anyone want a personal AI agent from Meta, the company that has spent years dealing with criticism over how it collects and uses people's data? Others simply made fun of the idea of Zuckerberg building an AI assistant for people's everyday lives. I think both reactions miss the more important part of what Meta is trying to build. Muse is not necessarily interesting because it is already the best AI product available. It is interesting because Meta is trying to make AI an everyday product for billions of people who already use its platforms. Alexandr Wang has been highlighting how competitive Muse Spark is relative to other frontier models while being significantly cheaper to run. That matters because Meta does not necessarily need to build the single best model in the world. If it can build a model that is good enough, cheap enough and reliable enough, it can deploy it across an enormous consumer base. At Meta's scale, the economics of inference matter just as much as benchmark performance. The bigger opportunity is distribution. OpenAI and Anthropic have built extremely valuable AI businesses, but they still need to convince consumers and businesses to use their products. Meta already has 3.6 billion daily active people across its family of apps. If personal AI agents become an important way of interacting with the internet, Meta can put that agent directly into products people already use instead of asking them to change their habits. Imagine that your AI assistant lives inside WhatsApp and Instagram. You could ask it to organize a trip, find a product, compare prices, communicate with a business, help plan an event or complete other everyday tasks. Meta is trying to move AI from something people deliberately open and ask questions to something that becomes part of their daily digital life. If that happens, the agent could become the interface between consumers and the internet. That would create opportunities in subscriptions, commerce, business messaging and other services, but I think the bigger opportunity is actually Meta's existing business. AI can make Meta's recommendation systems better, help advertisers create better campaigns, improve targeting and measurement, make creators more productive and increase the value advertisers receive from the platform. Those benefits do not necessarily show up as "AI revenue." They can simply appear as better economics across the existing advertising business. This is important when thinking about Meta's enormous AI spending. This is where Zuckerberg's comments about compute are important. Meta is receiving offers to sell computing capacity at a premium to what it paid for it. But Zuckerberg said he believes there can be significantly more value in building intelligence on top of that compute than simply selling the compute itself. He described the decision as a trade-off between monetizing an asset today and using it to develop future assets, and said he would not want Meta to simply sell all of its compute for a short-term profit. I like that way of thinking. If Meta believed the highest-return use of its infrastructure was simply selling compute today, then selling it would make sense. But if the company believes that the real value comes from using that infrastructure to build intelligence that improves advertising, powers agents, enables business products and creates new consumer experiences, then sacrificing some short-term revenue could be the right decision. The market has spent enormous amounts of money valuing companies such as OpenAI and Anthropic based on the future businesses they might build. Meta is already generating enormous amounts of revenue and cash from an existing business, while simultaneously investing heavily in the technology that could become the next major computing platform. That creates an interesting setup for investors. We do not have to believe that every dollar of AI spending will generate a great return. We only need to believe that some of the opportunities Meta is building become large enough to justify the investment. And Meta has several ways to win, has I have said before, AI can strengthen the advertising business, create subscription products, power business agents, improve commerce, support developers and eventually become a major part of Meta's hardware strategy. The company's AI glasses are particularly interesting because they could eventually give Meta an interface between AI and the physical world, while WhatsApp and Instagram give it an interface with the digital world. This is also why I don't think the right way to value Meta is to treat all of its AI spending as money disappearing from the business. Some of that spending is clearly an expense in the short term, but the ultimate question is what asset Meta is creating with it. The important thing about Meta is that you are not buying an unprofitable AI company and hoping it eventually becomes enormous. You are buying one of the world's largest advertising businesses, with billions of users and substantial existing cash generation, while also getting exposure to what could become a very large AI opportunity. That distinction matters. Meta does not need AI to justify the entire value of the company. The existing advertising business is already worth a substantial amount on its own. AI is an additional opportunity sitting on top of that business. This means that, unlike an AI company whose valuation depends almost entirely on future AI revenue, Meta can potentially fund the AI buildout from a business that already generates enormous amounts of revenue. If advertising continues to grow while AI improves monetization and creates new products, investors are getting exposure to both businesses. The stock is already recognized as one of the highest-quality businesses in technology, and the market knows that AI could become a major opportunity. The question is whether today's valuation fully reflects the potential return on the infrastructure Meta is building. I think it does not. Of course, Meta is not risk-free. The spending could prove excessive, AI competition could remain intense, agents could take much longer to become useful than expected, and the returns on the infrastructure could disappoint. The concerns around Meta's handling of user data also aren't going away simply because the company has a compelling AI strategy. If anything, a personal AI agent will make privacy and trust even more important. The market seems very comfortable paying enormous valuations for companies whose future depends almost entirely on AI becoming a huge business. With Meta, much of the value is already supported by a massive existing business, while the company is investing aggressively to secure a position in the next generation of computing. I don't need Muse to become the best AI agent in the world. I don't need Meta to win every AI benchmark. I don't even need every part of the AI strategy to work. I want Meta to build a good enough AI system, distribute it to billions of people, improve its existing businesses and find new ways to monetize the infrastructure it is building. If AI becomes as important to computing as many people expect, I would rather own a company that already has billions of users, enormous cash generation and the ability to distribute AI at scale than one that has to build that distribution from scratch. The market is looking at Meta's AI spending and asking how much it will cost. I think investors should be asking what those billions of dollars can become over the next five to ten years. If Meta is right, we may eventually look back at this period not as the time when Meta spent too much on AI, but as the period when it built the infrastructure and distribution needed to participate in the next major computing platform. That is why I continue to like Meta.

Bag_Investing's tweet photo. $META

I have been thinking (again) about Meta and its AI strategy, especially after the launch of Muse.
The initial reaction from some people was almost predictable. Why would anyone want a personal AI agent from Meta, the company that has spent years dealing with criticism over how it collects and uses people's data? Others simply made fun of the idea of Zuckerberg building an AI assistant for people's everyday lives. I think both reactions miss the more important part of what Meta is trying to build.

Muse is not necessarily interesting because it is already the best AI product available. It is interesting because Meta is trying to make AI an everyday product for billions of people who already use its platforms.

Alexandr Wang has been highlighting how competitive Muse Spark is relative to other frontier models while being significantly cheaper to run. That matters because Meta does not necessarily need to build the single best model in the world. If it can build a model that is good enough, cheap enough and reliable enough, it can deploy it across an enormous consumer base. At Meta's scale, the economics of inference matter just as much as benchmark performance.

The bigger opportunity is distribution. OpenAI and Anthropic have built extremely valuable AI businesses, but they still need to convince consumers and businesses to use their products. Meta already has 3.6 billion daily active people across its family of apps. If personal AI agents become an important way of interacting with the internet, Meta can put that agent directly into products people already use instead of asking them to change their habits.

Imagine that your AI assistant lives inside WhatsApp and Instagram. You could ask it to organize a trip, find a product, compare prices, communicate with a business, help plan an event or complete other everyday tasks. Meta is trying to move AI from something people deliberately open and ask questions to something that becomes part of their daily digital life.

If that happens, the agent could become the interface between consumers and the internet. That would create opportunities in subscriptions, commerce, business messaging and other services, but I think the bigger opportunity is actually Meta's existing business.

AI can make Meta's recommendation systems better, help advertisers create better campaigns, improve targeting and measurement, make creators more productive and increase the value advertisers receive from the platform. Those benefits do not necessarily show up as "AI revenue." They can simply appear as better economics across the existing advertising business.

This is important when thinking about Meta's enormous AI spending.

This is where Zuckerberg's comments about compute are important.

Meta is receiving offers to sell computing capacity at a premium to what it paid for it. But Zuckerberg said he believes there can be significantly more value in building intelligence on top of that compute than simply selling the compute itself. He described the decision as a trade-off between monetizing an asset today and using it to develop future assets, and said he would not want Meta to simply sell all of its compute for a short-term profit.

I like that way of thinking.

If Meta believed the highest-return use of its infrastructure was simply selling compute today, then selling it would make sense. But if the company believes that the real value comes from using that infrastructure to build intelligence that improves advertising, powers agents, enables business products and creates new consumer experiences, then sacrificing some short-term revenue could be the right decision.

The market has spent enormous amounts of money valuing companies such as OpenAI and Anthropic based on the future businesses they might build. Meta is already generating enormous amounts of revenue and cash from an existing business, while simultaneously investing heavily in the technology that could become the next major computing platform.

That creates an interesting setup for investors.

We do not have to believe that every dollar of AI spending will generate a great return. We only need to believe that some of the opportunities Meta is building become large enough to justify the investment.

And Meta has several ways to win, has I have said before, AI can strengthen the advertising business, create subscription products, power business agents, improve commerce, support developers and eventually become a major part of Meta's hardware strategy. The company's AI glasses are particularly interesting because they could eventually give Meta an interface between AI and the physical world, while WhatsApp and Instagram give it an interface with the digital world.

This is also why I don't think the right way to value Meta is to treat all of its AI spending as money disappearing from the business. Some of that spending is clearly an expense in the short term, but the ultimate question is what asset Meta is creating with it.

The important thing about Meta is that you are not buying an unprofitable AI company and hoping it eventually becomes enormous. You are buying one of the world's largest advertising businesses, with billions of users and substantial existing cash generation, while also getting exposure to what could become a very large AI opportunity.

That distinction matters.

Meta does not need AI to justify the entire value of the company. The existing advertising business is already worth a substantial amount on its own. AI is an additional opportunity sitting on top of that business.

This means that, unlike an AI company whose valuation depends almost entirely on future AI revenue, Meta can potentially fund the AI buildout from a business that already generates enormous amounts of revenue. If advertising continues to grow while AI improves monetization and creates new products, investors are getting exposure to both businesses.

The stock is already recognized as one of the highest-quality businesses in technology, and the market knows that AI could become a major opportunity. The question is whether today's valuation fully reflects the potential return on the infrastructure Meta is building.

I think it does not.

Of course, Meta is not risk-free. The spending could prove excessive, AI competition could remain intense, agents could take much longer to become useful than expected, and the returns on the infrastructure could disappoint. The concerns around Meta's handling of user data also aren't going away simply because the company has a compelling AI strategy. If anything, a personal AI agent will make privacy and trust even more important.

The market seems very comfortable paying enormous valuations for companies whose future depends almost entirely on AI becoming a huge business. With Meta, much of the value is already supported by a massive existing business, while the company is investing aggressively to secure a position in the next generation of computing.

I don't need Muse to become the best AI agent in the world. I don't need Meta to win every AI benchmark. I don't even need every part of the AI strategy to work.

I want Meta to build a good enough AI system, distribute it to billions of people, improve its existing businesses and find new ways to monetize the infrastructure it is building.

If AI becomes as important to computing as many people expect, I would rather own a company that already has billions of users, enormous cash generation and the ability to distribute AI at scale than one that has to build that distribution from scratch.

The market is looking at Meta's AI spending and asking how much it will cost.

I think investors should be asking what those billions of dollars can become over the next five to ten years.

If Meta is right, we may eventually look back at this period not as the time when Meta spent too much on AI, but as the period when it built the infrastructure and distribution needed to participate in the next major computing platform.

That is why I continue to like Meta.

Manuel Magalhães @MM221923

15 days ago

@___Jota___ Planalto reserva branco, Reguengos reserva branco e Dona Graça

195

Manuel Magalhães @MM221923

22 days ago

@SICNoticias Quando são estrangeiros não referem nada. Que acabe a Sic

292

MM221923 retweeted

Pablito_Aimar @PablitoAimar6

26 days ago

@DTransferencias @Now_Canal_ Pedro Sousa se fosse trabalhar, ganhávamos todos mais

MM221923 retweeted

Vitor Azevedo @VFAzevedo

29 days ago

Marcos, cliente que estava numa loja do LIDL deteve até a chegada da polícia, um indivíduo que tentou agredir uma funcionária e roubar Abaixo está o comentário de um jumento atrasado mental,que já foi deputado pelo inútil @pcp_pt Se o vires chama a polícia, ele é perigoso

VFAzevedo's tweet photo. Marcos, cliente que estava numa loja do LIDL deteve até a chegada da polícia, um indivíduo que tentou agredir uma funcionária e roubar
Abaixo está o comentário de um jumento atrasado mental,que já foi deputado pelo inútil @pcp_pt
Se o vires chama a polícia, ele é perigoso https://t.co/OYr6whuanM

885

57K

MM221923 retweeted

Alexandre Guerreiro

@ATGuerreiro

28 days ago

Um homem vítima do "gang do Rolex" perseguiu e atropelou mortalmente um dos ladrões. Vai responder criminalmente pela morte. E eu ajudo a pagar a indemnização.

471

26K