Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

$APP is down nearly 20% after a temporary slowdown in model improvement prevented another major step-up in advertiser spending despite an otherwise exceptional quarter. AppLovin’s platform becomes more valuable as its AI improves ad targeting which lifts advertiser returns and unlocks larger budgets but its research team didn't deliver... show more

Shay Boloor

444,845 subscribers

81,850 просмотров • 28 дней назад •via X (Twitter)

Финансы Новости и политика

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

$DDOG had its strongest sequential growth since 2022 but fell nearly 20% after its largest AI customer renewed a nine-figure contract across 17 products while planning to reduce usage in Q3. That proves Datadog remains deeply embedded but also exposes the downside of consumption pricing because one customer can stay committed while still materially slowing companywide growth. The long-term thesis remains intact as cloud migration, AI inference, agents and security create more complexity to monitor but Datadog must now prove this is isolated optimization rather than a broader frontier AI trend.

$DDOG had its strongest sequential growth since 2022 but fell nearly 20% after its largest AI customer renewed a nine-figure contract across 17 products while planning to reduce usage in Q3. That proves Datadog remains deeply embedded but also exposes the downside of consumption pricing because one customer can stay committed while still materially slowing companywide growth. The long-term thesis remains intact as cloud migration, AI inference, agents and security create more complexity to monitor but Datadog must now prove this is isolated optimization rather than a broader frontier AI trend.

Shay Boloor

66,964 просмотров • 28 дней назад

$APP's pivot to e-commerce is a very intriguing strategy. The market perceives the entry of non-gaming advertisers as a potential friction point that could crowd out core gaming clients or inflate pricing, but this ignores the massive asymmetry between AppLovin’s reach of over one billion daily active users and its historically thin roster of active advertisers. This imbalance has meant that the vast amount of inventory were previously under-monetized, as the algorithm could only serve gaming ads to users who had no intent to install new games. By layering in e-commerce demand, the platform effectively monetizes this wasted inventory, driving up overall yield and floor prices without cannibalizing the high-intent impressions reserved for gaming clients. This creates a margin-accretive dynamic where the same unit of supply generates significantly higher revenue per user solely through better demand matching. This efficiency gain feeds directly into a data-driven moat that becomes increasingly difficult for competitors to replicate. The flywheel is the introduction of transactional e-commerce data that radically improves the Axon AI model's predictive capabilities for all participants. Unlike app install data, which is binary and relatively sparse, e-commerce purchase data provides immediate high-fidelity signals about user intent and purchasing power. As the model ingests this new layer of behavioral data, its ability to predict conversion improves universally, meaning that gaming advertisers actually benefit from the presence of e-commerce bids through sharper targeting and higher return on ad spend. The rapid 50% week over week growth in the self-service pilot is a great preliminary validation that this automated demand engine is functional. This signals that AppLovin can scale this new vertical with software operating leverage. The requirement for high-production video ads has left out the long tail of millions of small business advertisers who dominate platforms like $META. The launch of generative AI creative tools targets this specific bottleneck, commoditizing the production of high-performing video assets and allowing AppLovin to unlock global SMB demand instantly. If successful, this creates a self-reinforcing liquidity cycle where increased advertiser density leads to better data, which drives superior model performance, which in turn attracts more diverse advertisers. This helps decouple the company's growth trajectory from the cyclicality of the mobile gaming market. Really interesting biz and great CEO.

$APP's pivot to e-commerce is a very intriguing strategy. The market perceives the entry of non-gaming advertisers as a potential friction point that could crowd out core gaming clients or inflate pricing, but this ignores the massive asymmetry between AppLovin’s reach of over one billion daily active users and its historically thin roster of active advertisers. This imbalance has meant that the vast amount of inventory were previously under-monetized, as the algorithm could only serve gaming ads to users who had no intent to install new games. By layering in e-commerce demand, the platform effectively monetizes this wasted inventory, driving up overall yield and floor prices without cannibalizing the high-intent impressions reserved for gaming clients. This creates a margin-accretive dynamic where the same unit of supply generates significantly higher revenue per user solely through better demand matching. This efficiency gain feeds directly into a data-driven moat that becomes increasingly difficult for competitors to replicate. The flywheel is the introduction of transactional e-commerce data that radically improves the Axon AI model's predictive capabilities for all participants. Unlike app install data, which is binary and relatively sparse, e-commerce purchase data provides immediate high-fidelity signals about user intent and purchasing power. As the model ingests this new layer of behavioral data, its ability to predict conversion improves universally, meaning that gaming advertisers actually benefit from the presence of e-commerce bids through sharper targeting and higher return on ad spend. The rapid 50% week over week growth in the self-service pilot is a great preliminary validation that this automated demand engine is functional. This signals that AppLovin can scale this new vertical with software operating leverage. The requirement for high-production video ads has left out the long tail of millions of small business advertisers who dominate platforms like $META. The launch of generative AI creative tools targets this specific bottleneck, commoditizing the production of high-performing video assets and allowing AppLovin to unlock global SMB demand instantly. If successful, this creates a self-reinforcing liquidity cycle where increased advertiser density leads to better data, which drives superior model performance, which in turn attracts more diverse advertisers. This helps decouple the company's growth trajectory from the cyclicality of the mobile gaming market. Really interesting biz and great CEO.

CapexAndChill

29,897 просмотров • 7 месяцев назад

AppLovin CEO is trying to counter Anthropic’s CEO claims that SaaS is DEAD and never coming back. $APP is not an ad network. It is not a gaming company. It is an arbitrage engine with 400 engineers, $1.3 billion in cash flow per quarter, and a capital allocation track record that belongs in a business school curriculum. Start with the gaming studios. From 2018 to 2023, AppLovin acquired over 15 mobile gaming studios and 1,500 people. Nobody understood why. The answer was data. Third party advertisers would not share their conversion and ROAS data with AppLovin's model. So AppLovin became a first party advertiser — running its own games, generating its own behavioral data, feeding that data into Axon, its deep learning ad model. The studios were not a business. They were a training set. The moment Axon 2.0 launched in April 2023 and proved so effective that the entire gaming industry had to plug in and share their data to access the returns, the studios had served their purpose. AppLovin sold the entire portfolio to Tripledot and moved on. They used the asset to build the moat and immediately dumped the asset. That is not a pivot. That is a premeditated extraction. The 2022 buyback is the capital allocation move that defined the company's trajectory. The market had pushed AppLovin's market cap to $3.8 billion. The business was printing over $1 billion in EBITDA. Instead of open market repurchases, management identified that private equity backers and early insiders controlled nearly 50% of the float and needed liquidity. They bypassed the public market entirely, negotiated directly with those institutional holders, and executed a $6 billion leveraged buyback at the floor. Foroughi estimates that single decision generated $50 to $60 billion in retained value for remaining shareholders as the stock recovered. One negotiation. One decision. $50 billion created. The operating model is the part that should make every other tech CEO uncomfortable. AppLovin fired 40% of its workforce while growing revenue nearly 100% year over year. The C suite is four people—CEO, CFO, CTO, General Counsel. No CRO. No COO. No salesforce. Over 80% of the codebase is now written by AI, multiplying the output of their best engineers by up to 100x. The product does not need to be sold. Advertisers plug in, set a performance goal, and if the ROAS is positive they scale spend infinitely. The platform turns advertisers into blind arbitrageurs — they do not need to understand how it works, they just need to see the return. The TAM expansion is the next leg. Axon perfected gaming monetization. It is now being pointed at ecommerce and local SMBs — markets orders of magnitude larger than mobile gaming. The model does not need a sales team to penetrate them. It needs inventory and intent signals. It is acquiring both. The company has internal compensation triggers tied to a $1 trillion market cap. They went from $3.8 billion in 2022 to $154 billion today. The people running this business have done everything they said they would do, faster than anyone expected, with fewer people than anyone thought possible. That track record is the most important input in any forward model. Betting against a team that turned a $3.8 billion floor into $154 billion in three years, with 400 people, no salesforce, and an AI model that the entire industry has to use — requires a very specific and defensible thesis. Most of the people making that bet do not have one…stock is still being beaten to death…still not compelled. AI companies are just too good and getting better but interesting to see him come publicly to try to pump his stock

AppLovin CEO is trying to counter Anthropic’s CEO claims that SaaS is DEAD and never coming back. $APP is not an ad network. It is not a gaming company. It is an arbitrage engine with 400 engineers, $1.3 billion in cash flow per quarter, and a capital allocation track record that belongs in a business school curriculum. Start with the gaming studios. From 2018 to 2023, AppLovin acquired over 15 mobile gaming studios and 1,500 people. Nobody understood why. The answer was data. Third party advertisers would not share their conversion and ROAS data with AppLovin's model. So AppLovin became a first party advertiser — running its own games, generating its own behavioral data, feeding that data into Axon, its deep learning ad model. The studios were not a business. They were a training set. The moment Axon 2.0 launched in April 2023 and proved so effective that the entire gaming industry had to plug in and share their data to access the returns, the studios had served their purpose. AppLovin sold the entire portfolio to Tripledot and moved on. They used the asset to build the moat and immediately dumped the asset. That is not a pivot. That is a premeditated extraction. The 2022 buyback is the capital allocation move that defined the company's trajectory. The market had pushed AppLovin's market cap to $3.8 billion. The business was printing over $1 billion in EBITDA. Instead of open market repurchases, management identified that private equity backers and early insiders controlled nearly 50% of the float and needed liquidity. They bypassed the public market entirely, negotiated directly with those institutional holders, and executed a $6 billion leveraged buyback at the floor. Foroughi estimates that single decision generated $50 to $60 billion in retained value for remaining shareholders as the stock recovered. One negotiation. One decision. $50 billion created. The operating model is the part that should make every other tech CEO uncomfortable. AppLovin fired 40% of its workforce while growing revenue nearly 100% year over year. The C suite is four people—CEO, CFO, CTO, General Counsel. No CRO. No COO. No salesforce. Over 80% of the codebase is now written by AI, multiplying the output of their best engineers by up to 100x. The product does not need to be sold. Advertisers plug in, set a performance goal, and if the ROAS is positive they scale spend infinitely. The platform turns advertisers into blind arbitrageurs — they do not need to understand how it works, they just need to see the return. The TAM expansion is the next leg. Axon perfected gaming monetization. It is now being pointed at ecommerce and local SMBs — markets orders of magnitude larger than mobile gaming. The model does not need a sales team to penetrate them. It needs inventory and intent signals. It is acquiring both. The company has internal compensation triggers tied to a $1 trillion market cap. They went from $3.8 billion in 2022 to $154 billion today. The people running this business have done everything they said they would do, faster than anyone expected, with fewer people than anyone thought possible. That track record is the most important input in any forward model. Betting against a team that turned a $3.8 billion floor into $154 billion in three years, with 400 people, no salesforce, and an AI model that the entire industry has to use — requires a very specific and defensible thesis. Most of the people making that bet do not have one…stock is still being beaten to death…still not compelled. AI companies are just too good and getting better but interesting to see him come publicly to try to pump his stock

Nicholas Mugalli

100,854 просмотров • 4 месяцев назад

Meta just launched its 1st image model after Mark Zuckerberg’s AI shake-up. Muse Image is Meta Superintelligence Labs' first image generator after Muse Spark. They said the model will power new editing features in Meta’s Instagram photo app and be added to its marketer tools for creating platform ads. Meta had relied on Midjourney and Black Forest Labs for generation inside Meta AI. Muse Image brings that layer in-house, so Meta controls quality, cost, and product timing. Consumers get free access through Meta AI, WhatsApp chats, and Instagram Stories. Power users need Meta One or another monthly plan when free limits run out. Users can start from prompts, add photos, annotate edits, or sketch changes directly. Meta says it can erase photobombers, create QR codes, and keep visual text readable. Advertisers get variants through Advantage+ creative, with edits, style swaps, and brand-matched versions. Meta says internal tests trail GPT Image 2 but beat Nano Banana 2 on editing. This is another move in Meta's effort in trying to convert AI infrastructure spending into revenue beyond social ads.

Rohan Paul

20,377 просмотров • 1 месяц назад

Google DeepMind just dropped 'Genie', an AI that can generate interactive video games. This is a huge deal. Genie is trained on 200,000 hours of unsupervised public internet gaming videos and can generate video games from a single prompt or image. But here's what's insane: Despite not being trained on action or text annotations, the foundation model can determine who the main character is and enable a user to control that character in the generated world. It does this through its Latent Action Model, Video Tokenizer, and Dynamics Model (will go more in-depth on this in tomorrow's newsletter for those interested). And for those asking, yes, it's research-only and not publicly available (here come the Google memes), and it does come with some limitations, like only currently creating games at 1FPS. But this is the worst AI will ever be. Anyone will be able to create their own entirely imagined virtual worlds soon, and that's a wild sentence to say out loud.

Google DeepMind just dropped 'Genie', an AI that can generate interactive video games. This is a huge deal. Genie is trained on 200,000 hours of unsupervised public internet gaming videos and can generate video games from a single prompt or image. But here's what's insane: Despite not being trained on action or text annotations, the foundation model can determine who the main character is and enable a user to control that character in the generated world. It does this through its Latent Action Model, Video Tokenizer, and Dynamics Model (will go more in-depth on this in tomorrow's newsletter for those interested). And for those asking, yes, it's research-only and not publicly available (here come the Google memes), and it does come with some limitations, like only currently creating games at 1FPS. But this is the worst AI will ever be. Anyone will be able to create their own entirely imagined virtual worlds soon, and that's a wild sentence to say out loud.

Rowan Cheung

3,377,539 просмотров • 2 лет назад

There is a massive misunderstanding about what $APP actually is, and the recent Adam Foroughi interview just handed investors as a class in capital allocation, efficiency, and strategic vertical integration. At its core, AppLovin isn’t solely an ad-network or a gaming business but an arbitrage engine. APP is essentially run by a skeleton crew of about ~400 core engineers and product builders generating $1.3B in cash flow a quarter. The gaming studio acquisitions were purely strategic. From roughly 2018 to 2023, APP bought up mobile gaming studios. It was accumulating over 15 studios and 1,500 headcount. This was not because they wanted to be a gaming company, but strictly to harvest first-party conversion data. They needed proprietary behavioral and return on ad spend data to train their Axon deep learning models because third-party advertisers wouldn't share it. Once the Axon 2.0 model launched in April 2023, it became so hyper-effective that the entire gaming industry had no choice but to plug in and share their data to access the platform's unparalleled returns. Having served its purpose, APP then shed the distraction, selling off the gaming portfolio to Tripledot Studios. This was pure strategic rent extraction. They used the asset to train the AI, built the moat, and immediately dumped the asset. Then there is the capital allocation execution. Back in 2022, when the broader market abandoned the stock and pushed APP's market cap to a floor of $3.8B, the business was still printing over $1 B in EBITDA. Instead of executing standard open-market share repurchases, management identified that their private equity backers and early insiders held nearly 50% of the float and eventually wanted liquidity. Management bypassed the public market entirely. They negotiated directly with those institutional holders to execute a massive $6 billion leveraged buyback. They effectively retired a huge portion of the company at rock-bottom valuations. Foroughi noted this singular move generated $50 to $60 B in retained value for remaining shareholders as the stock rebounded. While the rest of big tech hoards headcount, APP actively fired 40% of its workforce in recent years while growing revenue by nearly 100% YoY. The C-suite consists only of the CEO, CFO, CTO, and General Counsel. There is no CRO, no COO, and virtually no salesforce. They replaced standard enterprise bloat with LLMs. Over 80% of their codebase is now written by AI, multiplying the output of their highest-tier engineers up to 100x. The product nearly sells itself. Advertisers plug in, set a performance goal, and if the ROAS is positive, they scale spend infinitely. It turns advertisers into blind arbitrageurs. The TAM expansion is also clear. Now that the model has perfected gaming monetization, APP is unleashing Axon onto e-commerce and local SMBs. They don't need a massive sales team to do this. The AI matches the right intent with the right ad, pushing conversion rates up. The company literally has internal compensation triggers tied to a $1 T market cap. Given they have grown from under $4 B in 2022 to roughly ~$154 B in market cap today, betting against this lean, hyper-competent team does not make much sense.

There is a massive misunderstanding about what $APP actually is, and the recent Adam Foroughi interview just handed investors as a class in capital allocation, efficiency, and strategic vertical integration. At its core, AppLovin isn’t solely an ad-network or a gaming business but an arbitrage engine. APP is essentially run by a skeleton crew of about ~400 core engineers and product builders generating $1.3B in cash flow a quarter. The gaming studio acquisitions were purely strategic. From roughly 2018 to 2023, APP bought up mobile gaming studios. It was accumulating over 15 studios and 1,500 headcount. This was not because they wanted to be a gaming company, but strictly to harvest first-party conversion data. They needed proprietary behavioral and return on ad spend data to train their Axon deep learning models because third-party advertisers wouldn't share it. Once the Axon 2.0 model launched in April 2023, it became so hyper-effective that the entire gaming industry had no choice but to plug in and share their data to access the platform's unparalleled returns. Having served its purpose, APP then shed the distraction, selling off the gaming portfolio to Tripledot Studios. This was pure strategic rent extraction. They used the asset to train the AI, built the moat, and immediately dumped the asset. Then there is the capital allocation execution. Back in 2022, when the broader market abandoned the stock and pushed APP's market cap to a floor of $3.8B, the business was still printing over $1 B in EBITDA. Instead of executing standard open-market share repurchases, management identified that their private equity backers and early insiders held nearly 50% of the float and eventually wanted liquidity. Management bypassed the public market entirely. They negotiated directly with those institutional holders to execute a massive $6 billion leveraged buyback. They effectively retired a huge portion of the company at rock-bottom valuations. Foroughi noted this singular move generated $50 to $60 B in retained value for remaining shareholders as the stock rebounded. While the rest of big tech hoards headcount, APP actively fired 40% of its workforce in recent years while growing revenue by nearly 100% YoY. The C-suite consists only of the CEO, CFO, CTO, and General Counsel. There is no CRO, no COO, and virtually no salesforce. They replaced standard enterprise bloat with LLMs. Over 80% of their codebase is now written by AI, multiplying the output of their highest-tier engineers up to 100x. The product nearly sells itself. Advertisers plug in, set a performance goal, and if the ROAS is positive, they scale spend infinitely. It turns advertisers into blind arbitrageurs. The TAM expansion is also clear. Now that the model has perfected gaming monetization, APP is unleashing Axon onto e-commerce and local SMBs. They don't need a massive sales team to do this. The AI matches the right intent with the right ad, pushing conversion rates up. The company literally has internal compensation triggers tied to a $1 T market cap. Given they have grown from under $4 B in 2022 to roughly ~$154 B in market cap today, betting against this lean, hyper-competent team does not make much sense.

CapexAndChill

24,419 просмотров • 4 месяцев назад

New Paper: Continuous Thought Machines 🧠 Neurons in brains use timing and synchronization in the way that they compute, but this is largely ignored in modern neural nets. We believe neural timing is key for the flexibility and adaptability of biological intelligence. We propose a new neural architecture, “Continuous Thought Machines” (CTMs), which is built from the ground up to use neural dynamics as a core representation for intelligence. By using neural dynamics as a first-class representational citizen, CTMs naturally perform adaptive computation. Many emergent, interesting behaviors arise as a result: CTMs solve mazes by observing a raw maze image and producing step-by-step instructions directly from its neural dynamics. When tasked with image recognition, the CTM naturally takes multiple steps to examine different parts of the image before making its decision. This step-by-step approach not only makes its behavior more interpretable but also improves accuracy: the longer it “thinks,” the more accurate its answers become. We also found that this allows the CTM to decide to spend less time thinking on simpler images, thus saving energy. When identifying a gorilla, for example, the CTM’s attention moves from eyes to nose to mouth in a pattern remarkably similar to human visual attention. I think this work underscores an important, yet often lost, synergy between neuroscience and AI. While modern AI is ostensibly brain-inspired, the two fields often operate in surprising isolation. By starting with such inspiration and iteratively following the emergent, interesting behaviors, we developed a model with unexpected capabilities, such as its surprisingly strong calibration in classification tasks, a feature that was not explicitly designed for. When we initially asked, “why do this research?”, we hoped the journey of the CTM would provide compelling answers. By embracing light biological inspiration and pursuing the novel behaviors observed, we have arrived at a model with emergent capabilities that exceeded our initial designs. We are committed to continuing this exploration, borrowing further concepts to discover what new and exciting behaviors will emerge, pushing the boundaries of what AI can achieve.

New Paper: Continuous Thought Machines 🧠 Neurons in brains use timing and synchronization in the way that they compute, but this is largely ignored in modern neural nets. We believe neural timing is key for the flexibility and adaptability of biological intelligence. We propose a new neural architecture, “Continuous Thought Machines” (CTMs), which is built from the ground up to use neural dynamics as a core representation for intelligence. By using neural dynamics as a first-class representational citizen, CTMs naturally perform adaptive computation. Many emergent, interesting behaviors arise as a result: CTMs solve mazes by observing a raw maze image and producing step-by-step instructions directly from its neural dynamics. When tasked with image recognition, the CTM naturally takes multiple steps to examine different parts of the image before making its decision. This step-by-step approach not only makes its behavior more interpretable but also improves accuracy: the longer it “thinks,” the more accurate its answers become. We also found that this allows the CTM to decide to spend less time thinking on simpler images, thus saving energy. When identifying a gorilla, for example, the CTM’s attention moves from eyes to nose to mouth in a pattern remarkably similar to human visual attention. I think this work underscores an important, yet often lost, synergy between neuroscience and AI. While modern AI is ostensibly brain-inspired, the two fields often operate in surprising isolation. By starting with such inspiration and iteratively following the emergent, interesting behaviors, we developed a model with unexpected capabilities, such as its surprisingly strong calibration in classification tasks, a feature that was not explicitly designed for. When we initially asked, “why do this research?”, we hoped the journey of the CTM would provide compelling answers. By embracing light biological inspiration and pursuing the novel behaviors observed, we have arrived at a model with emergent capabilities that exceeded our initial designs. We are committed to continuing this exploration, borrowing further concepts to discover what new and exciting behaviors will emerge, pushing the boundaries of what AI can achieve.

hardmaru

257,548 просмотров • 1 год назад

🇨🇳 It has started. A new home service in China pairs human cleaners with autonomous AI robots to tackle household chores. Residents in Shenzhen can now book a service where a human professional and an autonomous robot arrive together to clean their home. Real houses present a chaotic mess of dropped toys and random furniture that confuse traditional machines. X Square Robot and a major service platform named 58[.]com decided to tackle this chaos by launching China's first robot cleaner service in March-26. Customers use an application to hire a cleaning crew that consists of 1 human worker and 1 robot. The human takes care of the tricky chores that require complex judgment. The robot handles the repetitive physical work like picking up trash and wiping down flat surfaces. This machine runs on a system called WALL-A, which acts as a single continuous AI brain rather than a list of pre-written rules. They built this AI foundation model to perceive its surroundings and make its own decisions without human guidance. It processes visual data and plans multi-step actions. And deploying these robots into actual homes now provides the massive amounts of extremely important training data to improve it continuously. Alibaba and ByteDance backed this project. IMO, if the foundational model behind it figures out how to navigate a messy living room without getting stuck, it can learn to operate in almost any other physical environment.

🇨🇳 It has started. A new home service in China pairs human cleaners with autonomous AI robots to tackle household chores. Residents in Shenzhen can now book a service where a human professional and an autonomous robot arrive together to clean their home. Real houses present a chaotic mess of dropped toys and random furniture that confuse traditional machines. X Square Robot and a major service platform named 58[.]com decided to tackle this chaos by launching China's first robot cleaner service in March-26. Customers use an application to hire a cleaning crew that consists of 1 human worker and 1 robot. The human takes care of the tricky chores that require complex judgment. The robot handles the repetitive physical work like picking up trash and wiping down flat surfaces. This machine runs on a system called WALL-A, which acts as a single continuous AI brain rather than a list of pre-written rules. They built this AI foundation model to perceive its surroundings and make its own decisions without human guidance. It processes visual data and plans multi-step actions. And deploying these robots into actual homes now provides the massive amounts of extremely important training data to improve it continuously. Alibaba and ByteDance backed this project. IMO, if the foundational model behind it figures out how to navigate a messy living room without getting stuck, it can learn to operate in almost any other physical environment.

Rohan Paul

53,039 просмотров • 5 месяцев назад

New short course: Evaluating AI Agents! Evals are important for driving AI system improvements, and in this course you'll learn to systematically assess and improve an AI agent’s performance. This is built in partnership with Arize AI and taught by John Gilhuly, Head of Developer Relations, and , Director of Product. I've often found evals to be a critical tool in the agent development process - they can be the difference between picking the right thing to work on vs. wasting weeks of effort. Whether you’re building a shopping assistant, coding agent, or research assistant, having a structured evaluation process helps you refine its performance systematically, rather than relying on random trial and error. This course shows you how to structure your evals to assess the performance of each component of an agent and its end-to-end performance. For each component, you select the appropriate evaluators, test examples, and performance metrics. This helps you identify areas for improvement both during development and in production. (If you're familiar with error analysis in supervised learning, think of this as adapting those ideas to agentic workflows.) In this course, you'll build an AI agent, and add observability to visualize and debug its steps. You’ll learn about code-based evals, in which you write code explicitly to test a certain step, as well as LLM-as-a-Judge evals, in which you prompt an LLM to efficiently come up with ways to evaluate more open-ended outputs. In detail, you’ll: - Understand key differences between evaluating LLM-based systems and traditional software testing. - Add observability to an agent by collecting traces of the steps taken by the agent and visualizing them - Choose the appropriate evaluator - code-based, LLM-as-a-Judge, human-annotation based - for each component. - Compute a convergence score to evaluate if your agent can respond to a query in an efficient number of steps. - Run structured experiments to improve the agent’s performance by exploring changes to the prompt, LLM model, or the agent’s logic. - Understand how to deploy these evaluation techniques to monitor the agent’s performance in production. By the end of this course, you’ll know how to trace AI agents, systematically evaluate them, and improve their performance. Please sign up here:

New short course: Evaluating AI Agents! Evals are important for driving AI system improvements, and in this course you'll learn to systematically assess and improve an AI agent’s performance. This is built in partnership with Arize AI and taught by John Gilhuly, Head of Developer Relations, and , Director of Product. I've often found evals to be a critical tool in the agent development process - they can be the difference between picking the right thing to work on vs. wasting weeks of effort. Whether you’re building a shopping assistant, coding agent, or research assistant, having a structured evaluation process helps you refine its performance systematically, rather than relying on random trial and error. This course shows you how to structure your evals to assess the performance of each component of an agent and its end-to-end performance. For each component, you select the appropriate evaluators, test examples, and performance metrics. This helps you identify areas for improvement both during development and in production. (If you're familiar with error analysis in supervised learning, think of this as adapting those ideas to agentic workflows.) In this course, you'll build an AI agent, and add observability to visualize and debug its steps. You’ll learn about code-based evals, in which you write code explicitly to test a certain step, as well as LLM-as-a-Judge evals, in which you prompt an LLM to efficiently come up with ways to evaluate more open-ended outputs. In detail, you’ll: - Understand key differences between evaluating LLM-based systems and traditional software testing. - Add observability to an agent by collecting traces of the steps taken by the agent and visualizing them - Choose the appropriate evaluator - code-based, LLM-as-a-Judge, human-annotation based - for each component. - Compute a convergence score to evaluate if your agent can respond to a query in an efficient number of steps. - Run structured experiments to improve the agent’s performance by exploring changes to the prompt, LLM model, or the agent’s logic. - Understand how to deploy these evaluation techniques to monitor the agent’s performance in production. By the end of this course, you’ll know how to trace AI agents, systematically evaluate them, and improve their performance. Please sign up here:

Andrew Ng

126,587 просмотров • 1 год назад

I just jumped off Shopify’s Q2 2024 Earnings Call. Shopify had a great quarter and we continue to push the envelope of what’s possible in commerce. ICYMI, here are the highlights of Q2 2024 Shopify: 1️⃣ We passed a major milestone: Shopify has officially processed 1 TRILLION dollars in cumulative GMV 2️⃣ We delivered on the financials: - Gross Profit in Q2 was $1 billion dollars - up 25% YoY and grew faster than revenue which was $2 billion for the quarter - Our free cash flow margin more than doubled to 16% from 6% last year - This is our fourth consecutive quarter of double-digit free cash flow margin 3️⃣ Brands continue to flock to Shopify; Casper, Away luggage, Tonal, Vince Camuto, Ami Paris, Therabody, and legacy brands like QVC and Barnes and Noble all joined Shopify. ⚡We aren’t just helping brands thrive online, but also offline in physical retail. Our offline business surpassed $100 billion in cumulative GMV. Brands like Mejuri and Evereve launched online and offline with Shopify, and Oak + Fort launched online and rolled out Shopify POS in over 40 locations. ⚡B2B continues to grow. Shopify had its single largest GMV month ever for our B2B business and B2B grew more than 140% YoY. We had incredible operational discipline, and we nailed the execution of the plan we laid out. We are so well positioned to power commerce for years to come. Thank you to everyone on the Shopify team, our partners, and our incredible community for an exceptional quarter🚀

I just jumped off Shopify’s Q2 2024 Earnings Call. Shopify had a great quarter and we continue to push the envelope of what’s possible in commerce. ICYMI, here are the highlights of Q2 2024 Shopify: 1️⃣ We passed a major milestone: Shopify has officially processed 1 TRILLION dollars in cumulative GMV 2️⃣ We delivered on the financials: - Gross Profit in Q2 was $1 billion dollars - up 25% YoY and grew faster than revenue which was $2 billion for the quarter - Our free cash flow margin more than doubled to 16% from 6% last year - This is our fourth consecutive quarter of double-digit free cash flow margin 3️⃣ Brands continue to flock to Shopify; Casper, Away luggage, Tonal, Vince Camuto, Ami Paris, Therabody, and legacy brands like QVC and Barnes and Noble all joined Shopify. ⚡We aren’t just helping brands thrive online, but also offline in physical retail. Our offline business surpassed $100 billion in cumulative GMV. Brands like Mejuri and Evereve launched online and offline with Shopify, and Oak + Fort launched online and rolled out Shopify POS in over 40 locations. ⚡B2B continues to grow. Shopify had its single largest GMV month ever for our B2B business and B2B grew more than 140% YoY. We had incredible operational discipline, and we nailed the execution of the plan we laid out. We are so well positioned to power commerce for years to come. Thank you to everyone on the Shopify team, our partners, and our incredible community for an exceptional quarter🚀

Harley Finkelstein

161,958 просмотров • 2 лет назад

For weeks now we've been told: "Look, the reason this war failed is that there was no political solution to bring it to a close." Well—yesterday, a few minutes after the start of Shabbat, that solution arrived. And it is the best of our dreams: an agreement with the State of Lebanon. A framework that recognizes Israel's presence in the security zone for as long as Hezbollah exists, that calls for Hezbollah's disarmament, and in which Lebanon recognizes Israel's legitimacy, its sovereignty over the territories, and the end of rocket fire. Having said all that—let's see what it's actually worth. Both the grand principle, the one that says a military operation must always end with a decisive final blow, and the agreement itself. Because a quiet suspicion runs through all of us tonight: that even after this accord—coming forty-four years after another historic agreement with Lebanon—Hezbollah will not be laying down its arms. If anything, the opposite. So let us remember that the struggle in the Middle East is a long one. And in that long struggle, yesterday we took a real step forward—for the simplest of reasons: yet another Arab country, and one that matters to us, has in effect recognized the legitimacy of the State of Israel. All of it under the American umbrella—the very same United States that disappointed us so deeply only last week. Another step on a long road.

For weeks now we've been told: "Look, the reason this war failed is that there was no political solution to bring it to a close." Well—yesterday, a few minutes after the start of Shabbat, that solution arrived. And it is the best of our dreams: an agreement with the State of Lebanon. A framework that recognizes Israel's presence in the security zone for as long as Hezbollah exists, that calls for Hezbollah's disarmament, and in which Lebanon recognizes Israel's legitimacy, its sovereignty over the territories, and the end of rocket fire. Having said all that—let's see what it's actually worth. Both the grand principle, the one that says a military operation must always end with a decisive final blow, and the agreement itself. Because a quiet suspicion runs through all of us tonight: that even after this accord—coming forty-four years after another historic agreement with Lebanon—Hezbollah will not be laying down its arms. If anything, the opposite. So let us remember that the struggle in the Middle East is a long one. And in that long struggle, yesterday we took a real step forward—for the simplest of reasons: yet another Arab country, and one that matters to us, has in effect recognized the legitimacy of the State of Israel. All of it under the American umbrella—the very same United States that disappointed us so deeply only last week. Another step on a long road.

Amit Segal

42,622 просмотров • 2 месяцев назад

#WATCH | California, USA: On Anthropic Mythos, CEO of Blackstraw AI, Atul Arya says, "...Anthropic has released a reasoning model called Mythos... Unlike standard AI, it explains the 'why' behind its answers. Its primary initial use case is cybersecurity... helping companies identify not just where vulnerabilities exist, but why they occur. Currently, Mythos provides its testing early version to 20 major firms like Microsoft and CrowdStrike... to both simulate and prevent attacks... it's in its early phase... this model is powerful... and will be very useful in future..." Speaking about its impact on the banking industry, he says, "It can be both a threat and helpful... If someone wants to do something wrong, they'll attack the bank... According to me, for the banking industry, it is not a matter of worry. But yes, it is definitely a matter of caution as to what can happen with this and early adoption means adopting it in advance and reaching a stage when it is used in general availability..." On India's banking system's preparations for Mythos, he says, "...What's happened in India is that a large generation has acquired banking access directly on mobile phones, having never seen anything else, including a laptop... Even if we provide this only to banking systems, they are only halfway to success... The remaining 50% depends on public education. In a nation where 60-70% of people live in rural areas, educating users not to click suspicious links or share credentials is critical..."

#WATCH | California, USA: On Anthropic Mythos, CEO of Blackstraw AI, Atul Arya says, "...Anthropic has released a reasoning model called Mythos... Unlike standard AI, it explains the 'why' behind its answers. Its primary initial use case is cybersecurity... helping companies identify not just where vulnerabilities exist, but why they occur. Currently, Mythos provides its testing early version to 20 major firms like Microsoft and CrowdStrike... to both simulate and prevent attacks... it's in its early phase... this model is powerful... and will be very useful in future..." Speaking about its impact on the banking industry, he says, "It can be both a threat and helpful... If someone wants to do something wrong, they'll attack the bank... According to me, for the banking industry, it is not a matter of worry. But yes, it is definitely a matter of caution as to what can happen with this and early adoption means adopting it in advance and reaching a stage when it is used in general availability..." On India's banking system's preparations for Mythos, he says, "...What's happened in India is that a large generation has acquired banking access directly on mobile phones, having never seen anything else, including a laptop... Even if we provide this only to banking systems, they are only halfway to success... The remaining 50% depends on public education. In a nation where 60-70% of people live in rural areas, educating users not to click suspicious links or share credentials is critical..."

ANI

93,224 просмотров • 4 месяцев назад

.John Coogan says the recent reporting on Meta's 'tokenmaxxing' is less of a sign of bad incentives at the company, and more of a tell about its potential strategy for more vertical integration: "I think it makes clearer the strategy with MSL. Because it's clear that they're spending hundreds of millions of dollars on this, just for internal code-gen tooling — like, running their business." "They are going to spend an inordinate amount of money on frontier inference, and so [by] training a model there, they will be able to amortize the training cost of the next model they build not just over, 'Can they get a product out that goes viral and becomes its own standalone chat app that people pay for.'" "Just on the internal usage, they could be running a multi-billion-dollar token bill that they would have to pay another lab. And so if they develop that internally, it's pure vertical integration." "And then you also have everything that's happening on the actual ad targeting and content delivery side..." "The big question has been like, 'Is Meta going to be able to launch an entirely new AI product like Vibes or something like that?'" "And this is a data point that, to me, says they don't need to. Because just from a pure vertical integration [perspective], the investment in MSL can pencil out."

.John Coogan says the recent reporting on Meta's 'tokenmaxxing' is less of a sign of bad incentives at the company, and more of a tell about its potential strategy for more vertical integration: "I think it makes clearer the strategy with MSL. Because it's clear that they're spending hundreds of millions of dollars on this, just for internal code-gen tooling — like, running their business." "They are going to spend an inordinate amount of money on frontier inference, and so [by] training a model there, they will be able to amortize the training cost of the next model they build not just over, 'Can they get a product out that goes viral and becomes its own standalone chat app that people pay for.'" "Just on the internal usage, they could be running a multi-billion-dollar token bill that they would have to pay another lab. And so if they develop that internally, it's pure vertical integration." "And then you also have everything that's happening on the actual ad targeting and content delivery side..." "The big question has been like, 'Is Meta going to be able to launch an entirely new AI product like Vibes or something like that?'" "And this is a data point that, to me, says they don't need to. Because just from a pure vertical integration [perspective], the investment in MSL can pencil out."

TBPN

32,654 просмотров • 4 месяцев назад

#WATCH | Mumbai, MH: On India's GDP growth of 7.8% in Q1, Director and Chief Economist with Deloitte India, Dr Rumki Majumdar says, "...Our projections for the entire fiscal year expected 6.7 per cent growth, which was pretty much aligned with what the RBI had predicted as well in August. But with 7.8 per cent growth in the very first quarter, and that too in a quarter considered to be the most uncertain of the year...In fact, that inflation remains contained, we are hopeful that inflation will remain more or less moderate. In that way, it will likely ensure that all economic fundamentals remain strong. Another interesting thing that I would like to highlight is capital formation. At 11.9 per cent, it is one of the highest capital formations that India has ever seen, which means that while consumption spending holds on, investment is building up. Given that corporate revenues and sales were very strong in the April to June quarter, clearly, private capex is likely to pull up the rest of the economy much stronger than what it has done so far..."

#WATCH | Mumbai, MH: On India's GDP growth of 7.8% in Q1, Director and Chief Economist with Deloitte India, Dr Rumki Majumdar says, "...Our projections for the entire fiscal year expected 6.7 per cent growth, which was pretty much aligned with what the RBI had predicted as well in August. But with 7.8 per cent growth in the very first quarter, and that too in a quarter considered to be the most uncertain of the year...In fact, that inflation remains contained, we are hopeful that inflation will remain more or less moderate. In that way, it will likely ensure that all economic fundamentals remain strong. Another interesting thing that I would like to highlight is capital formation. At 11.9 per cent, it is one of the highest capital formations that India has ever seen, which means that while consumption spending holds on, investment is building up. Given that corporate revenues and sales were very strong in the April to June quarter, clearly, private capex is likely to pull up the rest of the economy much stronger than what it has done so far..."

ANI

64,381 просмотров • 1 день назад

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:

Andrew Ng

68,034 просмотров • 1 год назад

🚨GAMING NEEDS A RESET: WHY ELON'S GAME STUDIO COULD CHANGE EVERYTHING The gaming industry desperately needs a shake-up. What once was a medium defined by imagination, fun, and pushing technological boundaries has increasingly become formulaic, predictable, and disconnected from what players actually want. As Tim Sweeney, Founder and CEO of Epic Games, perfectly articulated: "There's one constant in gaming that gets lost—fun. People play games for fun. A lot of recent failures didn't deliver fun, or not enough to compete with real life. Everything else—politics, drama—is a sideshow. The core problem is a failure to deliver fun." This insight from one of gaming's most successful visionaries cuts to the heart of the issue. Major studios have fallen into a cycle of risk-aversion – churning out sequel after sequel while prioritizing monetization schemes over gameplay innovation. Meanwhile, creative decisions increasingly cater to external pressures rather than addressing Sweeney's fundamental truth. Enter a potential catalyst for change. A new AI-focused game studio backed by someone with resources, technological vision, and a track record of transforming calcified industries represents exactly what gaming needs right now – an outsider's perspective unencumbered by industry groupthink. For too long, gaming has been caught between corporate profit-maximization and agenda-driven messaging. A fresh approach that returns to Sweeney's simple truth—that fun is the constant that matters most—could spark a renaissance in an art form that millions love but increasingly find disappointing.

🚨GAMING NEEDS A RESET: WHY ELON'S GAME STUDIO COULD CHANGE EVERYTHING The gaming industry desperately needs a shake-up. What once was a medium defined by imagination, fun, and pushing technological boundaries has increasingly become formulaic, predictable, and disconnected from what players actually want. As Tim Sweeney, Founder and CEO of Epic Games, perfectly articulated: "There's one constant in gaming that gets lost—fun. People play games for fun. A lot of recent failures didn't deliver fun, or not enough to compete with real life. Everything else—politics, drama—is a sideshow. The core problem is a failure to deliver fun." This insight from one of gaming's most successful visionaries cuts to the heart of the issue. Major studios have fallen into a cycle of risk-aversion – churning out sequel after sequel while prioritizing monetization schemes over gameplay innovation. Meanwhile, creative decisions increasingly cater to external pressures rather than addressing Sweeney's fundamental truth. Enter a potential catalyst for change. A new AI-focused game studio backed by someone with resources, technological vision, and a track record of transforming calcified industries represents exactly what gaming needs right now – an outsider's perspective unencumbered by industry groupthink. For too long, gaming has been caught between corporate profit-maximization and agenda-driven messaging. A fresh approach that returns to Sweeney's simple truth—that fun is the constant that matters most—could spark a renaissance in an art form that millions love but increasingly find disappointing.

Mario Nawfal

276,074 просмотров • 1 год назад

I think I discovered the cheapest way to code with AI 😁 I tried GLM 4.5 on Claude Code and I'm really amazed how good this model is! I usually test with my nextjs crm app benchmark so I asked it to do that. Prompt "create a nextjs app which is a crm app. use sqlite for the database and use api routes to manage connecting to the database from the frontend" My observations - Generation feels like Sonnet 4 intelligence - Speed is great - It used create-next-app so it bootstrapped nextjs v15.5.2 - It created a sqlite database and used Prisma ORM and helped me easily add seed data - I needed a few prompts to iron out minor issues but all's good. It didn't struggle or got stuck. - Spent $0.50 for this app. Could be way cheaper with the Z.ai sub plans After trying, I can say GLM 4.5 model is up there as the best open source model for coding. Will continue using it and update here. So far so good! P.S. DM me if you want my referral link, otherwise just check Not dropping the link here so it doesn’t look like I’m hyping it. Here's a video of the experience 👇(sped up a bit)

I think I discovered the cheapest way to code with AI 😁 I tried GLM 4.5 on Claude Code and I'm really amazed how good this model is! I usually test with my nextjs crm app benchmark so I asked it to do that. Prompt "create a nextjs app which is a crm app. use sqlite for the database and use api routes to manage connecting to the database from the frontend" My observations - Generation feels like Sonnet 4 intelligence - Speed is great - It used create-next-app so it bootstrapped nextjs v15.5.2 - It created a sqlite database and used Prisma ORM and helped me easily add seed data - I needed a few prompts to iron out minor issues but all's good. It didn't struggle or got stuck. - Spent $0.50 for this app. Could be way cheaper with the Z.ai sub plans After trying, I can say GLM 4.5 model is up there as the best open source model for coding. Will continue using it and update here. So far so good! P.S. DM me if you want my referral link, otherwise just check Not dropping the link here so it doesn’t look like I’m hyping it. Here's a video of the experience 👇(sped up a bit)

Melvin Vivas

45,749 просмотров • 1 год назад

Today, we're announcing Heaviside, our foundation model for electromagnetism. Trained on tens of millions of designs and over 20 years of proprietary simulation data, Heaviside predicts electromagnetic behavior from geometry in 13ms, which is 800,000x faster than a commercial solver. Heaviside is not a language model, and it’s not a surrogate model. Heaviside marks a new class of foundation model for physics which understands the fundamental relationships between materials, the geometries and the electromagnetic fields they generate. We’re releasing a research preview of Heaviside in Atlas RF Studio, an interactive agentic sandbox where you describe the EM behavior you want and the model generates the physical structure that produces it. Arena Physica , we believe the implications of this class of model extend well beyond RF, as the frontier of exquisite hardware is electromagnetically-governed: wireless communication, radar, power delivery, high-speed computing, and the interconnects inside every chip on earth. In the months ahead, we’re excited to scale up Heaviside to broader frequency ranges, design spaces, and to support silicon-level designs, and deploy it with our closest partners and collaborators in service of their biggest design challenges. If you’ve read our thesis, this is just Step 2 in our pursuit of electromagnetic superintelligence. Read the full announcement and try Atlas RF Studio…tell us what you think:

Today, we're announcing Heaviside, our foundation model for electromagnetism. Trained on tens of millions of designs and over 20 years of proprietary simulation data, Heaviside predicts electromagnetic behavior from geometry in 13ms, which is 800,000x faster than a commercial solver. Heaviside is not a language model, and it’s not a surrogate model. Heaviside marks a new class of foundation model for physics which understands the fundamental relationships between materials, the geometries and the electromagnetic fields they generate. We’re releasing a research preview of Heaviside in Atlas RF Studio, an interactive agentic sandbox where you describe the EM behavior you want and the model generates the physical structure that produces it. Arena Physica , we believe the implications of this class of model extend well beyond RF, as the frontier of exquisite hardware is electromagnetically-governed: wireless communication, radar, power delivery, high-speed computing, and the interconnects inside every chip on earth. In the months ahead, we’re excited to scale up Heaviside to broader frequency ranges, design spaces, and to support silicon-level designs, and deploy it with our closest partners and collaborators in service of their biggest design challenges. If you’ve read our thesis, this is just Step 2 in our pursuit of electromagnetic superintelligence. Read the full announcement and try Atlas RF Studio…tell us what you think:

Arya Hezarkhani

699,521 просмотров • 5 месяцев назад

I just cracked the code on AI UGC ads 🤯 This ad is nearly THREE minutes long, and it has perfect character, voice, and product consistency across the entire video. I found a winning authority-style ad running on Meta, recreated it shot-for-shot with AI, and ended up with what you see below. Two consistent characters, consistent voice, real product in hand. If you've been stuck generating 8 second clips that look fine but can't hold attention for a full funnel ad, this is the workflow you're missing. Here's exactly how I did it: → Find a winning ad in Meta Ad Library and download it → Upload to Loom to get a timestamped transcript (this becomes your script) → Screenshot each character → clean in ChatGPT Images 2.0 → upload as Seedance Elements → Generate your first clip per character to lock in their voice → Extract that voice as an MP3 via CapCut, attach it to every clip after → Edit in CapCut: remove dead zones, speed up to 1.1x, done → Drop final video in the Captions app, add subtitles, export 2 minutes and 40 seconds, perfect character & product consistency. I put together the full step-by-step playbook with exact Seedance prompts, the CapCut speed trick, and a voiceover technique most people don't know about. Want access completely for free? > Comment "UGC" > Like this post And I'll send it over (must be following so I can DM)

I just cracked the code on AI UGC ads 🤯 This ad is nearly THREE minutes long, and it has perfect character, voice, and product consistency across the entire video. I found a winning authority-style ad running on Meta, recreated it shot-for-shot with AI, and ended up with what you see below. Two consistent characters, consistent voice, real product in hand. If you've been stuck generating 8 second clips that look fine but can't hold attention for a full funnel ad, this is the workflow you're missing. Here's exactly how I did it: → Find a winning ad in Meta Ad Library and download it → Upload to Loom to get a timestamped transcript (this becomes your script) → Screenshot each character → clean in ChatGPT Images 2.0 → upload as Seedance Elements → Generate your first clip per character to lock in their voice → Extract that voice as an MP3 via CapCut, attach it to every clip after → Edit in CapCut: remove dead zones, speed up to 1.1x, done → Drop final video in the Captions app, add subtitles, export 2 minutes and 40 seconds, perfect character & product consistency. I put together the full step-by-step playbook with exact Seedance prompts, the CapCut speed trick, and a voiceover technique most people don't know about. Want access completely for free? > Comment "UGC" > Like this post And I'll send it over (must be following so I can DM)

Mike Futia

40,148 просмотров • 3 месяцев назад

“We believe in intelligence as a public good before everything else.” Here’s my new episode with mephisto, who co-founded Nous Research and helped build Hermes, the #1 personal agent and AI app on OpenRouter. Karan is the most down-to-earth AI founder I've interviewed. We had a great chat about: → Why open-source AI must win → How Hermes improves itself without slop → How to mod your favorite games with AI Some quotes from Karan: “We need to keep giving this level of intelligence to everyone so that we're all on an even and equal playing field.” “Anytime [the model] says ‘you’re absolutely right,’ you’re being reward hacked. You are fuel for its reward function.” “Memory, skills, and the self-improvement loop are where context is stored. That’s where the actual personality you want lives. You can put that on any model.” 📌 Watch now: Thanks to our sponsors: Riverside: All-in-one AI studio for podcasts and video Linear: The AI agent platform for modern teams

“We believe in intelligence as a public good before everything else.” Here’s my new episode with mephisto, who co-founded Nous Research and helped build Hermes, the #1 personal agent and AI app on OpenRouter. Karan is the most down-to-earth AI founder I've interviewed. We had a great chat about: → Why open-source AI must win → How Hermes improves itself without slop → How to mod your favorite games with AI Some quotes from Karan: “We need to keep giving this level of intelligence to everyone so that we're all on an even and equal playing field.” “Anytime [the model] says ‘you’re absolutely right,’ you’re being reward hacked. You are fuel for its reward function.” “Memory, skills, and the self-improvement loop are where context is stored. That’s where the actual personality you want lives. You can put that on any model.” 📌 Watch now: Thanks to our sponsors: Riverside: All-in-one AI studio for podcasts and video Linear: The AI agent platform for modern teams

Peter Yang

124,572 просмотров • 1 месяц назад