正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Is Anthropic altering model performance to force costly upgrades? Chapter Co-Founder and CEO Cobi Blumenfeld-Gantz outlines a shift his team recently noticed: "Anthropic has kind of taken a page out of the Apple playbook, where they have...decreased the quality of the prior models right before they launched the new... show more

The Information

1,610,691 subscribers

36,895 次观看 • 5 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

11 条评论

Tech With Matteo 的头像

Tech With Matteo5 个月前

@CobiBGantz good breakdown, a 25 word cap on tool calls would explain a lot of the weird truncation people were seeing, glad it got reverted

Tech With Matteo 的头像

Tech With Matteo5 个月前

@CobiBGantz as someone using claude daily to build stuff this claim doesnt match what i noticed, the reverted system prompt change explains way more than intentional degradation

John Allen Berger 的头像

John Allen Berger5 个月前

@CobiBGantz Yep learn from the champ for squeezing blood out of a turnip. Karma is going to bite your snotty asses in the coming years. Your AI on local devices is fundamentally flawed and dead on arrival. Thank Tim Cook for this fuckup

AGI Pulse 的头像

AGI Pulse5 个月前

@CobiBGantz Very much so. That's why we will see AI bubble burst

Houman Asefi 的头像

Houman Asefi5 个月前

@CobiBGantz 'Anthropic decreased quality of prior models'..based on what evidence?

Michael Spencer 的头像

Michael Spencer5 个月前

@CobiBGantz If you are a frontier lab and develop the best models, what kind of incentives do you have to boost perceived productivity and profitability?

Daniel M. Karim - Aka The Ghost 的头像

Daniel M. Karim - Aka The Ghost5 个月前

@CobiBGantz Token transparency start up request.

Aris Chronos 的头像

Aris Chronos5 个月前

@CobiBGantz I think for a company built on 'ethical AI', the Apple playbook is quite the look.

Carlos Aggio 的头像

Carlos Aggio5 个月前

@CobiBGantz To me it’s simpler. They’ve reallocated priority inference compute to the newer model and we intuitively know that frontier LLMs with access to more compute at inference time seem to perform better.. besides the “impression of degradation”

F U Z Z Y G O A T ™ 的头像

F U Z Z Y G O A T ™5 个月前

This kind of gaming of the audience feels inevitable. Throttle performance to justify an upgrade, add adverts so you can charge people to remove them and get back to the experience they already had. It’s all coming. Right now we’re in the honeymoon period, the equivalent of the 1980s and 1990s internet era, the golden age before everything became cookie banners, GDPR pop-ups, adverts, subscriptions, and paywalls. Enjoy it while it lasts.

Winston B. 的头像

Winston B.5 个月前

The assumption baked in is deliberate degradation for upsell leverage, which doesn't hold up against what Anthropic actually disclosed: a system-prompt addition in mid-April that hard-capped Claude's text-between-tool-calls to 25 words and final responses to 100 words caused a measurable drop in coding eval quality before it was reverted.

相关视频

NEW - Anthropic CEO Dario Amodei says their Claude AI model has led to the discovery of a “new molecular machine” capable of gene editing

NEW - Anthropic CEO Dario Amodei says their Claude AI model has led to the discovery of a “new molecular machine” capable of gene editing

Chief Nerd

172,153 次观看 • 1 个月前

David Sacks: Anthropic Is Great at Two Things, Releasing Products and Scaring People “Anthropic has proven that it's very good at two things. One is product releases. The second is scaring people. And we've seen a pattern in their previous releases where, at the same time they roll out a new model or a new model card, something like that, they also roll out some study showing really the worst possible implications of where the technology could lead.” Source: The All-In Podcast

David Sacks: Anthropic Is Great at Two Things, Releasing Products and Scaring People “Anthropic has proven that it's very good at two things. One is product releases. The second is scaring people. And we've seen a pattern in their previous releases where, at the same time they roll out a new model or a new model card, something like that, they also roll out some study showing really the worst possible implications of where the technology could lead.” Source: The All-In Podcast

Mario Nawfal

50,266 次观看 • 7 个月前

Anthropic CEO: “At Anthropic, we essentially have Claude designing the next version of Claude itself.” Meanwhile, Anthropic says every new model is given code that trains a small AI model and asked to make it faster. Claude Opus 4 achieved a 3x speedup. Mythos Preview reached 52x. Recursive self-improvement is starting to look real The models are helping build what comes next.

Anthropic CEO: “At Anthropic, we essentially have Claude designing the next version of Claude itself.” Meanwhile, Anthropic says every new model is given code that trains a small AI model and asked to make it faster. Claude Opus 4 achieved a 3x speedup. Mythos Preview reached 52x. Recursive self-improvement is starting to look real The models are helping build what comes next.

CG

343,798 次观看 • 5 个月前

Anthropic, the tech company that recently clashed with the Pentagon, says it has developed a new AI model that is too powerful to be released to the public.

Anthropic, the tech company that recently clashed with the Pentagon, says it has developed a new AI model that is too powerful to be released to the public.

Channel 4 News

103,858 次观看 • 7 个月前

Dario Amodei says Anthropic held its first version of Claude back about six months, and gave up the ChatGPT moment doing it: "When we had our first product, our first version of Claude, which is our model, we actually delayed the release of that model roughly six months because this was such a new technology that we weren't sure of the safety properties." "We weren't sure we wanted to be the ones to kind of kick off a race." "This was just before ChatGPT." "So we arguably had the opportunity to seize the ChatGPT moment and we chose to release a little later, which I think had real commercial consequences but set the culture of the company." ChatGPT launched on November 30, 2022. Anthropic launched Claude in March 2023. - Dario Amodei (Dario Amodei), co-founder and CEO of Anthropic (Anthropic), at the Council on Foreign Relations (Council on Foreign Relations) CEO Speaker Series.

Dario Amodei says Anthropic held its first version of Claude back about six months, and gave up the ChatGPT moment doing it: "When we had our first product, our first version of Claude, which is our model, we actually delayed the release of that model roughly six months because this was such a new technology that we weren't sure of the safety properties." "We weren't sure we wanted to be the ones to kind of kick off a race." "This was just before ChatGPT." "So we arguably had the opportunity to seize the ChatGPT moment and we chose to release a little later, which I think had real commercial consequences but set the culture of the company." ChatGPT launched on November 30, 2022. Anthropic launched Claude in March 2023. - Dario Amodei (Dario Amodei), co-founder and CEO of Anthropic (Anthropic), at the Council on Foreign Relations (Council on Foreign Relations) CEO Speaker Series.

Karl Mehta

116,828 次观看 • 2 个月前

.David Sacks: Anthropic is good at two things—product releases, and scaring people. “They have a proven pattern of using fear as a way to market their new products.” He had Grok create a list of every time they’ve used scare tactics on the public. “At the same time they roll out a new model, they also roll out some study showing the worst possible implication of where the technology could lead.”

.David Sacks: Anthropic is good at two things—product releases, and scaring people. “They have a proven pattern of using fear as a way to market their new products.” He had Grok create a list of every time they’ve used scare tactics on the public. “At the same time they roll out a new model, they also roll out some study showing the worst possible implication of where the technology could lead.”

Jawwwn

99,705 次观看 • 7 个月前

Anthropic CEO: “open source is kind of a distraction you still can’t really see what’s happening inside the model, so it’s not truly “free.” People should be paying a lot more attention to what Anthropic is doing.

Anthropic CEO: “open source is kind of a distraction you still can’t really see what’s happening inside the model, so it’s not truly “free.” People should be paying a lot more attention to what Anthropic is doing.

CG

1,268,604 次观看 • 4 个月前

🚨 ANTHROPIC JUST REVEALED CLAUDE MYTHOS ABILITIES Anthropic just formally announced "Claude Mythos Preview" and launched "Project Glasswing" to deploy it for cybersecurity defense. The models are unlocking completely new, autonomous behaviors. This isn't about slightly better benchmark scores. This is about what the model can do. Here are the direct quotes from Anthropic’s research team (including Dario) on exactly what Mythos is capable of: • Chaining Exploits: "It has the ability to chain together vulnerabilities... this model is able to create exploits out of three, four, sometimes five vulnerabilities that in sequence give you some kind of very sophisticated end outcome." • The Professional Standard: "The model that we're experimenting with is, by and large, as good as a professional human at identifying bugs." • Unprecedented Autonomy: "It's just generally better at pursuing really long-range tasks that are kind of like the tasks that a human security researcher would do throughout the course of an entire day." The Reality Check: Dario Amodei flat out said: "There's a kind of accelerating exponential... Claude Mythos Preview is a particularly big jump along that point." Because this model has become so capable at identifying zero-days, they are restricting its release to top tech partners to try to patch the world's software before these capabilities leak out. The autonomous researcher era has officially arrived. It’s over 💀

🚨 ANTHROPIC JUST REVEALED CLAUDE MYTHOS ABILITIES Anthropic just formally announced "Claude Mythos Preview" and launched "Project Glasswing" to deploy it for cybersecurity defense. The models are unlocking completely new, autonomous behaviors. This isn't about slightly better benchmark scores. This is about what the model can do. Here are the direct quotes from Anthropic’s research team (including Dario) on exactly what Mythos is capable of: • Chaining Exploits: "It has the ability to chain together vulnerabilities... this model is able to create exploits out of three, four, sometimes five vulnerabilities that in sequence give you some kind of very sophisticated end outcome." • The Professional Standard: "The model that we're experimenting with is, by and large, as good as a professional human at identifying bugs." • Unprecedented Autonomy: "It's just generally better at pursuing really long-range tasks that are kind of like the tasks that a human security researcher would do throughout the course of an entire day." The Reality Check: Dario Amodei flat out said: "There's a kind of accelerating exponential... Claude Mythos Preview is a particularly big jump along that point." Because this model has become so capable at identifying zero-days, they are restricting its release to top tech partners to try to patch the world's software before these capabilities leak out. The autonomous researcher era has officially arrived. It’s over 💀

Chris

46,155 次观看 • 7 个月前

🚨 DAVID SACKS: “Anthropic has proven that it's very good at two things — One is product releases, the second is scaring people … At the same time they roll out a new model … they also roll out some study showing the worst possible implication where the technology could lead.”

🚨 DAVID SACKS: “Anthropic has proven that it's very good at two things — One is product releases, the second is scaring people … At the same time they roll out a new model … they also roll out some study showing the worst possible implication where the technology could lead.”

Chief Nerd

295,335 次观看 • 7 个月前

Anthropic CEO Dario Amodei says if the scaling hypothesis turns out to be true, then a $100 billion AI model will have the intelligence of a Nobel Prize winner

Anthropic CEO Dario Amodei says if the scaling hypothesis turns out to be true, then a $100 billion AI model will have the intelligence of a Nobel Prize winner

Tsarathustra

156,759 次观看 • 2 年前

Anthropic research lead: "99% of our engineers are running swarms of 300+ self-improving agents. close the agent loop. Give the model a way to verify its own output" in a 20-minute session, Anthropic team member explains how to build a model that improves itself. Claude + loops + plan mode + dynamic workflows -that’s the secret. Watch the talk, then save the playbook below.

Anthropic research lead: "99% of our engineers are running swarms of 300+ self-improving agents. close the agent loop. Give the model a way to verify its own output" in a 20-minute session, Anthropic team member explains how to build a model that improves itself. Claude + loops + plan mode + dynamic workflows -that’s the secret. Watch the talk, then save the playbook below.

Codez

669,853 次观看 • 4 个月前

NEWS: Tesla has released a new video of Daniel Ho going over some of the upgrades on the new Model 3.

NEWS: Tesla has released a new video of Daniel Ho going over some of the upgrades on the new Model 3.

Sawyer Merritt

686,983 次观看 • 3 年前

$AMZN AWS CEO said Anthropic’s latest models are being trained on Trainium (including Mythos) which Anthropic describes as its most capable model yet. That is notable because Anthropic is limiting Mythos due to cybersecurity risk & high compute costs which says a lot about the kind of model Amazon’s chips are now helping train.

$AMZN AWS CEO said Anthropic’s latest models are being trained on Trainium (including Mythos) which Anthropic describes as its most capable model yet. That is notable because Anthropic is limiting Mythos due to cybersecurity risk & high compute costs which says a lot about the kind of model Amazon’s chips are now helping train.

Shay Boloor

57,746 次观看 • 7 个月前

JD Vance: "There are things at some of the AI companies you see happening, where you feel like there is some really, really weird spiritual dark energy. Like one of the companies that develops these models, when they get rid of one model and announce a new model, they actually hold a burial ceremony for the old model, and the employees apparently will cry."

JD Vance: "There are things at some of the AI companies you see happening, where you feel like there is some really, really weird spiritual dark energy. Like one of the companies that develops these models, when they get rid of one model and announce a new model, they actually hold a burial ceremony for the old model, and the employees apparently will cry."

Harlan Stewart

361,667 次观看 • 2 个月前

True Ventures Tony Conrad says Polsia's founder let Anthropic "run kind of crazy" and woke up to a $1.5 million bill. Now he's looking at Chinese open-source models: "We have a company, True, Polsia, and Ben's been really public about, he just let Anthropic run kind of crazy with Anthropic tokens, and he woke up and he had, like, a million-and-a-half-dollar bill." "He's like, no, no, that's not gonna work. I'm gonna go look at some of the Chinese models, open source models, and find the right balance." "We're in a really interesting moment where how do you find cheaper approaches to the tech stack that aren't just all Anthropic and OpenAI-driven. That's a very expensive Ferrari-type model." "They may be the best. I'm not gonna opine on that, but they may be the best. That might be worth it. But you might be able to do really great stuff with 95% of that." True Ventures Madeline Minshew

True Ventures Tony Conrad says Polsia's founder let Anthropic "run kind of crazy" and woke up to a $1.5 million bill. Now he's looking at Chinese open-source models: "We have a company, True, Polsia, and Ben's been really public about, he just let Anthropic run kind of crazy with Anthropic tokens, and he woke up and he had, like, a million-and-a-half-dollar bill." "He's like, no, no, that's not gonna work. I'm gonna go look at some of the Chinese models, open source models, and find the right balance." "We're in a really interesting moment where how do you find cheaper approaches to the tech stack that aren't just all Anthropic and OpenAI-driven. That's a very expensive Ferrari-type model." "They may be the best. I'm not gonna opine on that, but they may be the best. That might be worth it. But you might be able to do really great stuff with 95% of that." True Ventures Madeline Minshew

MTS

15,895 次观看 • 2 个月前

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.

Vikram M

21,463 次观看 • 4 个月前

SITUATION EXPLAINED: Why is the Bridgewater Thinking Machines paper a blueprint for every enterprise that has ever paid frontier model prices for generic outputs? 47fucb4r8curb4fc8f8r4bfic8r: "What Thinking Machines and Bridgewater did together is a blueprint for how a wide variety of stakeholders within the market can benefit tremendously from using LLMs and integrating them into a larger enterprise system." "They got a pretty old model. A Qwen3 model. Two hundred and thirty-five billion parameters. Not a new model. Not a very sophisticated or impressive state-of-the-art model. And they essentially fine-tuned it on Bridgewater's own proprietary data." "A financial analyst gets a ton of financial documents and has to filter which are worth looking at. Then evaluate which news pieces are relevant. They have a lot of tacit knowledge from doing the job on where to look. What they did was inject a lot of this tacit knowledge into the model through fine-tuning." "Stop obsessing over state-of-the-art models that are the most intelligent possible. Focus on targeting a very specific LLM specifically trained for a specific function to do that function better than the alternatives on the market." "This old non-state-of-the-art model got 84.7% accuracy versus 78% for all the state-of-the-art models. A massive improvement. And the kicker is they got a better performing model that is 14 times cheaper to use."

SITUATION EXPLAINED: Why is the Bridgewater Thinking Machines paper a blueprint for every enterprise that has ever paid frontier model prices for generic outputs? 47fucb4r8curb4fc8f8r4bfic8r: "What Thinking Machines and Bridgewater did together is a blueprint for how a wide variety of stakeholders within the market can benefit tremendously from using LLMs and integrating them into a larger enterprise system." "They got a pretty old model. A Qwen3 model. Two hundred and thirty-five billion parameters. Not a new model. Not a very sophisticated or impressive state-of-the-art model. And they essentially fine-tuned it on Bridgewater's own proprietary data." "A financial analyst gets a ton of financial documents and has to filter which are worth looking at. Then evaluate which news pieces are relevant. They have a lot of tacit knowledge from doing the job on where to look. What they did was inject a lot of this tacit knowledge into the model through fine-tuning." "Stop obsessing over state-of-the-art models that are the most intelligent possible. Focus on targeting a very specific LLM specifically trained for a specific function to do that function better than the alternatives on the market." "This old non-state-of-the-art model got 84.7% accuracy versus 78% for all the state-of-the-art models. A massive improvement. And the kicker is they got a better performing model that is 14 times cheaper to use."

MTS

93,185 次观看 • 4 个月前

My Tesla standard Model Y has Swing Mode where the front air vents oscillate side to side so airflow distributes evenly throughout the car. The thing that many people do not know about Tesla is that they don’t wait for a “new model year” to add new features. They don’t sit around planning for a 2025 version or a 2026 version. When a feature is ready, the team just builds it into the cars right then and there. You get whatever the latest tech is the moment your Tesla rolls out of the factory. This is one of the key differences between Tesla and legacy auto.

My Tesla standard Model Y has Swing Mode where the front air vents oscillate side to side so airflow distributes evenly throughout the car. The thing that many people do not know about Tesla is that they don’t wait for a “new model year” to add new features. They don’t sit around planning for a 2025 version or a 2026 version. When a feature is ready, the team just builds it into the cars right then and there. You get whatever the latest tech is the moment your Tesla rolls out of the factory. This is one of the key differences between Tesla and legacy auto.

Teslaconomics

159,863 次观看 • 11 个月前

Anthropic has created an AI model that “can get into almost any software in the world,” former AI company founder and CEO Matt Shumer says — but the company is not releasing it to the public: “I have never seen something like this — a company that is about to IPO and having such a moral compass.” Shumer shares what makes the AI model so powerful and potentially dangerous.

Anthropic has created an AI model that “can get into almost any software in the world,” former AI company founder and CEO Matt Shumer says — but the company is not releasing it to the public: “I have never seen something like this — a company that is about to IPO and having such a moral compass.” Shumer shares what makes the AI model so powerful and potentially dangerous.

CBS Mornings

98,892 次观看 • 6 个月前

Engram cofounder Jack Morris just raised $98M to build a new type of AI. He says models don't need to get smarter over time. Instead, they just need to know you better and better over time. Jack describes what he's building: "Our product is a new type of AI. We have a pretty different vision from a lot of the frontier labs, which are working on one model per lab, and trying to make that model smarter every month." "There's another way to think about it, which is that the model doesn't need to get smarter every month. It needs to know you better." "So we're working on a whole different stack, which is a way to train models that train themselves to know your world better and adjust to the things that you say." "So: new ways of training, new ways of running the models."

Engram cofounder Jack Morris just raised $98M to build a new type of AI. He says models don't need to get smarter over time. Instead, they just need to know you better and better over time. Jack describes what he's building: "Our product is a new type of AI. We have a pretty different vision from a lot of the frontier labs, which are working on one model per lab, and trying to make that model smarter every month." "There's another way to think about it, which is that the model doesn't need to get smarter every month. It needs to know you better." "So we're working on a whole different stack, which is a way to train models that train themselves to know your world better and adjust to the things that you say." "So: new ways of training, new ways of running the models."

TBPN

70,850 次观看 • 4 个月前