#compute — Public Fediverse posts

"Unfortunately, it’s difficult to make a crisp comparison, but the proxies that we have suggest that demand is growing much faster. For instance, both the quantity of tokens processed by Google in the last year, and by all providers according to Exponential View, have been growing by around 10×/year.

From another angle, we can look at token demand from today’s most intensive AI users: software engineers. Recent reports claim that some of Apple’s software engineers are permitted to use up to $300 in tokens per day, which works out to about 5 million output tokens per day with Claude Opus 4.7 API pricing, or 25 million output tokens per day with Kimi K2.6.16 Another point of comparison comes from Meta, whose 85,000 employees used 60 trillion tokens in one month across the organization. That figure included both input and output tokens; assuming a 25,000:1,000 input-to-output token ratio, that would be around 1 million output tokens per day and employee.

There were about 30 million software engineers worldwide as of 2025 (estimates range from 20 million to 50 million), and Stack Overflow’s 2025 survey on AI usage suggested that only around 47% of developers used AI on a daily basis, as of mid-2025. If all SWEs using AI daily were using it as intensely as Meta or Apple, they would demand somewhere between 10 and 350 trillion tokens per day in aggregate, i.e., between 200 million and 4 billion tokens per second. At the longest context sizes of 128,000:1,000, today’s Blackwell chips would struggle to serve all this potential demand for coding agents using models as large as Kimi K2.6. It also seems likely that both the number of developers using AI, and the intensity of their use will continue to grow rapidly."

https://epoch.ai/gradient-updates/is-a-compute-crunch-coming

#AI #GenerativeAI #AIBubble #Compute #AIHype #DataCenters #Inference

"Unfortunately, it’s difficult to make a crisp comparison, but the proxies that we have suggest that demand is growing much faster. For instance, both the quantity of tokens processed by Google in the last year, and by all providers according to Exponential View, have been growing by around 10×/year.

From another angle, we can look at token demand from today’s most intensive AI users: software engineers. Recent reports claim that some of Apple’s software engineers are permitted to use up to $300 in tokens per day, which works out to about 5 million output tokens per day with Claude Opus 4.7 API pricing, or 25 million output tokens per day with Kimi K2.6.16 Another point of comparison comes from Meta, whose 85,000 employees used 60 trillion tokens in one month across the organization. That figure included both input and output tokens; assuming a 25,000:1,000 input-to-output token ratio, that would be around 1 million output tokens per day and employee.

There were about 30 million software engineers worldwide as of 2025 (estimates range from 20 million to 50 million), and Stack Overflow’s 2025 survey on AI usage suggested that only around 47% of developers used AI on a daily basis, as of mid-2025. If all SWEs using AI daily were using it as intensely as Meta or Apple, they would demand somewhere between 10 and 350 trillion tokens per day in aggregate, i.e., between 200 million and 4 billion tokens per second. At the longest context sizes of 128,000:1,000, today’s Blackwell chips would struggle to serve all this potential demand for coding agents using models as large as Kimi K2.6. It also seems likely that both the number of developers using AI, and the intensity of their use will continue to grow rapidly."

https://epoch.ai/gradient-updates/is-a-compute-crunch-coming

#AI #GenerativeAI #AIBubble #Compute #AIHype #DataCenters #Inference

"Unfortunately, it’s difficult to make a crisp comparison, but the proxies that we have suggest that demand is growing much faster. For instance, both the quantity of tokens processed by Google in the last year, and by all providers according to Exponential View, have been growing by around 10×/year.

From another angle, we can look at token demand from today’s most intensive AI users: software engineers. Recent reports claim that some of Apple’s software engineers are permitted to use up to $300 in tokens per day, which works out to about 5 million output tokens per day with Claude Opus 4.7 API pricing, or 25 million output tokens per day with Kimi K2.6.16 Another point of comparison comes from Meta, whose 85,000 employees used 60 trillion tokens in one month across the organization. That figure included both input and output tokens; assuming a 25,000:1,000 input-to-output token ratio, that would be around 1 million output tokens per day and employee.

There were about 30 million software engineers worldwide as of 2025 (estimates range from 20 million to 50 million), and Stack Overflow’s 2025 survey on AI usage suggested that only around 47% of developers used AI on a daily basis, as of mid-2025. If all SWEs using AI daily were using it as intensely as Meta or Apple, they would demand somewhere between 10 and 350 trillion tokens per day in aggregate, i.e., between 200 million and 4 billion tokens per second. At the longest context sizes of 128,000:1,000, today’s Blackwell chips would struggle to serve all this potential demand for coding agents using models as large as Kimi K2.6. It also seems likely that both the number of developers using AI, and the intensity of their use will continue to grow rapidly."

https://epoch.ai/gradient-updates/is-a-compute-crunch-coming

#AI #GenerativeAI #AIBubble #Compute #AIHype #DataCenters #Inference

"Unfortunately, it’s difficult to make a crisp comparison, but the proxies that we have suggest that demand is growing much faster. For instance, both the quantity of tokens processed by Google in the last year, and by all providers according to Exponential View, have been growing by around 10×/year.

From another angle, we can look at token demand from today’s most intensive AI users: software engineers. Recent reports claim that some of Apple’s software engineers are permitted to use up to $300 in tokens per day, which works out to about 5 million output tokens per day with Claude Opus 4.7 API pricing, or 25 million output tokens per day with Kimi K2.6.16 Another point of comparison comes from Meta, whose 85,000 employees used 60 trillion tokens in one month across the organization. That figure included both input and output tokens; assuming a 25,000:1,000 input-to-output token ratio, that would be around 1 million output tokens per day and employee.

There were about 30 million software engineers worldwide as of 2025 (estimates range from 20 million to 50 million), and Stack Overflow’s 2025 survey on AI usage suggested that only around 47% of developers used AI on a daily basis, as of mid-2025. If all SWEs using AI daily were using it as intensely as Meta or Apple, they would demand somewhere between 10 and 350 trillion tokens per day in aggregate, i.e., between 200 million and 4 billion tokens per second. At the longest context sizes of 128,000:1,000, today’s Blackwell chips would struggle to serve all this potential demand for coding agents using models as large as Kimi K2.6. It also seems likely that both the number of developers using AI, and the intensity of their use will continue to grow rapidly."

https://epoch.ai/gradient-updates/is-a-compute-crunch-coming

#AI #GenerativeAI #AIBubble #Compute #AIHype #DataCenters #Inference

"Unfortunately, it’s difficult to make a crisp comparison, but the proxies that we have suggest that demand is growing much faster. For instance, both the quantity of tokens processed by Google in the last year, and by all providers according to Exponential View, have been growing by around 10×/year.

From another angle, we can look at token demand from today’s most intensive AI users: software engineers. Recent reports claim that some of Apple’s software engineers are permitted to use up to $300 in tokens per day, which works out to about 5 million output tokens per day with Claude Opus 4.7 API pricing, or 25 million output tokens per day with Kimi K2.6.16 Another point of comparison comes from Meta, whose 85,000 employees used 60 trillion tokens in one month across the organization. That figure included both input and output tokens; assuming a 25,000:1,000 input-to-output token ratio, that would be around 1 million output tokens per day and employee.

There were about 30 million software engineers worldwide as of 2025 (estimates range from 20 million to 50 million), and Stack Overflow’s 2025 survey on AI usage suggested that only around 47% of developers used AI on a daily basis, as of mid-2025. If all SWEs using AI daily were using it as intensely as Meta or Apple, they would demand somewhere between 10 and 350 trillion tokens per day in aggregate, i.e., between 200 million and 4 billion tokens per second. At the longest context sizes of 128,000:1,000, today’s Blackwell chips would struggle to serve all this potential demand for coding agents using models as large as Kimi K2.6. It also seems likely that both the number of developers using AI, and the intensity of their use will continue to grow rapidly."

https://epoch.ai/gradient-updates/is-a-compute-crunch-coming

#AI #GenerativeAI #AIBubble #Compute #AIHype #DataCenters #Inference

Microsoft is canceling most Claude Code licenses for developers and steering staff to GitHub Copilot CLI after internal AI coding costs surged. 💸
Uber exhausted its 2026 AI coding budget in four months as rising token use exposed compute costs that can exceed employee pay. ⚙️

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #Anthropic #Claude #ClaudeCode #GitHub #GitHubCopilot #AI #ArtificialIntelligence #Copilot #LLM #Privacy #FOSS #Cloud #Developers #Automation #Compute #Enterprise #Uber

Microsoft is canceling most Claude Code licenses for developers and steering staff to GitHub Copilot CLI after internal AI coding costs surged. 💸
Uber exhausted its 2026 AI coding budget in four months as rising token use exposed compute costs that can exceed employee pay. ⚙️

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #Anthropic #Claude #ClaudeCode #GitHub #GitHubCopilot #AI #ArtificialIntelligence #Copilot #LLM #Privacy #FOSS #Cloud #Developers #Automation #Compute #Enterprise #Uber

Microsoft is canceling most Claude Code licenses for developers and steering staff to GitHub Copilot CLI after internal AI coding costs surged. 💸
Uber exhausted its 2026 AI coding budget in four months as rising token use exposed compute costs that can exceed employee pay. ⚙️

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #Anthropic #Claude #ClaudeCode #GitHub #GitHubCopilot #AI #ArtificialIntelligence #Copilot #LLM #Privacy #FOSS #Cloud #Developers #Automation #Compute #Enterprise #Uber

Microsoft is canceling most Claude Code licenses for developers and steering staff to GitHub Copilot CLI after internal AI coding costs surged. 💸
Uber exhausted its 2026 AI coding budget in four months as rising token use exposed compute costs that can exceed employee pay. ⚙️

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #Anthropic #Claude #ClaudeCode #GitHub #GitHubCopilot #AI #ArtificialIntelligence #Copilot #LLM #Privacy #FOSS #Cloud #Developers #Automation #Compute #Enterprise #Uber

Microsoft is canceling most Claude Code licenses for developers and steering staff to GitHub Copilot CLI after internal AI coding costs surged. 💸
Uber exhausted its 2026 AI coding budget in four months as rising token use exposed compute costs that can exceed employee pay. ⚙️

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #Anthropic #Claude #ClaudeCode #GitHub #GitHubCopilot #AI #ArtificialIntelligence #Copilot #LLM #Privacy #FOSS #Cloud #Developers #Automation #Compute #Enterprise #Uber

Microsoft is reportedly ending most Claude Code licenses after AI coding costs surged, shifting staff to GitHub Copilot CLI 💸
Uber exhausted its 2026 AI coding budget in four months, showing token-heavy AI workflows can outpace labor savings and raise compute dependence 🔓

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #ClaudeCode #GitHubCopilot #Anthropic #AI #GenAI #LLM #OpenSource #Privacy #Cloud #Compute #Automation #Developers #BigTech #ArtificialIntelligence #Claude #GitHub #Coding #Uber

Microsoft is reportedly ending most Claude Code licenses after AI coding costs surged, shifting staff to GitHub Copilot CLI 💸
Uber exhausted its 2026 AI coding budget in four months, showing token-heavy AI workflows can outpace labor savings and raise compute dependence 🔓

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #ClaudeCode #GitHubCopilot #Anthropic #AI #GenAI #LLM #OpenSource #Privacy #Cloud #Compute #Automation #Developers #BigTech #ArtificialIntelligence #Claude #GitHub #Coding #Uber

Microsoft is reportedly ending most Claude Code licenses after AI coding costs surged, shifting staff to GitHub Copilot CLI 💸
Uber exhausted its 2026 AI coding budget in four months, showing token-heavy AI workflows can outpace labor savings and raise compute dependence 🔓

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #ClaudeCode #GitHubCopilot #Anthropic #AI #GenAI #LLM #OpenSource #Privacy #Cloud #Compute #Automation #Developers #BigTech #ArtificialIntelligence #Claude #GitHub #Coding #Uber

Microsoft is reportedly ending most Claude Code licenses after AI coding costs surged, shifting staff to GitHub Copilot CLI 💸
Uber exhausted its 2026 AI coding budget in four months, showing token-heavy AI workflows can outpace labor savings and raise compute dependence 🔓

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #ClaudeCode #GitHubCopilot #Anthropic #AI #GenAI #LLM #OpenSource #Privacy #Cloud #Compute #Automation #Developers #BigTech #ArtificialIntelligence #Claude #GitHub #Coding #Uber

Microsoft is reportedly ending most Claude Code licenses after AI coding costs surged, shifting staff to GitHub Copilot CLI 💸
Uber exhausted its 2026 AI coding budget in four months, showing token-heavy AI workflows can outpace labor savings and raise compute dependence 🔓

🔗 https://fortune.com/2026/05/22/microsoft-ai-cost-problem-tokens-agents/

#TechNews #Microsoft #ClaudeCode #GitHubCopilot #Anthropic #AI #GenAI #LLM #OpenSource #Privacy #Cloud #Compute #Automation #Developers #BigTech #ArtificialIntelligence #Claude #GitHub #Coding #Uber

#xAI is flexing its #compute… while everyone scrambles to have enough to run their models, they sign deals with #Anthropic and #cursor while giving #X Premium subscribers more access via #OpenClaw and #opencode… it’s the under appreciated AI player. https://x.ai/news/grok-opencode

The orchestrator touches across all layers. With its open-source approach, with: →the integrated #datacatalog that lists/surfaces all →#explicitly define assets →resources for integrating #compute as well as #storage →integrating best practices →general #automations →#governed data platform

@OrionKidder

There will be no #aibubble collapse.

Not in the sense most #Luddites are praying for.

Sure at some stage there will be market consolidation, but there will be no tumbleweeds rolling through abandoned data centres.

In the short term, the #compute will he sold for cents on the dollar, the compute market is already commoditfying.

Instead of multiple #Ai firms, there will emerge 2-3 monoliths, in the same way #dotcom crash left dark fibre in the ground and gave us google and Facebook.

@OrionKidder

There will be no #aibubble collapse.

Not in the sense most #Luddites are praying for.

Sure at some stage there will be market consolidation, but there will be no tumbleweeds rolling through abandoned data centres.

In the short term, the #compute will he sold for cents on the dollar, the compute market is already commoditfying.

Instead of multiple #Ai firms, there will emerge 2-3 monoliths, in the same way #dotcom crash left dark fibre in the ground and gave us google and Facebook.

@OrionKidder

There will be no #aibubble collapse.

Not in the sense most #Luddites are praying for.

Sure at some stage there will be market consolidation, but there will be no tumbleweeds rolling through abandoned data centres.

In the short term, the #compute will he sold for cents on the dollar, the compute market is already commoditfying.

Instead of multiple #Ai firms, there will emerge 2-3 monoliths, in the same way #dotcom crash left dark fibre in the ground and gave us google and Facebook.

Needle: We Distilled Gemini Tool Calling into a 26M Model

https://github.com/cactus-compute/needle

#HackerNews #Needle #Gemini #Tool #Model #AI #Distillation #Cactus #Compute

Needle: We Distilled Gemini Tool Calling into a 26M Model

https://github.com/cactus-compute/needle

#HackerNews #Needle #Gemini #Tool #Model #AI #Distillation #Cactus #Compute

Needle: We Distilled Gemini Tool Calling into a 26M Model

https://github.com/cactus-compute/needle

#HackerNews #Needle #Gemini #Tool #Model #AI #Distillation #Cactus #Compute

Needle: We Distilled Gemini Tool Calling into a 26M Model

https://github.com/cactus-compute/needle

#HackerNews #Needle #Gemini #Tool #Model #AI #Distillation #Cactus #Compute

Needle: We Distilled Gemini Tool Calling into a 26M Model

https://github.com/cactus-compute/needle

#HackerNews #Needle #Gemini #Tool #Model #AI #Distillation #Cactus #Compute

https://www.europesays.com/people/61825/ DOD planning to address compute ‘bottleneck’ that could hinder AI proliferation #ArtificialIntelligence(ai) #CameronStanley #cdao #compute #DefenseSecretaryPeteHegseth #EmilMichael #IranWar #MavenSmartSystem #OperationEpicFury #PeteHegseth

In my #homelab I have a mix of #arm ( #raspberrypi ) and #intel / #amd #compute nodes. I have never actually tested if the VPC #network works between the two architectures. Decided to test it today, seems to work fine.

Below are two #selfhosted #vm on different bare metals with a shared VPC network using iperf.

#technology #virtualization #linux

In my #homelab I have a mix of #arm ( #raspberrypi ) and #intel / #amd #compute nodes. I have never actually tested if the VPC #network works between the two architectures. Decided to test it today, seems to work fine.

Below are two #selfhosted #vm on different bare metals with a shared VPC network using iperf.

#technology #virtualization #linux

In my #homelab I have a mix of #arm ( #raspberrypi ) and #intel / #amd #compute nodes. I have never actually tested if the VPC #network works between the two architectures. Decided to test it today, seems to work fine.

Below are two #selfhosted #vm on different bare metals with a shared VPC network using iperf.

#technology #virtualization #linux

In my #homelab I have a mix of #arm ( #raspberrypi ) and #intel / #amd #compute nodes. I have never actually tested if the VPC #network works between the two architectures. Decided to test it today, seems to work fine.

Below are two #selfhosted #vm on different bare metals with a shared VPC network using iperf.

#technology #virtualization #linux

In my #homelab I have a mix of #arm ( #raspberrypi ) and #intel / #amd #compute nodes. I have never actually tested if the VPC #network works between the two architectures. Decided to test it today, seems to work fine.

Below are two #selfhosted #vm on different bare metals with a shared VPC network using iperf.

#technology #virtualization #linux

DOD planning to address compute ‘bottleneck’ that could hinder AI proliferation

The Pentagon is preparing to take additional steps to address a major bottleneck that could limit the military’s…
#UnitedStates #US #USA #artificialintelligenceAI #cameronstanley #cdao #compute #DefenseSecretaryPeteHegseth #Død #emilmichael #iranwar #MavenSmartSystem #OperationEpicFury #pentagon #petehegseth #SecretaryofDefense
https://www.europesays.com/2973312/

Wow, a #compute #deal with SpaceX! 🚀 Because nothing says "cutting-edge tech" like more #cloud minutes for Claude to contemplate its existence. Meanwhile, Bezos is crying into his #BezosBot at not being invited to this party of zeros and ones. 😂
https://www.anthropic.com/news/higher-limits-spacex #SpaceX #computing #tech #humor #HackerNews #ngated

Wow, a #compute #deal with SpaceX! 🚀 Because nothing says "cutting-edge tech" like more #cloud minutes for Claude to contemplate its existence. Meanwhile, Bezos is crying into his #BezosBot at not being invited to this party of zeros and ones. 😂
https://www.anthropic.com/news/higher-limits-spacex #SpaceX #computing #tech #humor #HackerNews #ngated

Wow, a #compute #deal with SpaceX! 🚀 Because nothing says "cutting-edge tech" like more #cloud minutes for Claude to contemplate its existence. Meanwhile, Bezos is crying into his #BezosBot at not being invited to this party of zeros and ones. 😂
https://www.anthropic.com/news/higher-limits-spacex #SpaceX #computing #tech #humor #HackerNews #ngated

Wow, a #compute #deal with SpaceX! 🚀 Because nothing says "cutting-edge tech" like more #cloud minutes for Claude to contemplate its existence. Meanwhile, Bezos is crying into his #BezosBot at not being invited to this party of zeros and ones. 😂
https://www.anthropic.com/news/higher-limits-spacex #SpaceX #computing #tech #humor #HackerNews #ngated

Wow, a #compute #deal with SpaceX! 🚀 Because nothing says "cutting-edge tech" like more #cloud minutes for Claude to contemplate its existence. Meanwhile, Bezos is crying into his #BezosBot at not being invited to this party of zeros and ones. 😂
https://www.anthropic.com/news/higher-limits-spacex #SpaceX #computing #tech #humor #HackerNews #ngated

OpenCL 3.1 is here.

The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products.

Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries.

Implementations are already underway across major vendors and open source projects.

- Full Blog: https://www.khronos.org/blog/opencl-3.1-is-here?utm_medium=social&utm_source=mastodon&utm_campaign=OpenCL_3.1_is_here&utm_content=blog
- OpenCL specification GitHub
- Khronos Discord

#OpenCL #HPC #GPU #Compute #SPIRV

OpenCL 3.1 is here.

The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products.

Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries.

Implementations are already underway across major vendors and open source projects.

- Full Blog: https://www.khronos.org/blog/opencl-3.1-is-here?utm_medium=social&utm_source=mastodon&utm_campaign=OpenCL_3.1_is_here&utm_content=blog
- OpenCL specification GitHub
- Khronos Discord

#OpenCL #HPC #GPU #Compute #SPIRV

OpenCL 3.1 is here.

The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products.

Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries.

Implementations are already underway across major vendors and open source projects.

- Full Blog: https://www.khronos.org/blog/opencl-3.1-is-here?utm_medium=social&utm_source=mastodon&utm_campaign=OpenCL_3.1_is_here&utm_content=blog
- OpenCL specification GitHub
- Khronos Discord

#OpenCL #HPC #GPU #Compute #SPIRV

OpenCL 3.1 is here.

The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products.

Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries.

Implementations are already underway across major vendors and open source projects.

- Full Blog: https://www.khronos.org/blog/opencl-3.1-is-here?utm_medium=social&utm_source=mastodon&utm_campaign=OpenCL_3.1_is_here&utm_content=blog
- OpenCL specification GitHub
- Khronos Discord

#OpenCL #HPC #GPU #Compute #SPIRV

OpenCL 3.1 is here.

The Khronos Group has moved several capabilities into the core spec, including SPIR-V kernels, subgroups, and integer dot products.

Also includes improvements to the memory model and synchronization, plus better alignment with Vulkan via device UUID queries.

Implementations are already underway across major vendors and open source projects.

- Full Blog: https://www.khronos.org/blog/opencl-3.1-is-here?utm_medium=social&utm_source=mastodon&utm_campaign=OpenCL_3.1_is_here&utm_content=blog
- OpenCL specification GitHub
- Khronos Discord

#OpenCL #HPC #GPU #Compute #SPIRV

"AI policy expert Lennart Heim is a useful guide to this machinery. He formerly led compute research at the RAND Center on AI, Security, and Technology and cofounded Epoch AI, which tracks the resources behind frontier AI models. His beat is where a cloud dashboard becomes a construction project—where digital demand collides with factories, transformers, chips and cables.

[An edited transcript of the telephone interview follows.]

Developers are saying the rate limits and blocked third-party tools look like a compute crunch. What does a compute shortage actually mean?

When we say “compute,” we mean computing power. For AI, training compute scales with model size: bigger neural networks need more data, and more data needs more processing power. What was underreported for years is that the same relationship holds for deployment. Running the model for users—inference—is incredibly compute-intensive because bigger models need more computing power to serve. So if more people use AI with more tokens and more intensity, you need more compute. If 10 times more people use AI 10 times more heavily, you need close to 100 times more compute.

Why does a flat-rate subscription break down for AI in a way it didn’t for earlier Internet services?

The Internet runs on flat-rate subscriptions: you pay $20 a month and get effectively unlimited use. That works when the marginal cost per user is low—a Google Workspace power user doesn’t cost Google much more than a light user. With AI, it breaks. Using AI 10 times more heavily costs the provider roughly 10 times more money. Paying per token means you literally pay for your resources; paying $20 flat means you’re often burning more compute than $20 can buy. That’s why we see rate limits mostly on monthly subscription plans. At some point, you have to rate limit."

https://www.scientificamerican.com/article/what-is-the-ai-compute-crunch-and-why-are-ai-tools-hitting-usage-limits/

#AI #GenerativeAI #DataCenters #Compute

April issue is out!

I’m tweaking the newsletter to better match what Python data engineers care about: less visualization, more data munging.

So expect a bit less Shiny and Streamlit, and more ADBC, SQLGlot, and practical ecosystem updates.

https://pythondataeng.substack.com/p/monthly-python-data-engineering-april-4e5

#python #dataeng #datascience #data #ml #machineLearning #database #compute

April issue is out!

I’m tweaking the newsletter to better match what Python data engineers care about: less visualization, more data munging.

So expect a bit less Shiny and Streamlit, and more ADBC, SQLGlot, and practical ecosystem updates.

https://pythondataeng.substack.com/p/monthly-python-data-engineering-april-4e5

#python #dataeng #datascience #data #ml #machineLearning #database #compute

April issue is out!

I’m tweaking the newsletter to better match what Python data engineers care about: less visualization, more data munging.

So expect a bit less Shiny and Streamlit, and more ADBC, SQLGlot, and practical ecosystem updates.

https://pythondataeng.substack.com/p/monthly-python-data-engineering-april-4e5

#python #dataeng #datascience #data #ml #machineLearning #database #compute

April issue is out!

I’m tweaking the newsletter to better match what Python data engineers care about: less visualization, more data munging.

So expect a bit less Shiny and Streamlit, and more ADBC, SQLGlot, and practical ecosystem updates.

https://pythondataeng.substack.com/p/monthly-python-data-engineering-april-4e5

#python #dataeng #datascience #data #ml #machineLearning #database #compute

April issue is out!

I’m tweaking the newsletter to better match what Python data engineers care about: less visualization, more data munging.

So expect a bit less Shiny and Streamlit, and more ADBC, SQLGlot, and practical ecosystem updates.

https://pythondataeng.substack.com/p/monthly-python-data-engineering-april-4e5

#python #dataeng #datascience #data #ml #machineLearning #database #compute

Stuff are connecting ! #shader #bonzomatic #compute