MadAIMan
@mayankvora.bsky.social
AI Newsletter Pioneer at The Shift | Get the latest breakthroughs, tools, and strategies in AI | Subscribe to my newsletter 👇🏻
https://theshiftai.beehiiv.com/subscribe
https://theshiftai.beehiiv.com/subscribe
AI slop is scaling faster than quality video, and the data is uncomfortable.
Kapwing just published a report on the rise of low-effort, auto-generated AI content built purely to farm views.
Not AI-assisted creativity. Slop.
Kapwing just published a report on the rise of low-effort, auto-generated AI content built purely to farm views.
Not AI-assisted creativity. Slop.
December 30, 2025 at 2:37 PM
AI slop is scaling faster than quality video, and the data is uncomfortable.
Kapwing just published a report on the rise of low-effort, auto-generated AI content built purely to farm views.
Not AI-assisted creativity. Slop.
Kapwing just published a report on the rise of low-effort, auto-generated AI content built purely to farm views.
Not AI-assisted creativity. Slop.
Games are quietly becoming both creation tools and training grounds.
- YouTube and NVIDIA are proving it from opposite directions.
One lowers the barrier to building games.
- The other turns gameplay into data for learning how to act.
- YouTube and NVIDIA are proving it from opposite directions.
One lowers the barrier to building games.
- The other turns gameplay into data for learning how to act.
December 24, 2025 at 3:08 PM
Games are quietly becoming both creation tools and training grounds.
- YouTube and NVIDIA are proving it from opposite directions.
One lowers the barrier to building games.
- The other turns gameplay into data for learning how to act.
- YouTube and NVIDIA are proving it from opposite directions.
One lowers the barrier to building games.
- The other turns gameplay into data for learning how to act.
OpenAI just pushed agentic coding into more serious territory.
GPT-5.2-Codex is now live across all Codex surfaces for paid ChatGPT users.
This one is clearly built for real software work, not toy demos.
GPT-5.2-Codex is now live across all Codex surfaces for paid ChatGPT users.
This one is clearly built for real software work, not toy demos.
December 22, 2025 at 2:16 PM
OpenAI just pushed agentic coding into more serious territory.
GPT-5.2-Codex is now live across all Codex surfaces for paid ChatGPT users.
This one is clearly built for real software work, not toy demos.
GPT-5.2-Codex is now live across all Codex surfaces for paid ChatGPT users.
This one is clearly built for real software work, not toy demos.
OpenAI is quietly turning ChatGPT into a platform, not just an AI product.
An App Directory just launched, the SDK is now open, and third-party tools can run directly inside the interface.
An App Directory just launched, the SDK is now open, and third-party tools can run directly inside the interface.
December 18, 2025 at 2:33 PM
OpenAI is quietly turning ChatGPT into a platform, not just an AI product.
An App Directory just launched, the SDK is now open, and third-party tools can run directly inside the interface.
An App Directory just launched, the SDK is now open, and third-party tools can run directly inside the interface.
ChatGPT Image 1.5 isn’t just another model refresh.
It’s OpenAI shifting image generation from “cool demo” to something you can actually work with.
The focus isn’t novelty. It’s reliability.
It’s OpenAI shifting image generation from “cool demo” to something you can actually work with.
The focus isn’t novelty. It’s reliability.
December 17, 2025 at 12:43 PM
ChatGPT Image 1.5 isn’t just another model refresh.
It’s OpenAI shifting image generation from “cool demo” to something you can actually work with.
The focus isn’t novelty. It’s reliability.
It’s OpenAI shifting image generation from “cool demo” to something you can actually work with.
The focus isn’t novelty. It’s reliability.
AI just hit a ceiling that bigger models alone can’t break.
Zoom’s latest result on Humanity’s Last Exam shows the future isn’t one model winning everything.
It’s multiple frontier models being orchestrated together.
Zoom’s latest result on Humanity’s Last Exam shows the future isn’t one model winning everything.
It’s multiple frontier models being orchestrated together.
December 16, 2025 at 1:52 PM
AI just hit a ceiling that bigger models alone can’t break.
Zoom’s latest result on Humanity’s Last Exam shows the future isn’t one model winning everything.
It’s multiple frontier models being orchestrated together.
Zoom’s latest result on Humanity’s Last Exam shows the future isn’t one model winning everything.
It’s multiple frontier models being orchestrated together.
Google Translate just crossed a quiet but massive threshold.
This isn’t a UI refresh. It’s a shift toward translations that actually sound human.
With Gemini powering the engine, slang, idioms, and live conversation finally translate with intent, not just words.
This isn’t a UI refresh. It’s a shift toward translations that actually sound human.
With Gemini powering the engine, slang, idioms, and live conversation finally translate with intent, not just words.
December 15, 2025 at 2:05 PM
Google Translate just crossed a quiet but massive threshold.
This isn’t a UI refresh. It’s a shift toward translations that actually sound human.
With Gemini powering the engine, slang, idioms, and live conversation finally translate with intent, not just words.
This isn’t a UI refresh. It’s a shift toward translations that actually sound human.
With Gemini powering the engine, slang, idioms, and live conversation finally translate with intent, not just words.
AI editing just took a huge leap into everyday creative workflows.
Figma is pulling Photoshop-level tools directly into its canvas, and Adobe is letting ChatGPT operate Photoshop, Acrobat, and Express with simple prompts.
Figma is pulling Photoshop-level tools directly into its canvas, and Adobe is letting ChatGPT operate Photoshop, Acrobat, and Express with simple prompts.
December 11, 2025 at 2:56 PM
AI editing just took a huge leap into everyday creative workflows.
Figma is pulling Photoshop-level tools directly into its canvas, and Adobe is letting ChatGPT operate Photoshop, Acrobat, and Express with simple prompts.
Figma is pulling Photoshop-level tools directly into its canvas, and Adobe is letting ChatGPT operate Photoshop, Acrobat, and Express with simple prompts.
AI isn’t getting adopted like software, it’s getting adopted like a dirty secret. Anthropic just interviewed 1,250 people and the story is clear, everyone’s using it, but nobody wants to be “caught” using it.
December 8, 2025 at 2:05 PM
AI isn’t getting adopted like software, it’s getting adopted like a dirty secret. Anthropic just interviewed 1,250 people and the story is clear, everyone’s using it, but nobody wants to be “caught” using it.
Mistral just fired a clean shot at Big AI. Not by building the biggest model, but by shipping the most deployable ones.
December 4, 2025 at 2:10 PM
Mistral just fired a clean shot at Big AI. Not by building the biggest model, but by shipping the most deployable ones.
DeepSeek V3.2 just made frontier reasoning feel cheap.
Two open models, V3.2 and V3.2-Speciale, claiming GPT-5 and Gemini 3 Pro level reasoning, but priced like a utility.
This is not a demo drop, it is a market pressure event.
Two open models, V3.2 and V3.2-Speciale, claiming GPT-5 and Gemini 3 Pro level reasoning, but priced like a utility.
This is not a demo drop, it is a market pressure event.
December 2, 2025 at 1:59 PM
DeepSeek V3.2 just made frontier reasoning feel cheap.
Two open models, V3.2 and V3.2-Speciale, claiming GPT-5 and Gemini 3 Pro level reasoning, but priced like a utility.
This is not a demo drop, it is a market pressure event.
Two open models, V3.2 and V3.2-Speciale, claiming GPT-5 and Gemini 3 Pro level reasoning, but priced like a utility.
This is not a demo drop, it is a market pressure event.
China just overtook the U.S. in open AI models, and almost nobody noticed.
MIT + Hugging Face analyzed 2.2B downloads. The shift is now official.
MIT + Hugging Face analyzed 2.2B downloads. The shift is now official.
December 1, 2025 at 2:37 PM
China just overtook the U.S. in open AI models, and almost nobody noticed.
MIT + Hugging Face analyzed 2.2B downloads. The shift is now official.
MIT + Hugging Face analyzed 2.2B downloads. The shift is now official.
The Shock Hook
The U.S. just launched the biggest AI science project since the Manhattan Project.
- Trump’s new Genesis Mission is a plan to fuse NASA, NIH, DoD, and 17 National Labs into one AI-powered discovery machine.
The U.S. just launched the biggest AI science project since the Manhattan Project.
- Trump’s new Genesis Mission is a plan to fuse NASA, NIH, DoD, and 17 National Labs into one AI-powered discovery machine.
November 27, 2025 at 1:40 PM
The Shock Hook
The U.S. just launched the biggest AI science project since the Manhattan Project.
- Trump’s new Genesis Mission is a plan to fuse NASA, NIH, DoD, and 17 National Labs into one AI-powered discovery machine.
The U.S. just launched the biggest AI science project since the Manhattan Project.
- Trump’s new Genesis Mission is a plan to fuse NASA, NIH, DoD, and 17 National Labs into one AI-powered discovery machine.
Claude Opus 4.5 Is Here, And It’s a Monster Upgrade
Anthropic just dropped its new flagship model, and it’s a huge leap in coding, reasoning, and long-form agentic execution.
Anthropic just dropped its new flagship model, and it’s a huge leap in coding, reasoning, and long-form agentic execution.
November 25, 2025 at 5:24 PM
Claude Opus 4.5 Is Here, And It’s a Monster Upgrade
Anthropic just dropped its new flagship model, and it’s a huge leap in coding, reasoning, and long-form agentic execution.
Anthropic just dropped its new flagship model, and it’s a huge leap in coding, reasoning, and long-form agentic execution.
OpenAI just dropped GPT-5.1-Codex-Max, the first agentic coding model built for multi-hour real engineering work.
Not a chat model. Not a code autocomplete.
Not a chat model. Not a code autocomplete.
November 20, 2025 at 1:39 PM
OpenAI just dropped GPT-5.1-Codex-Max, the first agentic coding model built for multi-hour real engineering work.
Not a chat model. Not a code autocomplete.
Not a chat model. Not a code autocomplete.
Google just launched Gemini 3, its most advanced AI yet.
Big jumps in reasoning, multimodality, accuracy, and agentic automation.
And it rolls out everywhere on day one: Search, Gemini app, and dev tools.
Big jumps in reasoning, multimodality, accuracy, and agentic automation.
And it rolls out everywhere on day one: Search, Gemini app, and dev tools.
November 19, 2025 at 2:23 PM
Google just launched Gemini 3, its most advanced AI yet.
Big jumps in reasoning, multimodality, accuracy, and agentic automation.
And it rolls out everywhere on day one: Search, Gemini app, and dev tools.
Big jumps in reasoning, multimodality, accuracy, and agentic automation.
And it rolls out everywhere on day one: Search, Gemini app, and dev tools.
Grok 4.1 quietly rolled out over the past two weeks… and it just took the top spot on LM Arena.
xAI pushed the update live across X and grok.com, where it won 64.78% of blind comparisons over the previous version.
This release signals a clear shift: Grok is no longer the underdog.
xAI pushed the update live across X and grok.com, where it won 64.78% of blind comparisons over the previous version.
This release signals a clear shift: Grok is no longer the underdog.
November 18, 2025 at 2:33 PM
Grok 4.1 quietly rolled out over the past two weeks… and it just took the top spot on LM Arena.
xAI pushed the update live across X and grok.com, where it won 64.78% of blind comparisons over the previous version.
This release signals a clear shift: Grok is no longer the underdog.
xAI pushed the update live across X and grok.com, where it won 64.78% of blind comparisons over the previous version.
This release signals a clear shift: Grok is no longer the underdog.
OpenAI just shipped GPT-5.1, a refinement of GPT-5 built for sharper reasoning, cleaner explanations, and more personal communication.
- It introduces two upgraded models and brand-new tone controls shaped by feedback from 800M users.
- It introduces two upgraded models and brand-new tone controls shaped by feedback from 800M users.
November 13, 2025 at 2:27 PM
OpenAI just shipped GPT-5.1, a refinement of GPT-5 built for sharper reasoning, cleaner explanations, and more personal communication.
- It introduces two upgraded models and brand-new tone controls shaped by feedback from 800M users.
- It introduces two upgraded models and brand-new tone controls shaped by feedback from 800M users.
ElevenLabs just made AI voices legal.
Their new Iconic Voice Marketplace lets brands license AI-generated voices of celebrities and historical icons, all with full consent and fair pay.
The age of ethical cloning has officially begun.
Their new Iconic Voice Marketplace lets brands license AI-generated voices of celebrities and historical icons, all with full consent and fair pay.
The age of ethical cloning has officially begun.
November 12, 2025 at 2:29 PM
ElevenLabs just made AI voices legal.
Their new Iconic Voice Marketplace lets brands license AI-generated voices of celebrities and historical icons, all with full consent and fair pay.
The age of ethical cloning has officially begun.
Their new Iconic Voice Marketplace lets brands license AI-generated voices of celebrities and historical icons, all with full consent and fair pay.
The age of ethical cloning has officially begun.
Humans vs AI agents: same tasks, totally different thinking.
In the largest real-world test yet, 48 people went head-to-head with ChatGPT, Claude, and OpenHands on 16 jobs, from coding to design to analysis.
The results? AI was faster and cheaper, but half as accurate.
In the largest real-world test yet, 48 people went head-to-head with ChatGPT, Claude, and OpenHands on 16 jobs, from coding to design to analysis.
The results? AI was faster and cheaper, but half as accurate.
November 11, 2025 at 1:53 PM
Humans vs AI agents: same tasks, totally different thinking.
In the largest real-world test yet, 48 people went head-to-head with ChatGPT, Claude, and OpenHands on 16 jobs, from coding to design to analysis.
The results? AI was faster and cheaper, but half as accurate.
In the largest real-world test yet, 48 people went head-to-head with ChatGPT, Claude, and OpenHands on 16 jobs, from coding to design to analysis.
The results? AI was faster and cheaper, but half as accurate.
AI is everywhere, but barely working.
- McKinsey’s 2025 report shows 88% of companies use AI, yet only 1 in 3 have scaled it beyond pilots.
- The rest are stuck experimenting while agentic AI quietly rewires workflows behind the scenes.
- McKinsey’s 2025 report shows 88% of companies use AI, yet only 1 in 3 have scaled it beyond pilots.
- The rest are stuck experimenting while agentic AI quietly rewires workflows behind the scenes.
November 10, 2025 at 1:55 PM
AI is everywhere, but barely working.
- McKinsey’s 2025 report shows 88% of companies use AI, yet only 1 in 3 have scaled it beyond pilots.
- The rest are stuck experimenting while agentic AI quietly rewires workflows behind the scenes.
- McKinsey’s 2025 report shows 88% of companies use AI, yet only 1 in 3 have scaled it beyond pilots.
- The rest are stuck experimenting while agentic AI quietly rewires workflows behind the scenes.
Google just put AI in orbit.
- Project Suncatcher aims to run large-scale AI workloads… in space.
- Solar-powered satellites with Google TPUs will process data above Earth’s grid, using 8× more sunlight and near-zero downtime.
- Project Suncatcher aims to run large-scale AI workloads… in space.
- Solar-powered satellites with Google TPUs will process data above Earth’s grid, using 8× more sunlight and near-zero downtime.
November 5, 2025 at 2:41 PM
Google just put AI in orbit.
- Project Suncatcher aims to run large-scale AI workloads… in space.
- Solar-powered satellites with Google TPUs will process data above Earth’s grid, using 8× more sunlight and near-zero downtime.
- Project Suncatcher aims to run large-scale AI workloads… in space.
- Solar-powered satellites with Google TPUs will process data above Earth’s grid, using 8× more sunlight and near-zero downtime.
OpenAI Expands Beyond Microsoft
- OpenAI just inked a 7-year, $38B partnership with AWS, its first major move outside Microsoft’s cloud.
- The deal unlocks new compute capacity worldwide, powered by AWS UltraServers using NVIDIA GB200 and GB300 GPUs.
- OpenAI just inked a 7-year, $38B partnership with AWS, its first major move outside Microsoft’s cloud.
- The deal unlocks new compute capacity worldwide, powered by AWS UltraServers using NVIDIA GB200 and GB300 GPUs.
November 4, 2025 at 1:36 PM
OpenAI Expands Beyond Microsoft
- OpenAI just inked a 7-year, $38B partnership with AWS, its first major move outside Microsoft’s cloud.
- The deal unlocks new compute capacity worldwide, powered by AWS UltraServers using NVIDIA GB200 and GB300 GPUs.
- OpenAI just inked a 7-year, $38B partnership with AWS, its first major move outside Microsoft’s cloud.
- The deal unlocks new compute capacity worldwide, powered by AWS UltraServers using NVIDIA GB200 and GB300 GPUs.
1/ Inside the Brockman Memo
- A 52-page internal document detailed how OpenAI’s board planned to remove Sam Altman a year before the 2023 crisis.
- It accused him of manipulation, and even explored a merger with Anthropic, led by Dario Amodei.
- A 52-page internal document detailed how OpenAI’s board planned to remove Sam Altman a year before the 2023 crisis.
- It accused him of manipulation, and even explored a merger with Anthropic, led by Dario Amodei.
November 3, 2025 at 4:04 PM
1/ Inside the Brockman Memo
- A 52-page internal document detailed how OpenAI’s board planned to remove Sam Altman a year before the 2023 crisis.
- It accused him of manipulation, and even explored a merger with Anthropic, led by Dario Amodei.
- A 52-page internal document detailed how OpenAI’s board planned to remove Sam Altman a year before the 2023 crisis.
- It accused him of manipulation, and even explored a merger with Anthropic, led by Dario Amodei.
The 2025 State of Generative Media report just dropped, and Google’s officially on top.
Across 300 creators and enterprises, Gemini and Veo now dominate image and video generation alike.
It’s not just leadership, it’s ecosystem lock-in.
Across 300 creators and enterprises, Gemini and Veo now dominate image and video generation alike.
It’s not just leadership, it’s ecosystem lock-in.
October 30, 2025 at 2:39 PM
The 2025 State of Generative Media report just dropped, and Google’s officially on top.
Across 300 creators and enterprises, Gemini and Veo now dominate image and video generation alike.
It’s not just leadership, it’s ecosystem lock-in.
Across 300 creators and enterprises, Gemini and Veo now dominate image and video generation alike.
It’s not just leadership, it’s ecosystem lock-in.