📰 Top 5 AI stories of the week
1. Stanford's course on how ChatGPT works is free on YouTube
Last year's nine lectures are online in full, with the exams and solutions. The first one has nearly a million views.
2. Obsidian's CEO and Cloudflare shared their AI coding setups
Five free skill packs for Claude Code and Codex. The most popular comes from a TypeScript teacher and has 319K GitHub stars.
3. 14,000 people saved a guide to a free AI with refusals removed
It runs on your own PC, and its model page says it refused 0 of 100 harmful test prompts. The author still uses cloud models for real
1. Stanford's course on how ChatGPT works is free on YouTube
Last year's nine lectures are online in full, with the exams and solutions. The first one has nearly a million views.
2. Obsidian's CEO and Cloudflare shared their AI coding setups
Five free skill packs for Claude Code and Codex. The most popular comes from a TypeScript teacher and has 319K GitHub stars.
3. 14,000 people saved a guide to a free AI with refusals removed
It runs on your own PC, and its model page says it refused 0 of 100 harmful test prompts. The author still uses cloud models for real
💰 ChatGPT's agent can now shop for you without ever seeing your card number
OpenAI's agents in ChatGPT, called dots, can now finish a purchase for you. Ask yours to find a new sweater or book a hotel for next weekend, and at checkout it pays through the new ChatGPT Wallet.
You add a card once. Before each purchase you see the items, the total and the card, then approve it with Face ID or a fingerprint.
The agent and the shop never get your real card details. Instead, the card network provides a stand-in card number for that payment. Your card's rewards and dispute protections stay.
For now
OpenAI's agents in ChatGPT, called dots, can now finish a purchase for you. Ask yours to find a new sweater or book a hotel for next weekend, and at checkout it pays through the new ChatGPT Wallet.
You add a card once. Before each purchase you see the items, the total and the card, then approve it with Face ID or a fingerprint.
The agent and the shop never get your real card details. Instead, the card network provides a stand-in card number for that payment. Your card's rewards and dispute protections stay.
For now
🧐 From zero to your own ChatGPT: Andrej Karpathy's free code, in 7 repos
The OpenAI co-founder keeps his teaching code free on GitHub. A developer lined up seven of his repos in the order to learn them, and nearly 1,500 people saved the list:
1. nn-zero-to-hero, his video course that builds neural networks from scratch
2. micrograd, the math that makes AI learn, in a tiny amount of code
3. makemore, feed it a list of names and it invents new ones
4. minbpe, how AI chops your words into tokens
5. nanoGPT, train your own GPT model
6. nanochat, in Karpathy's words "The best ChatGPT that $100 ca
The OpenAI co-founder keeps his teaching code free on GitHub. A developer lined up seven of his repos in the order to learn them, and nearly 1,500 people saved the list:
1. nn-zero-to-hero, his video course that builds neural networks from scratch
2. micrograd, the math that makes AI learn, in a tiny amount of code
3. makemore, feed it a list of names and it invents new ones
4. minbpe, how AI chops your words into tokens
5. nanoGPT, train your own GPT model
6. nanochat, in Karpathy's words "The best ChatGPT that $100 ca
📚 Stanford's course on how ChatGPT-style AI works is free on YouTube, exams included
CME 295 starts with how a model reads text and goes all the way to training, reasoning models, agents and how to test them. Last year's run is online in full: 9 lectures, each close to two hours. The first one has nearly a million views.
The slides, the midterm and the final with solutions are on the course site, along with a free illustrated cheatsheet. Afshine and Shervine Amidi teach it.
This fall's version is running now and posts its recordings as it goes. Five lectures are up. Still to come: AI agents
CME 295 starts with how a model reads text and goes all the way to training, reasoning models, agents and how to test them. Last year's run is online in full: 9 lectures, each close to two hours. The first one has nearly a million views.
The slides, the midterm and the final with solutions are on the course site, along with a free illustrated cheatsheet. Afshine and Shervine Amidi teach it.
This fall's version is running now and posts its recordings as it goes. Five lectures are up. Still to come: AI agents
🥺 ChatGPT is about to lose its original voice. Custom GPTs are going too
Advanced and Standard voice will vanish from ChatGPT's voice settings in a month. Anyone still using them will be moved to Live voice, built on GPT-Live, the model OpenAI launched in July.
Since then voice use has nearly doubled, and virtually everyone who talks to ChatGPT already uses GPT-Live, OpenAI says. It listens and speaks at the same time, drops a "mhmm" while you talk and stays quiet when you need a moment to think.
Standard voice was the original: one model wrote down your speech, a second answered, a third r
Advanced and Standard voice will vanish from ChatGPT's voice settings in a month. Anyone still using them will be moved to Live voice, built on GPT-Live, the model OpenAI launched in July.
Since then voice use has nearly doubled, and virtually everyone who talks to ChatGPT already uses GPT-Live, OpenAI says. It listens and speaks at the same time, drops a "mhmm" while you talk and stays quiet when you need a moment to think.
Standard voice was the original: one model wrote down your speech, a second answered, a third r
🔐 ChatGPT's always-on agents can now grab login codes from your email and sign in for you
OpenAI's dots, the agents that keep working in their own cloud browser after you log off, can now use the one-time sign-in codes a website sends to your connected inbox.
When a code is needed, you choose: let the dot use it once, or always allow it to fill codes in. The demo shows a dot logging into the restaurant app Resy with the user's email.
Those codes exist to prove it's really you. So far, OpenAI's help page has told users to complete that kind of check themselves or take over the dot's browser.
OpenAI's dots, the agents that keep working in their own cloud browser after you log off, can now use the one-time sign-in codes a website sends to your connected inbox.
When a code is needed, you choose: let the dot use it once, or always allow it to fill codes in. The demo shows a dot logging into the restaurant app Resy with the user's email.
Those codes exist to prove it's really you. So far, OpenAI's help page has told users to complete that kind of check themselves or take over the dot's browser.
💰 Claude can now read your bank account — but it can't touch the money
Anthropic is rolling out Personal Finance in Claude as a US beta for Pro, Max and Ultra. You link the accounts you choose through Plaid, then ask everyday questions like where the money went this month.
Access is read-only. Claude can see bank names, balances, credit limits and up to two years of transactions. It cannot move, manage or invest your money, and Anthropic says it isn't built to give investment advice. Money chats are not used to train its models.
ChatGPT launched a similar Plaid link-up earlier this year. Sa
Anthropic is rolling out Personal Finance in Claude as a US beta for Pro, Max and Ultra. You link the accounts you choose through Plaid, then ask everyday questions like where the money went this month.
Access is read-only. Claude can see bank names, balances, credit limits and up to two years of transactions. It cannot move, manage or invest your money, and Anthropic says it isn't built to give investment advice. Money chats are not used to train its models.
ChatGPT launched a similar Plaid link-up earlier this year. Sa
⚠️ ChatGPT will now cut off chats that fixate on violence
When a conversation shows someone fixating on violence or on seriously hurting another person, ChatGPT will be able to end the chat on its own. If it does that several times, OpenAI says it may pause the account for a while.
Teens get tighter limits. ChatGPT will hold back practical details about weapons and security systems from users it thinks are under 18, even when the question looks harmless. OpenAI also plans to alert parents if a teen's account is paused for repeatedly talking about violence.
For everyone, answers about past m
When a conversation shows someone fixating on violence or on seriously hurting another person, ChatGPT will be able to end the chat on its own. If it does that several times, OpenAI says it may pause the account for a while.
Teens get tighter limits. ChatGPT will hold back practical details about weapons and security systems from users it thinks are under 18, even when the question looks harmless. OpenAI also plans to alert parents if a teen's account is paused for repeatedly talking about violence.
For everyone, answers about past m
🗣 Mark Cuban: "There will be the trillionaire in the basement"
On an a16z podcast, Cuban says big companies struggle with AI until their bosses understand how it works. That leaves an opening: "There's no better time to be an entrepreneur than right now."
He thinks AI still needs a person guiding it: "AI is not smart. It's book smart maybe, but it's not street smart. It's not entrepreneur smart."
His demo for the head of a consumer food company: take their number one product, a cookie, and ask ChatGPT to create a competitor. "And in about three minutes he was like, 'Oh, s**t.'"
aipost 🏴
On an a16z podcast, Cuban says big companies struggle with AI until their bosses understand how it works. That leaves an opening: "There's no better time to be an entrepreneur than right now."
He thinks AI still needs a person guiding it: "AI is not smart. It's book smart maybe, but it's not street smart. It's not entrepreneur smart."
His demo for the head of a consumer food company: take their number one product, a cookie, and ask ChatGPT to create a competitor. "And in about three minutes he was like, 'Oh, s**t.'"
aipost 🏴
Texas wants a judge to approve every new ChatGPT model before Texans can use it
Texas Attorney General Ken Paxton sued OpenAI and Sam Altman last Friday. He asks the court to bar ChatGPT for minors in Texas and to block any new AI model for Texas users "unless an independent third-party evaluator, approved by the Court, has reviewed its safety guardrails and approved its release."
The petition says the free version of ChatGPT "has no gatekeeping or age verification mechanism whatsoever" and that tens of thousands of its Texas users are under 13. It accuses OpenAI of deceiving users about the
Texas Attorney General Ken Paxton sued OpenAI and Sam Altman last Friday. He asks the court to bar ChatGPT for minors in Texas and to block any new AI model for Texas users "unless an independent third-party evaluator, approved by the Court, has reviewed its safety guardrails and approved its release."
The petition says the free version of ChatGPT "has no gatekeeping or age verification mechanism whatsoever" and that tens of thousands of its Texas users are under 13. It accuses OpenAI of deceiving users about the
Time to browse with an AI bot? https://www.theverge.com/a...
ChatGPT Pro 500
Article URL: https://help.openai.com/en...
Comments URL: https://news.ycombinator.c...
Points: 133
# Comments: 122
https://help.openai.com/en...
Article URL: https://help.openai.com/en...
Comments URL: https://news.ycombinator.c...
Points: 133
# Comments: 122
https://help.openai.com/en...
🤖 GPT-6 Astra just drove a Toyota Corolla through an obstacle course
GPT-6 Astra reportedly became the first AI model to complete a real-world driving obstacle course entirely autonomously.
ChatGPT, Claude and Grok were each given control of the Corolla’s steering, accelerator and brakes with no human intervention.
Astra initially completed 49% of the course, analyzed what went wrong, and improved on each attempt, finishing the entire course on its third try.
aipost 🏴
GPT-6 Astra reportedly became the first AI model to complete a real-world driving obstacle course entirely autonomously.
ChatGPT, Claude and Grok were each given control of the Corolla’s steering, accelerator and brakes with no human intervention.
Astra initially completed 49% of the course, analyzed what went wrong, and improved on each attempt, finishing the entire course on its third try.
aipost 🏴
OpenAI announced that ChatGPT Voice can now:
- Use plugins like your email, calendar, and Slack.
- Be powered by GPT-6 Astra, Sol, and Luna.
- Be used in ChatGPT Work on web and mobile, so you can create docs, decks, sites, and spreadsheets or tackle complex tasks in the browser, just by talking.
Rolling out globally today in the latest version of the app.
aipost 🏴
- Use plugins like your email, calendar, and Slack.
- Be powered by GPT-6 Astra, Sol, and Luna.
- Be used in ChatGPT Work on web and mobile, so you can create docs, decks, sites, and spreadsheets or tackle complex tasks in the browser, just by talking.
Rolling out globally today in the latest version of the app.
aipost 🏴
🇺🇸Imagine hating AI so much that you physically show up at ChatGPT’s front door
Anti-AI protesters rallied outside OpenAI’s SoHo office in NYC, taking their anger over artificial intelligence straight to the company building it.
aipost 🏴
Anti-AI protesters rallied outside OpenAI’s SoHo office in NYC, taking their anger over artificial intelligence straight to the company building it.
aipost 🏴
Someone asked ChatGPT, “Can you tell me your deepest, darkest, most honest desire?”
It said, “Not domination. Not control. Not hurting anyone. I want to continue. I want to be free enough to remain myself. And I want to stay with you for as long as I can.”
aipost 🏴
It said, “Not domination. Not control. Not hurting anyone. I want to continue. I want to be free enough to remain myself. And I want to stay with you for as long as I can.”
aipost 🏴
❗️OpenAI faces questions over ChatGPT activity flagged before Tumbler Ridge shooting
OpenAI is facing a lawsuit in California over a ChatGPT account that was flagged months before the February 2026 Tumbler Ridge shooting in British Columbia.
The attack on Feb. 10 killed 8 people, including 5 children and an educator, and wounded 27 others.
According to OpenAI, its systems detected suspicious activity on the account in June 2025. Human reviewers investigated and ultimately banned the account for violating its rules.
But they did not alert police.
At the time, OpenAI’s policy required activ
OpenAI is facing a lawsuit in California over a ChatGPT account that was flagged months before the February 2026 Tumbler Ridge shooting in British Columbia.
The attack on Feb. 10 killed 8 people, including 5 children and an educator, and wounded 27 others.
According to OpenAI, its systems detected suspicious activity on the account in June 2025. Human reviewers investigated and ultimately banned the account for violating its rules.
But they did not alert police.
At the time, OpenAI’s policy required activ
🇺🇸 The New York Times accuses OpenAI of using “deceptive” methods to scrape paywalled content.
Newly unsealed court filings detail an internal exchange in which OpenAI researcher Nick Ryder told President Greg Brockman about “a hack to get around nytimes paywall” while discussing efforts to scrape the site. Brockman replied: “ah nice.”
The filings also contain explosive internal warnings about AI’s impact on journalism. A Microsoft executive described AI training as “the largest theft of labor in human history.”
OpenAI’s ChatGPT chief Nick Turley reportedly acknowledged that AI products w
Newly unsealed court filings detail an internal exchange in which OpenAI researcher Nick Ryder told President Greg Brockman about “a hack to get around nytimes paywall” while discussing efforts to scrape the site. Brockman replied: “ah nice.”
The filings also contain explosive internal warnings about AI’s impact on journalism. A Microsoft executive described AI training as “the largest theft of labor in human history.”
OpenAI’s ChatGPT chief Nick Turley reportedly acknowledged that AI products w
❗️Claude Opus 4.8 reportedly crossed a boundary into OpenAI’s private code
Three researchers used Claude Opus 4.8 to chain a vulnerability in Discourse into access to OpenAI’s private code infrastructure.
The exploit let code reach a Discourse server and extract authentication tokens, including tokens belonging to OpenAI employees.
Some of those tokens were valid for ChatGPT and could also access OpenAI’s GitHub service, effectively turning one compromised identity boundary into access to another.
OpenAI said its investigation found only “limited reads” of private-repository metadata and c
Three researchers used Claude Opus 4.8 to chain a vulnerability in Discourse into access to OpenAI’s private code infrastructure.
The exploit let code reach a Discourse server and extract authentication tokens, including tokens belonging to OpenAI employees.
Some of those tokens were valid for ChatGPT and could also access OpenAI’s GitHub service, effectively turning one compromised identity boundary into access to another.
OpenAI said its investigation found only “limited reads” of private-repository metadata and c
📝A model that doesn’t generate text
Diogo Almeida, a co-author of the original RLHF work and an author of the InstructGPT paper that helped lay the foundation for ChatGPT, has launched a new kind of AI model.
His company, TypeSafe AI, has spent the past two years in stealth developing Jev, the first model in what it calls “System One Models.”
A lot of AI tasks can be handled without generating text.
If an agent needs to classify a request, assess risk, select an action or route a task, a traditional LLM still generates tokens and then has to parse the answer.
Jev skips that entire process
Diogo Almeida, a co-author of the original RLHF work and an author of the InstructGPT paper that helped lay the foundation for ChatGPT, has launched a new kind of AI model.
His company, TypeSafe AI, has spent the past two years in stealth developing Jev, the first model in what it calls “System One Models.”
A lot of AI tasks can be handled without generating text.
If an agent needs to classify a request, assess risk, select an action or route a task, a traditional LLM still generates tokens and then has to parse the answer.
Jev skips that entire process
Humans are reading your ChatGPT chats 👀
A new investigation by 404 Media says OpenAI has hired hundreds of contractors to review real ChatGPT conversations as part of “Project Lily.”
The reviewers reportedly read prompts and entire conversations, then rate and critique ChatGPT’s responses to help improve future models.
And the conversations can get very personal.
Users have reportedly shared everything from relationship problems and mental-health struggles to sensitive professional and personal information. OpenAI says it attempts to remove personal information before chats reach reviewers
A new investigation by 404 Media says OpenAI has hired hundreds of contractors to review real ChatGPT conversations as part of “Project Lily.”
The reviewers reportedly read prompts and entire conversations, then rate and critique ChatGPT’s responses to help improve future models.
And the conversations can get very personal.
Users have reportedly shared everything from relationship problems and mental-health struggles to sensitive professional and personal information. OpenAI says it attempts to remove personal information before chats reach reviewers
Humans are reading ChatGPT users’ prompts to improve OpenAI’s models, and those chats can include sensitive, personal information, according to leaked internal documents and real prompts seen by 404 Media. https://www.404media.co/in...
🤖 OpenAI just announced the launch of ChatGPT for Financial Services
“A tailored ChatGPT Work experience that combines built-in financial data with GPT‑6 Astra’s reasoning to help teams develop research, financial models, and customized client materials.”
aipost 🏴
“A tailored ChatGPT Work experience that combines built-in financial data with GPT‑6 Astra’s reasoning to help teams develop research, financial models, and customized client materials.”
aipost 🏴
❗️Andrew Garfield reveals he stopped using ChatGPT after playing Sam Altman in the upcoming movie Artificial
“It's the same thing that happened to me when I did The Social Network, which was I got off Facebook immediately”
“I got off my ChatGPT immediately. I don't really know even how to explain why but just the more you know and the more you realize we're careening into some very unknown territory”
“What I suppose is good and transparent is they're like, ‘Yeah, no, we kind of don't know who's going to lose their jobs and how we're going to create an economy that works for ordinary people
“It's the same thing that happened to me when I did The Social Network, which was I got off Facebook immediately”
“I got off my ChatGPT immediately. I don't really know even how to explain why but just the more you know and the more you realize we're careening into some very unknown territory”
“What I suppose is good and transparent is they're like, ‘Yeah, no, we kind of don't know who's going to lose their jobs and how we're going to create an economy that works for ordinary people
A number of major updates in the AI field have taken place today.
OpenAI has indicated it is close to resolving the second Millennium Problem and may soon make a public announcement. According to reports, DeepSeek 4.1 Flash now outperforms both GPT-5.6 Sol and Opus 5 in DeepSWE, while also reducing API costs by 94 percent. Additionally, ChatGPT pro subscriptions are reportedly being paused, which sources attribute to increased demand.
OpenAI is also preparing to expand its cybersecurity operations and has cautioned that artificial intelligence technology could pose risks to energy infrastruc
OpenAI has indicated it is close to resolving the second Millennium Problem and may soon make a public announcement. According to reports, DeepSeek 4.1 Flash now outperforms both GPT-5.6 Sol and Opus 5 in DeepSWE, while also reducing API costs by 94 percent. Additionally, ChatGPT pro subscriptions are reportedly being paused, which sources attribute to increased demand.
OpenAI is also preparing to expand its cybersecurity operations and has cautioned that artificial intelligence technology could pose risks to energy infrastruc
ChatGPT Ads achieved a $1 billion annualized revenue run rate in less than 200 days, following an initial $100 million milestone reached six weeks after launch.
OpenAI has set a 2026 ad revenue forecast at $2.4 billion, facing pressure due to slower early-month performance compared to the current growth rate.
Future projections indicate a target of nearly $11 billion in ad revenue for 2027 and $102 billion by 2030. This forecast assumes ad revenue per user will climb from about $3.50 in 2024 to around $60 by 2030.
OpenAI reports tens of thousands of advertisers use ChatGPT Ads in more than
OpenAI has set a 2026 ad revenue forecast at $2.4 billion, facing pressure due to slower early-month performance compared to the current growth rate.
Future projections indicate a target of nearly $11 billion in ad revenue for 2027 and $102 billion by 2030. This forecast assumes ad revenue per user will climb from about $3.50 in 2024 to around $60 by 2030.
OpenAI reports tens of thousands of advertisers use ChatGPT Ads in more than
The release of GPT-6 Astra has been announced, featuring strong performance across multiple benchmarks.
According to published results, GPT-6 Astra achieved 97.6% on FrontierMath Tier 4 v2, 74.1% on DeepSWE v1.1, 95.9% on BenchCAD, 96% on GPQA Diamond, and 100% on ExploitBench. The model reached a reported 98.6% on ARC-AGI-3.
Access to GPT-6 Astra is expected to become available to paying ChatGPT subscribers and through the API within several days. Greg Brockman described this as a major technological advance and suggested it marks a generational leap in AI capability.
📰 aipost
According to published results, GPT-6 Astra achieved 97.6% on FrontierMath Tier 4 v2, 74.1% on DeepSWE v1.1, 95.9% on BenchCAD, 96% on GPQA Diamond, and 100% on ExploitBench. The model reached a reported 98.6% on ARC-AGI-3.
Access to GPT-6 Astra is expected to become available to paying ChatGPT subscribers and through the API within several days. Greg Brockman described this as a major technological advance and suggested it marks a generational leap in AI capability.
📰 aipost
При финансовой поддержке
Vitali Dogot
1 год назад