Online shopping is broken. It's tedious and inefficient. It's slow and repetitive. We @AIStudioLab decided to fix it. Introducing Add To Cart AI: the fastest way to shop online. It's the first and one true AI Agent for e-commerce stores. 🧵

typo: on your own*^ docs site.

Hey @sama @gdb @OpenAI I believe it's deeply wrong that you allow ChatGPT agent everywhere else except in ChatGPT itself and your own docs site. Why can't I use your product on your on website?

it takes lots of courage to do and publish this kind of study. it brings forth the evident truth to light that obtaining online visibility is getting harder and more competitive. the effects of the great decoupling continue to exacerbate.

Ranking volatility on Google post June 2025 Core Update 🧵

🆕 How we as SEO Industry is letting this happen? -> why Google is kicking out publishers from even top 3 positions in results. so far from Jan to end of October I've done 78 meetings with different publishers whose traffic graph from Google looks like this! Over 30 people in editorial team on average! And traffic isn't getting stolen by AI Overviews/Mode rather its top 3 is getting wiped out. Mostly replaced with lower quality content from Reddit and yes I agree that Reddit is better sometimes but that doesn't mean they are always better! Google is literally rolling over the publishers and destroying their businesses and also saying this again its not the AI Overviews/Mode - why Google is kicking out publishers from even top 3 positions in results.

The IRON robot, in its full bare metal glory! XPeng's CEO shares an inside look at the humanoid in the lab.

People ask me about what private medical datasets we have access to, believing that to be the limiting aspect of medical AI development. But I'll stay something a bit controversial: sometimes private data is overrated If a model trained on public data beats one trained on private data, what is the point of having private data? People fail to realize that there is so much performance that can be squeezed out of the public datasets. The strategy should be to push performance on public data as far as possible, which then serves as a baseline for further work with private datasets. We'll be sharing an example of this shortly.

Codex for running overnight. <looks inside> > continue > continue > continue > continue > continue > continue > continue > continue .....

Those clicks aren't coming back yall, keep pushing to change your KPIs. We just updated our research from earlier this year, if you thought it couldn't get worse... it did. seerinteractive.com/insights…

Introducing Nested Learning: A new ML paradigm for continual learning that views models as nested optimization problems to enhance long context processing. Our proof-of-concept model, Hope, shows improved performance in language modeling. Learn more: goo.gle/47LJrzI @GoogleAI

News thread for Thursday, Nov 6th. Looks like multiple announcements today. Kimi K2-Thinking is live in Kimi-chat now, not in app for me yet. Benchmarks are leaking ahead of the official announcement.

Google Search to drop more search features and structured data types: - Practice problem - Dataset - Today's Doodle box - Nutrition facts - Nearby offers and events - Local bikeshare station status - TV season selector - and more seroundtable.com/google-drop…

$36,000

Important case. In my opinion, Amazon's claims are wrong.

Today senators Mark Warner and Josh Hawley announced the AI-Related Job Impacts Clarity Act. It would require submitting quarterly reports to the Department of Labor detailing any job losses or reduced hiring due to replacement or automation by AI.

You can now interrupt long-running queries and add new context without restarting or losing progress. This is especially useful for refining deep research or GPT-5 Pro queries as the model will adjust its response with your new requirements. Just hit update in the sidebar and type in any additional details or clarifications.

New on the Anthropic Engineering blog: tips on how to build more efficient agents that handle more tools while using fewer tokens. Code execution with the Model Context Protocol (MCP): anthropic.com/engineering/co…

With the Atlantic hurricane season nearly over, a new champion has emerged in predicting both track and intensity of hurricanes: Google DeepMind. According to preliminary analysis by Brian McNoldy, a senior researcher at the University of Miami, Google's AI models had the lowest mean position error in measuring the track of all 13 named storms this year from 0 to 120 hours. See the "Track Verification" chart below (lower lines are better because they signify lower errors). Google's models also consistently beat all models other than the National Hurricane Center's official forecast (OFCL) in forecasting hurricane intensity. See the "Intensity Verification" graph, which shows Google and OFCL running essentially neck-in-neck for all forecasts between 0 and 120 hours (again, lower line is better). Google's experimental Weather Lab service launched in June of this year, so this was the first hurricane season for which Google's AI models made forecasts. This is a really remarkable result!