- AI Toast
- Posts
- Google dropped Personal Intelligence
In partnership with
Welcome to AI Toast!
Here’s what’s on the menu today:
Google introduces personal intelligence
Exodus at Thinking Machines Lab shakes industry
The Legendary Erdős Problems cracked by GPT-5.2
2Superhero tools for automation
Seven ways to use NotebookLM
Quick Toasts: Wikimedia’s AI Partnerships
Total read time: About 8 minutes, perfect for a quick claude vibe code break.
Google has updated the Gemini app to better understand your personal context and digital habits. This new level of integration aims to make the assistant more useful in daily life.
Personal intelligence connects Gemini directly to your Google apps with a single tap.
The system pulls information from Gmail, Photos, and YouTube to provide tailored answers.
Users can ask for specific details like a license plate number stored in a photo.
It can reason across multiple sources to plan trips or suggest products you might like.
Privacy remains a focus as these connections are off by default and require manual activation.
The assistant does not train its core models on your private emails or library photos.
Correcting the AI is possible if it makes wrong assumptions about your interests or history.
This update is currently rolling out to paid subscribers in the United States.
Managing a digital life often feels like manual labor when searching through old threads and folders. Moving toward an assistant that already knows the context of your last vacation or car repair. This makes AI feel less like a tool and more like an actual assistant.
Personal Take: The future of these assistants lies in their ability to act as a memory layer for our digital lives. As these systems become more proactive, we will likely spend less time digging for information and more time using it to make decisions. The only concerning thing would be the privacy of the user.
Help us make better ads
Did you recently see an ad for beehiiv in a newsletter? We’re running a short brand lift survey to understand what’s actually breaking through (and what’s not).
It takes about 20 seconds, the questions are super easy, and your feedback directly helps us improve how we show up in the newsletters you read and love.
If you’ve got a few moments, we’d really appreciate your insight.
Image Source: Own Design
Mira Murati’s high-profile startup faces significant internal pressure following the exit of several key leaders. These departures test the stability of a company recently valued at twelve billion dollars.
Co-founders Barret Zoph and Luke Metz are rejoining OpenAI after less than a year.
Murati announced that the company parted ways with Zoph following allegations of unethical conduct.
Researcher Sam Schoenholz is also returning to OpenAI as part of a coordinated move.
Reports suggest Zoph may have shared confidential data with competitors during his brief tenure.
Soumith Chintala, known for co-creating PyTorch, has been named the new chief technology officer.
The startup recently raised two billion dollars from major investors like Nvidia and Andreessen Horowitz.
OpenAI executives confirmed the rehiring had been in progress for several weeks before the announcement.
This marks the third major founding member to leave since the company launched last February.
Personal Take: Such a young company suggests that massive funding cannot always resolve disagreements. Replacing a technical co-founder with a veteran like Chintala may stabilize the ship, but the optics remain difficult. Future talent wars will likely see more "boomerang" employees returning to established labs as startups struggle to maintain momentum.
The CRM that saves teams hours every week
HubSpot Smart CRM learns how your team works and adapts to help everyone perform better, which means you'll spot who needs support, celebrate wins instantly, and track what actually matters. The result? You get back hours every week to focus on growth instead of admin work. Start free today.
Image generated with Gemini
Software engineer and researcher Neel Somani watched as OpenAI’s latest model, GPT-5.2, independently generated a complete, verifiable proof for a high-level mathematics problem in about fifteen minutes.
Somani’s original goal was to identify where large language models still struggled with open math problems.
Instead, GPT-5.2 produced a full solution using step-by-step reasoning without human guidance.
The proof was formally verified using Harmonic, confirming its correctness.
The model applied advanced concepts such as Legendre’s formula and Bertrand’s postulate, showing genuine mathematical reasoning rather than pattern matching.
Why Erdős Problems Matter for AI?
Erdős left behind more than a thousand unsolved problems across number theory, combinatorics, and graph theory. These problems have long represented the edge of mathematical creativity and now they’re becoming a testing ground for AI discovery.
What makes this breakthrough credible isn’t just generation but it’s verification.
Formal proof systems like Lean allow computers to confirm proofs with absolute precision.
Tools like Harmonic’s Aristotle automate the translation of reasoning into machine-checkable logic.
This creates a powerful loop: AI generates ideas → verification tools validate them → humans gain trust in the results.
Researchers are increasingly adopting these tools in real academic workflows, signaling that AI is moving from novelty to infrastructure.
AI isn’t replacing mathematicians, it’s becoming a force multiplier.
It can scan massive bodies of literature instantly.
It can generate hypotheses humans might never attempt.
It can explore solution paths at machine scale.
Humans remain the architects of intuition, framing and interpretation but AI is accelerating the frontier itself.
Personal Take: When AI starts producing verified proofs in open mathematics it crosses from automation into discovery. The real story isn’t speed or convenience. It’s that we’re watching the first signs of a new kind of intelligence participating in humanity’s oldest intellectual pursuit.
Superhero Tools for automation
n8n is an extensible workflow tool that lets you build complex AI-native automations using a visual interface while maintaining the option to add custom code.
Zapier connects thousands of different applications through simple triggers and actions to help you scale business processes without any technical knowledge.
Make provides a visual-first platform for designing multi-step workflows that can orchestrate sophisticated AI agents and data integrations across your entire tech stack.
UiPath offers a robust enterprise suite for robotic process automation that uses software robots to handle repetitive manual tasks and legacy system interactions.
Bardeen is an AI-powered browser extension that automates manual tasks like web scraping and lead enrichment directly within your browser.
Seven ways to use NotebookLM
NotebookLM is an AI-powered research assistant designed to organize and synthesize your personal data.
Access the platform by visiting notebooklm.google and signing in with your Google account.
Create a new notebook to house your project, which acts as a distinct container for specific sets of documents and notes.
Upload your primary sources, such as PDFs, Google Docs, website links, or YouTube URLs, to provide the AI with the necessary data to analyze.
Review the source guide to see automatically generated summaries and suggested questions that help you understand the breadth of your material.
Use the chat interface to ask specific questions about your sources or click the audio overview button to generate a deep-dive discussion of your data.
Let’s see some real use cases with an example. For a complex project like a product launch, these seven features transform scattered data into a structured strategy.
Generate an audio overview to turn dense technical specs and market research into a conversational, podcast-style discussion that team members can listen to during their commute.
Create an interactive study guide for your sales team by uploading product manuals to automatically generate quizzes and flashcards on key features and pricing.
Build a visual mind map to see the connections between different customer interview transcripts, uncovering recurring pain points and feature requests.
Summarize complex YouTube videos of competitor demos or industry panels by providing the URL to get a structured breakdown of their claims and audience reactions.
Draft professional briefings or blog posts by using the "Studio" suite to transform your raw internal brainstorming notes into polished public-facing copy.
Organize project documentation by funneling meeting minutes, budget drafts, and legal requirements into a single, queryable knowledge base for the entire launch team.
Conduct deep research with inline citations that link every AI-generated response back to the exact passage in your original source to ensure your marketing claims are accurate.
Do you have different use cases for NotebookLM? Reply and let me know.
Quick AI News Bites
The Wikimedia Foundation has established new AI partnerships with major tech firms to ensure Wikipedia remains a reliable source for LLMs. These agreements focus on maintaining data quality and attribution while providing companies with a structured way to access community-governed knowledge.
Taiwan is looking to position itself as an AI partner with the United States through a new trade and tariff framework. The government hopes this deal will secure its role in the global semiconductor supply chain and foster closer ties with American technology developers.
Google has released TranslateGemma, a collection of open models designed for long-form translation across diverse languages. These models aim to help developers build more accurate translation tools that can handle complex grammar and varying cultural contexts.
OpenAI is expanding its research capabilities by investing in Merge Labs to study how digital systems can better interact with physical environments. This collaboration will likely influence the development of more capable autonomous systems and advanced robotics software.
Anthropic researchers have developed a set of economic index primitives to measure the impact of AI on different sectors of the global workforce. The project provides a framework for understanding how automated systems influence productivity and wealth distribution in the modern economy.
Boost revenue and gain new customers by partnering with us
Reach over 35K AI enthusiasts with your product.
Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!
Got feedback, a story worth toasting, or a wild tech question? I’d love to hear from you; just reply, or find me on X.
Cheers,
— Poonam Soni