ChatGPT now remembers every thing

Good morning, AI enthusiasts. OpenAI just gave ChatGPT an enormous memory boost, enabling it to routinely remember and reference key information across all of your conversations without being asked.

Is that this persistent memory the breakthrough that finally turns ChatGPT into the truly personalized AI assistant we have been waiting for?

In today’s AI rundown:

• OpenAI gives ChatGPT a memory boost

• Mira Murati’s AI startup eyes record funding

• Turn YouTube videos into high-ranking blogs

• Study: AI still struggles with software debugging

• 4 recent AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🧠 OpenAI gives ChatGPT a memory boost

The Rundown: OpenAI just rolled out a massive update to ChatGPT’s memory, enabling the AI assistant to routinely remember and reference information across all user conversations, delivering more personalized and relevant responses.

The small print:

• ChatGPT will cut across all conversations, listening in on a regular basis and capturing users’ preferences, interests, needs, and even things they don’t like.

• With all this information, the assistant will then tailor its responses to every user, engaging in conversations “that feel noticeably more relevant and useful.”

• Unlike previous versions where users needed to specifically request that information be remembered, the system now does this routinely.

• If you ought to change what ChatGPT knows about you, simply ask within the chat through a prompt.

Why it matters: This feature is a game changer for lively users who hate switching between chats or continuously repeating themselves. ChatGPT’s prolonged memory is a begin to an exciting future where AI systems genuinely get to know you over time, becoming increasingly personalized and useful.

Privacy note: Users can opt out of the memory feature via ChatGPT’s settings or use temporary chat mode for conversations they don’t need remembered.

TOGETHER WITH INNOVATING WITH AI

💼 Start your profession as an AI Consultant

The Rundown: Innovating with AI’s recent program, AI Consultancy Project, equips AI enthusiasts with all of the resources to capitalize on the rapidly growing AI consulting market – which is ready to 8x to $54.7B by 2032.

This system offers:

• Tools and framework to seek out clients and deliver top-notch services

• A 6-month roadmap to construct a 6-figure AI consulting business

• Student landing their first AI client in as little as 3 days

Click here to request early access to The AI Consultancy Project.

THINKING MACHINES LAB

💰 Mira Murati’s AI startup eyes record funding

The Rundown: Former OpenAI CTO Mira Murati’s Pondering Machines Lab, an AI startup led by several OpenAI veterans, is reportedly pursuing one among the most important seed funding rounds in history.

The small print:

• Fresh out of stealth with nearly half of the founding team from OpenAI, Pondering Machines Lab is in talks to boost $2B at a valuation of “at the least” $10B.

• The worth of the round is double what Murati was initially targeting, though details can change because the round remains to be said to be in progress.

• Murati launched the AI startup six months after leaving OpenAI, where she spent nearly seven years working on AI systems, including ChatGPT.

• While much stays under the wraps, the direction of Pondering Machines is towards “widely understood, customizable, and usually capable” AI systems.

Why it matters: With Pondering Machines Lab’s record-setting raise and Ilya Sutskever’s SSI also reportedly raising at a $30B valuation, the AI funding frenzy continues to succeed in recent heights. What’s striking, though, is that each these startups are pulling massive investments with out a public product or revenue roadmap.

AI TRAINING

📝 Turn YouTube videos into high-ranking blogs

The Rundown: On this tutorial, you’ll learn the right way to use NotebookLM to show YouTube video content into Search engine optimisation-optimized blog posts by analyzing transcripts and generating structured, search engine-targeted content.

Step-by-step:

1. Create a notebook in NotebookLM and add your YouTube video transcript as a source via YouTube link or pasted text.

2. Prepare your Search engine optimisation strategy by identifying primary and secondary keywords (e.g., for AI automation: “AI workflow tools,” “business process automation”).

3. Craft an in depth prompt including your keywords and desired structure, then generate your content.

4. Enhance your post with images, links, formatting, and a compelling call-to-action before publishing.

Pro tip: For technical topics, include a glossary section defining complex terms to enhance search visibility for beginners looking for explanations of those concepts.

PRESENTED BY IMAGINE AI LIVE

⏰ AI’s future won’t wait — will you retain up?

The Rundown: IMAGINE AI LIVE ’25 isn’t just one other event— it’s your ticket to leapfrog competitors who’re still scrambling to know AI’s potential. Join industry pioneers on the Fontainebleau Las Vegas and get practical strategies to implement AI that drives immediate results.

Attend on May 28-30 on the Fontainebleau Las Vegas and you may:

• Speed up your AI journey by years, not months

• Connect with cross-industry innovators who’ve already solved your challenges

• Transform theoretical AI knowledge into actionable implementation roadmaps

Secure Your Competitive Edge with code AIDOESNTWAIT15 and save 15% for those who register by today!

MICROSOFT RESEARCH

🤖 Study: AI still struggles with software debugging

The Rundown: Microsoft Research published a study revealing that AI agents—even those powered by essentially the most advanced models available—still struggle with most software debugging tasks that human programmers routinely solve.

The small print:

• Microsoft used nine LLMs, including Claude 3.7 Sonnet, to power a “single prompt-based agent” tasked with 300 debugging issues from SWE-bench Lite.

• Within the test, the agent struggled to finish half of the assigned tasks, even when using frontier models that excel at coding as its backbone.

• With debugging tools, 3.7 Sonnet performed best, solving 48.4% of issues, followed by OpenAI’s o1 and o3-mini with a 30.2% and 22.1% success rate.

• The team found that the performance gap is as a result of a scarcity of sequential decision-making data (human debugging traces) within the LLMs’ training corpus.

Why it matters: As investors and executives from firms like Google and Meta proceed pouring billions into AI coding agents, this study serves as a reality check of the present state. Despite impressive progress in code generation, AI still falls significantly short in debugging, one among programming’s most important skills.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 Gemini 2.5 Flash – Google’s recent faster and cheaper reasoning model

• 🤖 Kimi-VL – Moonshot’s 3B-parameter VLM matching 10x larger models

• 🎥 Pika Twists – Manipulate objects or characters in footage while keeping the scene intact

• 💬 askplexbot – Perplexity’s recent Telegram channel, allowing users to interact with or bring the chatbot into group chats

💼 AI Job Opportunities

• ⚖️ OpenAI – Senior IP Paralegal

• 📈 Meta – Product Marketing Manager

• 🛠️ Harvey – User Operations Manager

• 🖥️ Waymo – Senior Software Engineer, Infrastructure

📰 Every thing else in AI today

Ilya Sutskever’s Protected Superintelligence (SSI) partnered with Google Cloud to make use of the corporate’s TPU chips to power its research and development efforts.

Google CEO Sundar Pichai confirmed that the corporate will adopt Anthropic’s open Model Context Protocol to let its models hook up with diverse data sources and apps.

Canva introduced Visual Suite 2.0, several AI features, and a voice-enabled AI creative partner that generates editable content at Canva Create 2025.

OpenAI countersued Elon Musk, citing a pattern of harassment and asking a federal judge to stop him from any “further illegal and unfair motion.”

OpenAI also open-sourced BrowseComp, a benchmark that measures the flexibility of AI agents to locate hard-to-find information on the web.

TikTok parent ByteDance announced Seed-Pondering-v1.5, a 200B reasoning model—with 20B lively parameters—that beats DeepSeek R1.

Elon Musk’s AI startup, xAI, made its flagship Grok-3 model available via API, with pricing starting at $3 and $15 per million input and output tokens.

AI company Author launched AI HQ, an end-to-end platform for constructing, activating, and supervising AI agents within the enterprise.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the tip of the workshop, you’ll have your personal AI assistant that hurries up your workflow and helps you get more done, with less manual effort.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Joey, Zach, Alvaro, Jason, and Shubham —The Rundown’s editorial team