The AI Hub Weekly: ChatGPT Gains a Smarter, Scalable Memory

It’s week two of June, and the AI world just did what it does best—be everywhere at once. From OpenAI quietly rolling out smarter memory in ChatGPT, to huge insurance companies letting AI fully handle crash claims, it’s clear the “agentic” era isn’t a buzzword anymore. But as everyday tools get more personal and powerful, the rules of who controls, benefits from, and profits off AI are shifting—sometimes in surprising ways.

TOP STORIES

ChatGPT Gets a Major Memory Upgrade to Personalize Conversations

• OpenAI is rolling out a more scalable “memory” system for ChatGPT, designed to better remember user preferences and conversations across hundreds of millions of users.
• This feature means ChatGPT will be more helpful and less forgetful—tracking your projects, correcting stale info, and scaling to long-term use.
• The new memory system will become available to more users in stages, with enterprise access prioritized.
• Source

S&P 500 Rejects SpaceX, Also Bars OpenAI and Anthropic From Index Entry

• The S&P 500 stock market index denied SpaceX’s fast-track request for inclusion; the decision also blocks other unprofitable AI companies like OpenAI and Anthropic from joining.
• Analysts say S&P’s refusal signals a hard line against “growth-at-all-costs” firms—even if they dominate headlines and investment.
• For regular investors, it means you won’t see OpenAI or Anthropic shares in your basic index funds anytime soon.
• Source

OpenAI’s New GPT-Rosalind Update Aims at Real-World Science and Drug Discovery

• OpenAI released an upgrade for GPT-Rosalind, its AI model tailored for life sciences, now powered by GPT-5.5 and able to run sophisticated “agent” workflows common in drug research.
• The update brings better scientific reasoning and stronger tool integration for industry teams working on discovery pipelines.
• Available now to enterprise customers; expect to see more biotech and pharma using AI for real, results-driven work.
• Source

AI-Powered Claims Handling Now Covers Most U.S. Car Accidents at Travelers Insurance

• Travelers, a major U.S. insurer, now resolves over 90% of accident claims via an “AI Claim Assistant” built with OpenAI tech.
• Customers who call after a crash can have their claim filed and completed (often in real-time) by AI, even in disaster-level events.
• This rollout is nationwide and signals how routine, high-volume tasks in big industries are quietly going fully AI-driven.
• Source

Endava Moves Its Entire Software Delivery Pipeline to AI Agents

• Endava, a global tech consultancy, announced it’s redesigning how it delivers software—putting AI agents at the center of workflows from project kickoff to code deployment.
• For clients, this means faster releases, fewer handoffs, and more dynamic updates—but also big shifts in job roles and required skills.
• Early adopters can expect a blend of automation and human oversight; broader rollouts are coming in months.
• Source

Wasmer Uses OpenAI Codex to Make Node.js Apps Run Securely “at the Edge”

• Wasmer engineers used Codex and GPT-5.5 to develop a new way to run Node.js (JavaScript) apps, agents, and microservices inside a secure sandboxed WebAssembly environment.
• This approach claims to speed up developer work by 10–20x, opening up secure, scalable web applications that can deploy anywhere—from cloud to your laptop.
• It’s a sign of AI lowering barriers for complex tech projects, making next-gen apps more accessible to smaller teams.
• Source

THIS WEEK IN AI

What’s more personal than memory? This week, OpenAI decided it was time to tackle one of AI’s most persistent annoyances: the fact that your AI assistant didn’t remember you. Rolling out a scalable, non-forgetful “memory” for ChatGPT is a deceptively simple update. But for millions of us who use ChatGPT as a work tool, sidekick, or substitute for tedious admin, this changes the game. The AI that knows your context, remembers your last request, and gets smarter the more it helps—you might wonder: is this the beginning of real AI companionship, or just the next level of digital lock-in?

Alongside this, we’re seeing a massive expansion of AI’s reach into “serious” industries. There’s something quietly radical about 90% of Travelers’ accident claims now routed through an AI agent. Not only does it mean faster resolutions (and likely cost cuts for insurers), it’s a signal to all of us: if you thought AI was just parlor tricks or novelty text generators, you’re missing the real story. The boring, repetitive work that makes giant businesses tick is being handed off to technology—efficient, tireless, and, for now, not looking for a promotion.

But it wouldn’t be a week in AI without reminders that there are boundaries still being enforced. The S&P 500’s curt “no” to SpaceX, OpenAI, and Anthropic’s index inclusion is a statement: profits still matter, and financial gatekeepers won’t cave to hype or vision alone. It raises hard questions—should the public be able to own a piece of these transformative (but risky) AI labs? Are Wall Street rules holding back progress, or preventing disaster?

The “agentic” trend, meanwhile, is seeping into every workflow imaginable—from Endava’s code factories to the biotech labs harnessing GPT-Rosalind for drug research. Every organization trying to move faster is essentially building itself around machines that act on their own, updating their memory and context with every task. But how do we deal with the friction this introduces for human workers? If AI can handle more context than a person ever could, where does that leave skills and careers that once seemed future-proof?

Here’s the challenge for all of us: Next time you use ChatGPT, actually switch on and try the new memory features. What part of your work or routine would genuinely be better if the AI just “knew” you and didn’t start from scratch every time? And when you see another headline about AI handling thousands of tasks per minute, ask yourself—are you watching from the sidelines, or are you learning how to direct these new agents yourself?

MORE TOP STORIES

OpenAI Releases Its First Public Policy Agenda, Vows to Democratize AI

• OpenAI published a detailed policy playbook describing how AI should benefit everyone, not just tech giants or a select few.
• Its five core principles emphasize resisting power concentration and ensuring user empowerment, aiming for inclusive global impact.
• The document signals how major labs hope to shape regulation—and public expectations—as AI gets more capable.
• Source

New “Democratic Governance” Blueprint for Frontier AI Released

• OpenAI detailed proposals for a U.S. federal approach to governing dangerous or advanced AI, outlining a three-part plan to align oversight.
• The blueprint addresses legal, technical, and institutional tools for handling the next generation of AI systems.
• It’s an attempt by industry to influence how potentially risky models are managed before government moves too slowly.
• Source

AI Overtakes Coding: 80% of New Wasmer Node.js Runtime Built With GPT-5.5 Codex

• The Wasmer project leveraged OpenAI tools to rapidly develop a complex, secure runtime for JavaScript apps on the web’s “edge.”
• They claim development speed increased by up to 20x, showing how agentic AI now shapes complex software, not just text output.
• Such advances make it easier for small teams to build and deploy next-gen applications previously only possible for giants.
• Source

Codex Now Used by Over 5 Million People Beyond Programmers

• OpenAI released new data showing its Codex tool, initially built for coding, has spread as a general productivity booster.
• More than 5 million people use Codex weekly to automate routine tasks across many fields, not just software development.
• The expansion points to AI assistants moving from specialist tools into daily business and knowledge work.
• Source

Google I/O Used Gemini Agents to Build Its Own Conference

• Google revealed that the flagship I/O 2026 developer event was itself largely “built by AI,” with Gemini models and agents doing everything from presentations to logistics.
• The company’s meta-move signals that AI agent workflows aren’t just demo material—they’re production-ready, even for high-profile events.
• Expect accelerated adoption across events, marketing, and operations in other major organizations.
• Source

Holo3.1 Launches Fast, Local Computer-Use Agents for Cloud and Device

• The Holo3.1 agent family lets users run “digital worker” tasks across desktop, mobile, or cloud—claiming seamless switching between environments.
• It boosts speed and privacy by supporting fully local execution on personal devices (not just cloud processing).
• Available now via Hugging Face, this tech is a glimpse at where “your AI assistant” could actually live.
• Source

ALSO THIS WEEK

• JetBrains launches Mellum2, a 12B-parameter Mixture-of-Experts AI model trained on code and language — Now available on Hugging Face, aiming to accelerate developer productivity. (Source)
• Momfluencers are pitching AI as a better ‘coparent’ than men — Social media influencers in parenting circles increasingly promote AI copilots for emotional and logistical parenting support. (Source)
• “When Claude changed, everything changed”: Managing AI blast radius in production — A look at how sudden model updates can break real-world business workflows overnight. (Source)
• Agentic AI solved coding—and exposed every other problem in software engineering — As AI accelerates code, bottlenecks shift to integration, testing, and real-world deployment. (Source)
• Microsoft AI chief says company was “set free” from OpenAI to pursue superintelligence — Leadership claims greater independence as Microsoft ramps its own AGI ambitions. (Source)
• Microsoft’s AI Futurist explains real-world Copilot use — Enterprise “agents” are being deployed in operations, supply chains, and productivity tools. (Source)
• AI agents are learning on the job—but corrections aren’t shared across teams — Highlights the fragmented way AI “learns” in business, and why context doesn’t transfer. (Source)
• Meta’s AI support agent allowed unauthorized account email recovery — Security failures let attackers hijack accounts with little oversight. (Source)
• Anthropic says 80% of its new production code is authored by Claude — The company leads the charge as AI-written code becomes the default for enterprise. (Source)
• OpenAI and Anthropic may be bitter rivals, but investors aren’t picking sides — VCs are hedging, betting on both major labs despite strategic differences. (Source)
• Trump’s AI executive order on safety testing faces pushback after government cybersecurity teams were gutted — Critics question whether voluntary model checks will prevent dangerous deployments. (Source)
• Microsoft’s Project Solara: Android for AI agents, not apps — Microsoft previews a new mobile OS where “agentic” workflows replace classic apps. (Source)
• The Meta hack highlights simple but dangerous AI security gaps — Some AI-driven attacks are easily overlooked, but still highly effective. (Source)
• Rehumanizing global healthcare with agentic AI — As AI takes over admin burdens, practitioners hope to spend more time with patients. (Source)
• How small businesses can leverage AI — A tutor shows how LLMs are quietly reshaping small-team workflows. (Source)
• Microsoft’s AI chief says superintelligence is near, but won’t take your job — Mustafa Suleyman, CEO of Microsoft AI, makes bold claims about AGI’s timeline. (Source)
• ZDNet: ChatGPT’s new memory upgrade is a double-edged sword — The new “profile” system makes ChatGPT more useful, but could amplify mistakes. (Source)
• AI masks America’s “post-literate” workforce — Automation may be hiding low literacy rates among U.S. workers. (Source)
• Trump’s push for U.S. stake in AI giants stirs debate — The former president argues a public share of AI’s upside could be “beautiful.” (Source)
• Children’s wellbeing slips nationwide — New research shows children’s health and success indicators worsening since 2019. (Source)
• Jewish Democrats warn about rising antisemitism ahead of 2028 — Concerns swell in the political realm as divisions grow. (Source)
• Iran faces World Cup visa gauntlet — U.S.-Iran tensions challenge FIFA logistics and travel arrangements. (Source)
• The business of AI is facing harsh new realities — Investors recalibrate expectations as markets question AI’s path to profit. (Source)
• AI bubble “revenge” — After years of hype, the AI economy faces its own cycles of boom and skepticism. (Source)
• Trump: U.S. stake in AI could be “beautiful” — Proposes government investment in the biggest AI labs. (Source)
• ZDNet: Budget laptop comparison — How AI is changing what you get for $899 vs. $1299 in laptops. (Source)
• ZDNet: ChatGPT vs. Gemini image generation — Subtle prompt tweaks yield wildly different results in AI art. (Source)
• C3 AI agents to automate predictive maintenance for Shell — Shell shifts from anomaly detection toward full AI-driven repairs and predictions. (Source)
• Meta Business Agent launches for conversational commerce — Retailers can now field orders and customer questions from Meta’s AI inside messaging apps. (Source)
• Microsoft’s Scout Autopilot expands agentic workflows in M365 — Now in wider testing, bringing cross-tool automation to enterprises. (Source)
• Amazon’s AI shopping assistant is now available to retailers, with Kate Spade as a launch customer — AI powers customer journeys and sales recommendations in-store. (Source)
• Walmart limits AI use as workflows meet balance sheet reality — Internal LLM-powered tool “Code Puppy” gets scaled back due to high demand costs. (Source)
• Microsoft’s Majorana 2 quantum chip showcases agentic AI in R&D — Boasts 1,000x more reliable qubits and shows AI shaping R&D at the hardware frontier. (Source)
• Anthropic IPO filing signals AI’s shift to enterprise utility — The move underlines commercialization and industry focus beyond pure research. (Source)
• GitHub Copilot users see use-based price hikes — As Copilot switches to token-based billing, some users will pay more for frequent AI use. (Source)
• China linked to fueling U.S. resistance to data centers — AI industry claims foreign actors are stoking local pushback against new buildouts. (Source)
• Who’s behind Trump’s AI strategy? — Energy Department undersecretary Darío Gil emerges as a key adviser. (Source)
• House Democrats join GOP to kill pushback on Trump’s Lebanon policy — A rare bipartisan move blocks Lebanon war powers legislation. (Source)