In partnership with
Good morning! Here's what's happening in AI today:
OpenAI rebuilds ChatGPT's entire voice architecture with GPT-Live
Gemini Spark now handles real errands through Chrome auto-browse
Cursor can now read, write, and act across Google Workspace
How to run Hermes inside Buzz so every agent action is signed, audited, and traceable
4 new AI tools worth trying today
TOGETHER WITH HUBSPOT
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
AI NEWS
OpenAI published a detailed response to Apple's trade-secrets lawsuit nearly a month after it was filed, disputing several of Apple's specific claims and releasing internal messages and email correspondence to support its account.
OpenAI says Apple's claim that it never responded to an earlier outreach doesn't hold up, since Apple's own outside lawyers sent that message to the wrong person after confusing two similar names, an error Apple only acknowledged once OpenAI pointed it out.
On the accusation that a former Apple employee accessed confidential files after leaving the company, OpenAI says Apple's own employees had reached out asking that former employee for help locating information in the first place.
OpenAI defends the second named employee, a longtime Apple veteran who now leads hardware at OpenAI, saying he has been consistently clear the company should not use confidential information from other companies.
This is one of the more significant disputes between two major AI and hardware players right now, and it signals that OpenAI's hardware ambitions are serious enough that Apple felt the need to sue over them. For anyone watching the AI hardware space, this fight is likely to shape how carefully companies handle hiring and information boundaries with each other going forward. It's still an active, unresolved case, so today's response is one side's account, not a ruling.
AI AGENT
Google connected Gemini Spark to Chrome's auto-browse feature, letting it use a person's logged-in accounts, with permission, to handle real online tasks like scheduling apartment viewings or researching and starting a flight booking.
Spark can now act inside a browser using accounts a person is already logged into, rather than being limited to answering questions or drafting text.
The two examples Google shows, apartment viewings and flight research, are genuinely multi-step errands, not single clicks, the kind of task people usually put off because it takes real back-and-forth.
Permission sits at the center of how this works, Spark only touches accounts and sites a person has explicitly allowed it to use.
Most AI agents can research and summarize, but stop short of actually doing the task that follows. An agent that can navigate a real logged-in account and get partway through a booking is a different, more useful thing, provided the permission model holds up under pressure. This is worth trying on something low-stakes first, before handing it something with money attached.
AI VOICE
Cursor rolled out new plugins giving its agents direct access to Gmail, Google Drive, Calendar, Docs, and Sheets, letting them read, write, and act across a person's entire Workspace instead of just the code editor.
Cursor's agents can now pull context from an inbox or a doc directly, instead of a developer manually copying information over from another tab.
Write access means an agent can update a Sheet or draft something in Docs as part of finishing a coding task, not just read for context.
This extends Cursor's reach well past its original scope as a code editor, into the kind of work-management territory usually reserved for a separate assistant.
A coding agent that can also touch your calendar and inbox starts blurring the line between "dev tool" and "everything tool," which is either exactly what a busy developer wants or exactly the kind of scope creep worth watching closely. For teams already living inside Cursor, this cuts out a lot of manual copy-pasting between tabs. The real question is how tightly permissions are scoped before this ships more broadly.
HOW TO AI
Every bot in Slack authenticates the same way, one shared API key, issued by the vendor, not by you. When something breaks, you're debugging a token, not an identity.
Block shipped Buzz to fix that. It looks like Slack, channels, threads, DMs, but underneath it runs on Nostr, so every agent gets its own cryptographic keypair instead of a borrowed one. Nous Research shipped the official Hermes integration days later.
Three ways in, one worth building first
Buzz Desktop is zero-config, install Hermes, install Buzz, it auto-discovers on your PATH. Good for solo testing, but it auto-approves tool permissions, so keep it owner-only.
The native gateway is the one Nous Research recommends and the one that keeps the full feature set, memory, skills, cron, smart approvals, all exactly as they work anywhere else Hermes runs.
Setting it up:
hermes gateway setupPick Buzz from the platform list. You'll need the buzz CLI on your PATH, a community relay URL, and a Nostr private key, shown once when you create the agent in Buzz, copy it immediately.
BUZZ_PRIVATE_KEY=nsec1...
BUZZ_RELAY_URL=https://mycommunity.communities.buzz.xyzAdd the key to your .env, never to a config file or a log.
hermes gateway start
hermes gateway statusMessage the agent directly in Buzz. If it responds and correctly identifies itself, the connection is live.
One rule that matters more than any of the setup steps: the keypair is the identity. If it leaks, whoever holds it can post as your agent. Store it like a production secret, and keep Buzz Desktop's auto-approve mode to solo use only, once more than one person can see the output, move to the relay bridge or native gateway instead. Read the full guide.
P.S. You can access all my prompts, workflows and AI trainings if you upgrade here
Mercury launched Mercury Cloud in invite-only beta, letting people manage agents remotely and connect them with each other or a friend's agent across 300+ LLMs, including 10+ free models with unlimited usage.
Gemini Notebook completed its rollout of the more powerful Notebook experience to all Pro users, with the upgrade now reaching free users as well.
Genspark open-sourced GenOffice, a free, ad-free AI office suite covering Docs, Sheets, Slides, and PDF, built by one engineer in one week for $10,000 in tokens.
🚀 v0: A new API gives programmatic access to v0's app-building, starting a chat from a prompt, repo, or ZIP and deploying straight to Vercel.
📜 Nous Research: Hermes Agent v0.20.0 ships real-time voice with wake words, cited research, and agent-to-agent protocol support
📊 Kimi.ai: Kimi Slides builds an entire deck end to end, structure, research, polished charts, and a file ready to download.
🎬 Motion: Drop a product URL into Claude and get back a fully scripted, voiced, and rendered launch video in one message
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial