In partnership with
Good morning! Here's what's happening in AI today:
xAI rolls out Grok 4.6, a significant upgrade over 4.5 at the same price
Claude's Chrome sessions now sync across desktop, web, and mobile
Qwen releases its biggest open-weight model yet with a 1M context window
How to use ChatGPT Work and automate your daily tasks
4 new AI tools worth trying today
TOGETHER WITH WISPR FLOW
Talk to your AI tools the way you'd talk to a colleague.
You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.
Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.
Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.
AI MODEL
xAI unveiled Grok 4.6, delivering frontier intelligence with a significant improvement over Grok 4.5 at the same price, outperforming rival models across intelligence indexing, coding, agent, and enterprise research benchmarks.
Grok 4.6 posted higher scores than Grok 4.5 across nearly every benchmark tested, including a jump on CursorBench and a sharp gain on Terminal-Bench, both of which measure real coding and agent task performance.
The model matches or beats competing frontier models like GPT-5.6 Sol Max and Fable 5 Max on several benchmarks, including AA-Briefcase and Harvey LAB legal validation tests.
xAI held pricing steady at the same rate as Grok 4.5, meaning the capability jump comes without a cost increase for existing users.
A meaningful capability jump at an unchanged price is a rare combination in a market where better performance usually comes with a premium. The benchmark spread, especially the gains on coding and agent tasks, suggests Grok 4.6 is aimed squarely at developers and technical users rather than being a broad, general-purpose refresh. Holding price steady while improving performance also puts pressure on competing labs to match value, not just capability, in their next release.
AI TOOL
Anthropic updated Claude in Chrome so sessions now carry over to desktop, web, and mobile, with conversations saved and skills and connectors working directly inside the browser, rolling out today on Max and Team with Pro access coming in the following weeks.
Conversations started in Chrome now persist across every surface, so a session begun in the browser can continue seamlessly on desktop or mobile without losing context.
Skills and connectors, previously limited to certain surfaces, now function directly inside the Chrome side panel, extending Claude Cowork's capabilities into everyday browsing.
The rollout is staged by plan tier, with Max and Team getting access first and Pro users following in the coming weeks rather than all at once.
Cross-device continuity has been a persistent gap for browser-based AI tools, where switching from laptop to phone often meant losing an entire conversation's context. Bringing skills and connectors into the Chrome side panel effectively turns the browser into a full Cowork surface rather than a separate, lighter experience. For anyone who moves between devices throughout the day, this removes one of the more common points of friction in actually relying on Claude for ongoing work.
AI MODEL
Qwen released Qwen3.8-Max, its biggest open-weight model yet at 2.4T total parameters with 95B active, built to take a goal and return with finished work, featuring a 1M token context window, adjustable reasoning, and parallel tool calls.
In testing, the model coded unattended for more than 10 days straight, building a self-evolving harness entirely from scratch without human intervention along the way.
Qwen3.8-Max reproduced a published research paper, then ran a 125-hour autonomous loop and ultimately beat the paper's original results.
The model also took a chip design from RTL through to layout, cutting die area by 81%, a task that normally requires specialized engineering tooling and expertise.
A 10-day unattended coding run and a 125-hour autonomous research loop are well beyond what most released models have publicly demonstrated, suggesting real gains in long-horizon task completion, not just benchmark scores. The chip design result stands out specifically because it's a domain-specific, high-stakes task with a clear, measurable outcome. As an open-weight release, this also puts frontier-level autonomous capability into the hands of anyone able to run it, not just users of closed, hosted models.
HOW TO AI
How to Use ChatGPT Work and Automate Your Daily Tasks
You've dug through Gmail for twenty minutes looking for one thread. You've had five tabs open just to answer one question that lives across an email, a doc, and a spreadsheet. ChatGPT could have answered it the whole time, it just never had access to any of it, and you never gave it a real job to do.
ChatGPT Work is the agent mode OpenAI added inside ChatGPT this July. Instead of answering one prompt at a time, it takes a real goal, works on its own for minutes or hours, and hands back a finished file, not a chat reply you have to reformat. A report. A working spreadsheet. A slide deck. A small web app. Whichever one the job actually calls for.
Setup & Overview
Work is a mode inside ChatGPT, not a separate app and not an extra subscription, running on OpenAI's newest model family and bundled into existing plans at no additional list price.
Turning it on takes one click. Open the ChatGPT desktop app or web, and switch the composer from Chat to Work. That's the whole toggle, no install, no separate sign-in.
One honest note before you connect anything: Work itself is available even on the free plan through the desktop app, but connecting Gmail or Drive specifically needs at least a Plus subscription.
Outputs: Reports, Sheets, Decks & Apps
Work produces four real kinds of output. A written report:
Using @Gmail and @Google Drive, research every active client
conversation this week: what they're waiting on from me, how long
it's been, and what I should send today. Write it up as a short
report.A working spreadsheet, a slide deck, or a small web app, depending on what the job needs. Same mode, four different outcomes, because the job decided the format, not you having to know which tool makes which file type.
Real Use Case: The Weekly Client Status Update
You freelance or consult for more than one client, and every week means writing the same kind of update from memory, half-remembering which thread you left hanging. Schedule it once, and Work pulls the real threads and writes the summary before you've opened your inbox:
Every Monday at 8am, check @Gmail and @Google Drive for anything
client-related that's been sitting unanswered for more than two
days, and send me a short summary.
Read the full guide.
P.S. You can access all the AI training, prompts and workflows if you upgrade.
Google Gemini added support for 14 new connected apps, letting people link their favorite services directly in settings to plan, create, and manage their to-do list all from within Gemini.
Vercel built a software factory for AI SDK where each development step runs as an agent and humans merge the changes, authoring up to 35% of merged pull requests, closing 70% of July's issues, and cutting open bugs by 25% four weeks in.
WeChat introduced WeLM, a family of large language models built by the Weixin team with resource efficiency at its core, including an 80B-parameter version already deployed in Xiaowei, Weixin's native AI assistant, for chat, search, and native function access.
🖥️ Delta: multiplayer environment for coding with agents, powered by DeltaDB, private beta invites going out now.
🚀 Antigravity: define your own custom agents, now available.
🐟 TinyFish: live web search, clean fetch, and web agent, now officially in ChatGPT.
💧 Liquid AI: LFM2.5-VL-3B, a lightweight vision-language model that reads screens, documents, and the physical world.
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