Anthropic's Most Powerful Model Ships to Everyone, and ChatGPT Memory Starts Updating Itself
A Mythos-class Claude is now in your plan, ChatGPT learned to keep itself current, and Grok plugged into your work apps. Here is what is worth your Wednesday.
What matters today
A Mythos-class Claude is now in your plan, ChatGPT learned to keep itself current, and Grok plugged into your work apps. Here is what is worth your Wednesday.
Welcome byte
The biggest model Anthropic has ever released to the public landed Tuesday, and you already pay for it. Claude Fable 5 reads up to a million tokens at once and finishes longer jobs without losing the thread, so a full data room read that used to eat an afternoon becomes a supervised 20-minute pass. It is free inside Pro, Max, Team, and Enterprise until June 22. ChatGPT also rebuilt how it remembers you: the new Dreaming system updates its own memory in the background and doubled capacity for Plus and Pro, so you re-explain yourself less. The Pro Tip is the exact Fable 5 prompt that turns a stack of contracts into a one-page risk memo. The Productivity Gem runs your Monday numbers review in Perplexity on Windows. Below those: Grok plugs into your work apps, an Apple Watch fitness read, and a story-building project for the kids.
Quick Hits
Grok now plugs into your work apps. xAI added connectors for SharePoint, Outlook, OneDrive, Google Workspace, Notion, GitHub, and Linear, plus bring-your-own MCP. Connect the two tools you live in and ask Grok one question that spans both.
ChatGPT added a way to see who is logged in. The new Active Sessions panel shows every device on your account with location and sign-in time, and lets you sign out anything you do not recognize. Worth a one-minute check this week.
Microsoft 365 Copilot's model picker grew. GPT-5.5 Instant and Claude Opus 4.7 are now selectable inside Copilot, so you can match the model to the task instead of taking the default. Set a team default and move on.
Grok shipped Voice to everyone, with voice cloning. Spoken back-and-forth is now on by default, and you can clone a voice from a short clip for the text-to-speech and voice agent tools. Handy for hands-free drafting on the move.
Microsoft made its Copilot small business plans permanent. Business Standard with Copilot and Business Premium with Copilot become standing SKUs on July 1 at 23.50 and 32 dollars per user per month. Budget for it now if you are piloting.
Top Updates
On June 9 Anthropic released Claude Fable 5, the first Mythos-class model available for general use. It holds up to a million tokens of context, produces up to 128k output, and works autonomously on a long task for longer than any prior Claude. High-risk questions in cybersecurity, biology, or chemistry route automatically to Opus 4.8. It is included in Pro, Max, Team, and Enterprise until June 22, then moves to credits; the API runs 10 dollars per million input and 50 per million output.
Action Steps
- Open Claude where you pay and select Fable 5 in the model picker.
- Hand it one long-document job: a data room, a contract set, or a quarter of reports.
- Paste everything at once; the million-token window means no chunking.
- Ask for a one-page output with citations and confidence labels.
- Run a task with a known answer before June 22 to judge the value.
Read the premium deep dive [Premium]
On June 4 OpenAI began rolling out Dreaming, a background process that captures context as it surfaces in conversation, revises stale facts as time passes, and doubles memory capacity for Plus and Pro. Tell it you are traveling in July, and after the trip the memory updates itself. OpenAI reports factual recall climbing from 41.5 to 82.8 percent on its own eval. It is live for Plus and Pro in the US first, with more tiers and countries to follow.
Action Steps
- Open Settings, Personalization, Memory and read what it has stored.
- Delete anything wrong or stale before it propagates on its own.
- Stop re-explaining your role and projects each session.
- Use a temporary chat for anything sensitive you do not want remembered.
- Spot-check what it kept after a week.
Read the premium deep dive [Premium]
On June 4 xAI rolled out connectors in Grok Web for SharePoint, Outlook, OneDrive, Google Workspace, Notion, GitHub, and Linear, plus a bring-your-own MCP server option for anything custom. Once connected, Grok reads across those tools inside a single chat, so the most common AI chore, copy-pasting context between tabs before you can even ask your question, disappears.
Action Steps
- Open Connectors in Grok Web and review the available integrations.
- Connect just two tools: your email plus your project tracker.
- Ask one question that spans both, such as open tickets plus the related email.
- Save the prompt and reuse it weekly.
- Clear it with your admin before connecting company tools.
Read the premium deep dive [Premium]
Productivity Gem
Perplexity's desktop agent reached Windows on June 3 and reads your Word, Excel, PowerPoint, and Outlook directly. Point it at your weekly numbers workbook with one reusable instruction and get the five figures that moved this week, each with the old value, new value, and percent change, plus one chart, a plain-language summary, and a data-quality flag, with the math shown so you can verify it. A ten-minute Monday scan becomes a one-minute read, no formulas or pivot tables required.
Read the premium deep dive [Premium]
Put it to work Strategic Brainstorming: Use AI to Generate and Refine Ideas
Pro Tip
Fable 5's million-token window lets it read a full data room in one pass. Drop a folder of long documents, contracts, vendor reports, financials, into a single Claude chat, then run one prompt that plans first, pauses for your approval, and returns a one-page memo: the five most material risks ranked, key obligations and deadlines, and three open questions, each citing the exact source line with a confidence label. You verify by spot-checking five citations, not by re-reading the stack. A half-day read in about 20 minutes.
Read the premium deep dive [Premium]
Copy-ready guide AI for Executive Summaries: Distill Complex Information Fast
Health Tip
Your Apple Watch has been tracking Cardio Fitness, an estimate of your VO2 max, for months. Open Health, then Heart, then Cardio Fitness, set the range to six months, and hand a screenshot of the trend to Claude. Ask it to describe the trend, flag the weeks it moved, ask you what was happening in those weeks, and then name the single lowest-effort change most likely to nudge the number up. Do that one thing for four weeks and re-measure. One change, so you learn what actually works. Not medical advice; consult a professional.
Read the premium deep dive [Premium]
Kids Tip
A weekend project for ages 8 to 16 that teaches branching logic disguised as play. The child maps the story on paper, one start, two choices, three or four endings, then uses the free Gemini tools to draft each scene and generate its art, keeping the same character and changing only the action. They assemble it in Google Slides, linking each choice to its slide so the reader clicks their way through, then debug any link that jumps to the wrong scene. The child is the author and the engineer; the AI is the illustrator. No paid tools or special hardware.
Read the premium deep dive [Premium]
That is the week. The most powerful Claude yet, free in your plan until June 22, a ChatGPT that keeps itself current, and a Grok that finally reads your work apps. Pick one thing above and actually try it before next Wednesday. That is the whole point of this newsletter.
See you next week, Pierre Bradshaw PromptHacker
Before next Wednesday
Three deep dives. Four useful moves. One email worth opening.
PromptHacker turns the AI firehose into practical next steps for work, health, family, and everything time keeps trying to steal.
No comments yet