In partnership with
OpenAI used its developer event to launch agents that keep working after you log off. Also today: a rival model that nearly matches Claude Mythos at cyberattacks, a mode that makes OpenAI's flagship model up to eight times faster, and a prompt that turns your chores into agent briefs.
Your AI prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning (constraints, edge cases, examples, tone) and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
Used by teams at Microsoft, Amazon, and Notion. Try Wispr Flow free on Mac, Windows, iPhone, and Android.
TODAY IN AI
3 stories worth your next 90 seconds
1. OpenAI launches Dots, always-on AI agents that work 24/7.
Each dot runs on GPT-6 Astra with its own cloud computer, and it can do background research with read-only tools while you are away. It connects to over 4,000 apps, lives in ChatGPT, Slack and Teams, and can message or call you first. The first dot is included at no extra cost for Pro and Business Premium subscribers, with a beta for Enterprise, Edu and Healthcare. Read OpenAI's announcement.
2. Anthropic finds a rival model, GLM-5.3, nearly matches Claude Mythos at cyberattacks.
On an exploit benchmark, GLM-5.3 built working attacks in 50 of 410 attempts, against 56 for Claude Mythos Preview. Researchers also used it to find previously unknown flaws in a browser's JavaScript engine and chain them into a working exploit in about a day. Anthropic says it will likely give attackers these capabilities without meaningful restrictions, and notes all tests ran in sandboxed environments. Read Anthropic's research.
3. OpenAI adds Ultrafast mode, up to 8x faster than standard speed.
The guide says Ultrafast runs up to eight times faster than Standard mode and works with GPT-6 Astra, with GPT-5.6 Sol in preview access. It costs more, so OpenAI recommends it only when speed justifies the price. It is open to all API users, though at low starting rate limits. See the Ultrafast guide.
10 AI Stocks Investors May Regret Ignoring
AI has already produced some of the stock market’s biggest winners, but the opportunity may be far from over.
This free report reveals 10 AI stocks positioned for the next wave of investment and explains how each business makes money, what could drive growth and which risks investors should understand.
FROM THE FRONTIER
OpenAI wants agents that never clock out. Are you ready?
The always-on bet. A dot keeps working toward a goal around the clock, researches in the background and can reach out to you by message or voice call. That is a different product from a chatbot, which waits for you to type.
The builder side. Developers get the same engine through the Agents API, which creates a cloud agent in a single call and charges only for the tokens and tools it uses. It is in public beta, and Amazon's Bedrock managed agents now offer the same Codex harness in preview.
The speed lever. Agents make many calls in a row, so slow responses add up. That is the case for Ultrafast mode, which OpenAI says is up to eight times quicker and recommends using over WebSockets for tool-heavy work.
The other side. The same week, Anthropic showed that GLM-5.3 reaches near-Mythos attack skills without comparable safeguards. Agents that act for you around the clock and agents that attack at scale run on the same underlying skills.
The reframe. The question is no longer what an agent can do. It is what you will let it do when nobody is watching, and most people have never written that down.
IN THE KNOW
What people are actually watching and sharing
An agent with a thumb. TapKit is a Mac app that lets AI agents see, tap and type across real iPhones, including downloading apps and handling two-factor codes. It costs $49 per month for each iPhone.
Save your reset. Codex now has banked resets, one-time refreshes of your usage limits that sit in your account until you use them. Unused resets expire and cannot be restored.
Agents with ID badges. In Amazon's Bedrock managed agents preview, each agent gets its own AWS identity and defined permissions, which is the kind of guardrail the reframe above calls for.
Pick your sandbox. The Agents API runs in OpenAI-managed sandboxes, your own compute, or partners including Modal, Cloudflare, Vercel and E2B.
Fine print on speed. Ultrafast mode supports US data residency and global processing only, so teams that need EU endpoints cannot use it yet.
State of Product reveals what AI still hasn’t solved
80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.
PROMPT STATION
Turn any repeat task into an agent brief
An always-on agent is only as good as its instructions, and vague instructions are how agents wander off. This prompt works in Claude, ChatGPT or Gemini and turns a chore you repeat into a clear brief with boundaries built in. Try it on the weekly task you would happily hand off tonight.
COPY AND PASTE THIS PROMPT
Act as a careful workflow designer who writes instructions for AI agents.
THE TASK I REPEAT: [DESCRIBE THE TASK]
HOW OFTEN IT HAPPENS: [FREQUENCY]
APPS AND FILES INVOLVED: [LIST THEM]
Write an agent brief with these parts: the goal in one sentence, the steps in order, what the agent may read, what it must ask my permission before doing, what a good result looks like, and how it should report back in under 100 words.
The agent must never spend money, send messages or delete anything without my approval. If any detail is missing, ask me before writing the brief.
Swap [DESCRIBE THE TASK] with something like "sort my inbox and draft replies", "update my sales sheet from order emails" or "collect competitor price changes". For a sharper result, add the line "List three ways this agent could go wrong and how the brief prevents each one."