In partnership with
SPD-BEEHIIV:32170096-a75f-4041-b102-be58035bf8cb:R21:e5c38eaa611d09f54a90612dDuet’s US preview has a waitlist—and background tasks draw from the same limits as your regular chats.
| Daily issue | Thursday, October 22, 2026 |
| | Our toolsSuperpower ChatGPT/WFH.team/Snipman | Today's briefing What matters todayClaude is taking on work between chats, while researchers use AI to pursue mathematical proofs and biological leads. Today’s lineup also brings a payroll warning for young workers, a critical PHP patch and a compromised LiteLLM release. | Inside today's briefing | | 01 | Duet’s background tasks count toward existing Claude limits, and invitations are rolling out only to US Pro and Max subscribers. | | 02 | Researchers—not autonomous agents—selected ideas and verified the argument in the AI-assisted proof of the KLS conjecture. | | 03 | August payroll data show a 4.4% annual employment decline for young workers in highly AI-exposed roles, without establishing AI as the cause. | | 04 | LiteLLM users exposed to its compromised release should rebuild affected environments and rotate accessible secrets. |
|  | Lead story / launch Anthropic opens a waitlist for Claude to keep working between chatsAnthropic opened a waitlist on October 21 for Claude Duet, a research-preview assistant that can prepare work and continue tasks while users are away. It is included for US Pro and Max subscribers through standalone mobile apps and claude.ai. Invitations will arrive in waves, with no guaranteed access date; Team and Enterprise subscribers are not eligible yet. Duet uses its own cloud computer and browser, so work can continue when a user’s phone or laptop is off. It inherits memory, skills and connected apps from the user’s Claude account, sends a morning brief and checks connected accounts every few hours. Larger assignments can return as editable research briefs, financial models, slide decks or dashboards. A separate Tasks tab highlights work needing input, and only the user can mark a task complete. Reading, drafting and preparation can happen without another prompt. Anthropic says Duet waits for approval before sending messages, making payments, signing documents or deleting information; securities trading and transfers between accounts remain prohibited even with permission. The company also warns that emails and webpages can contain malicious instructions and that no defense against them is perfect. The preview has no extra fee, but it also has no separate usage allowance. Briefs, account checks and other background tasks draw from the same five-hour and weekly limits as ordinary Claude chats, so subscribers may exhaust them sooner. Checking every few hours helps contain consumption, but urgent messages can arrive between checks. Users also cannot watch or take over Duet’s browser during the preview, and Anthropic acknowledges it may misread threads or misjudge urgency. Read full story ↗ |
|
|
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.
 | research Researchers prove a 31-year-old math conjecture with AI assistanceMicrosoft Research says Xinzhi Zhang and Zhao Song proved the KLS conjecture with help from AI agents that explored more than 100 approaches. The researchers chose directions, assembled the argument and verified the proof. A later machine-checked formalization combined their criterion with additional research. Results that previously assumed KLS can now drop that assumption, while the guided-versus-unguided AI comparison remains the researchers’ own account. Continue reading ↗ |
|  | A tool for your workflowSuperpower ChatGPT | | Search, organize, and export your ChatGPT history without breaking your flow. | | See features |
|  | research Stanford tracks a 4.4% employment drop for young workers in AI-exposed rolesU.S. employment among 22- to 25-year-olds in the most AI-exposed roles fell 4.4% year over year in August, according to Canaries Dashboard findings. Those roles have recorded monthly contractions since October 2023. The Stanford–ADP dashboard uses payroll records as an early-warning tool, not evidence that AI caused the decline. Its separate signal of slower job growth for young women remains unconfirmed and is not reflected in longer-term government statistics. Continue reading ↗ |
|  | security risk PHP fixes critical flaws that AI helped turn into server-takeover attacksPHP released a patch after researchers demonstrated server takeover without logging in on WordPress, Drupal and Joomla deployments. Wiz says AI helped adapt working exploits across applications in hours; the flaws sit in PHP itself, not website code. The four bugs can also expose sensitive process memory, and Wiz calls the update the only complete remediation. PHP’s broad usage is not a measure of how many sites are exploitable. Continue reading ↗ |
|
|
Stop Making Shoppers Spell It Out
“Running shos” shouldn’t return zero results. Neither should “red waterprof trail runners under $150.”
Traditional search needs exact keywords. Gladly understands what shoppers mean, typos and all, and turns dead-end searches into products that fit what they asked for.
| | Beyond the feed / AI events | | | The Claude Impact Lab Hackathon will bring small-business leaders, experienced professionals (including technical builders), and students or recent graduates together for a one-day sprint creating open-source AI tools to tackle real business challenges. Sponsored by Anthropic with WTIA, the event will feature team demos and judging, with Claude API credits for participants and prizes for the winning team. | | See the event → Explore more events |
| Daily tool drop 6 AI tools worth knowing today | |
| | | The Internet Had a Point | ChatGPT’s awkward answer to a crush question |  |
| |
| Your daily challenge Shift / Easy / 2–5 min Change one letter at a time. Can you get from LEND to PAVE in 4 moves? Take the challenge →Explore all games | | AI PULSE / YOUR PERSPECTIVE Would you act on an important AI recommendation without independently checking it? Take part → | Reader check-in Help shape tomorrow's briefingOne click tells us what to keep, improve, or tighten. | |
| Prefer one email a week? Get the Monday briefing, including Friday and weekend news. |
|
