{{first_name | Hey}}, welcome back.
In today’s issue, DeepSeek opens its inference playbook, Anthropic introduces an AI agent that works in place of you, critical PHP bugs trigger emergency patches, and ChatGPT gets a new way to lock private chats.
These are today's updates.
DeepSeek open-sources its inference framework
Anthropic launches a proactive AI agent that works across your apps
Critical PHP vulnerabilities trigger emergency patches and service disruptions
Tools, jobs, resources, and last issue’s winning prompt ⬇️
Spotlight: HubSpot Workshop
Image source: HubSpot
HubSpot is hosting a hands-on Agent Builder Workshop. You'll get to work with HubSpot's Data Hub product team to build an AI agent using your CRM data, connect it to the tools it needs, and put it to work on a real business workflow.
You’ll learn how to:
Turn a business goal into a working AI agent
Use agent instructions, actions, and knowledge to shape its behavior
Connect your agent to external tools so it can take action on your behalf
Design and test an agent using a customer health scoring use case
It's happening on October 29, 2026, at 10 AM EDT
Top News
DeepSeek open-sourced DeepInfer, the framework it uses to serve V4.1 Flash in production, under Apache 2.0. DeepSeek says one cluster handles 155 trillion tokens a day and brings in about $38 per GPU hour at off-peak prices. It runs on Nvidia Blackwell chips and Huawei Ascend 950.
DeepInfer is most relevant for teams running AI models on their own infrastructure. However, its minimum efficient setup requires 64 GPUs per stage, and the release lacks caching and load-balancing code. Teams will need to weigh those requirements before adopting it. It came a day after a US order to remove DeepSeek from federal civilian agencies. Explore DeepInfer here.
Anthropic opened a research preview of Claude Duet, built on Claude Opus 5.5. It sends a morning brief across your connected apps, checks them every few hours, and messages you only when something needs a decision. It runs on its own virtual machine, so tasks continue even when your laptop is closed.
Professionals and small business owners can use it to track the numerous tasks they run and find it difficult to handle, since it works with over 6,000 connectors. It asks for approval before sending, paying, signing or deleting, and background checks count toward your usage limits. Access is by waitlist for US Pro and Max users. Join the waitlist.
PHP, which runs about 70% of websites, shipped emergency patches on Oct 21 for two critical remote code execution bugs, CVE-2026-109618 and CVE-2026-109619. Attackers used AI tools, without logging in, to find flaws present in the code for years. Finland temporarily shut government web services while the BBC and Edinburgh council took pages offline.
Node.js, GitLab, LiteLLM, a Japanese certificate authority and several hosting platforms also saw fixes or incidents. Microsoft says a China-aligned group used an AI pen-testing framework to reach privileged cloud access in under three hours. Read to know why these bugs are causing a cyber emergency and the steps to stay safe.
Don’t miss this: ChatGPT now lets you lock chats behind a four-digit PIN, Face ID or Touch ID. Locked chats move to their own section with their own memory, away from the normal chats. It is available on ChatGPT’s Free, Go, Plus, Pro, Business, Enterprise and Edu plans. Read this post to know what it does not protect.
Signals
Tools
Gemini 4 Argon: Excels at deep work in complex, real-world tasks
Mistral Voxtral 3: Mistral’s speech models for transcription, TTS and agents
Loom Record for agents: Hit record and hand off to AI, now in the new Loom app
Resources
Poll
Choose a prompt you want us to publish next
Market
Funding
Whop riased $200M to expand globally and improve payment and advertising markets
Deepseek raised $15B to expand model training and scale computing infrastructure
Roles in AI
Tutorial
Can AI make people more productive without pushing them out of work? MIT economist and Nobel laureate Daron Acemoglu explores what businesses get wrong about AI adoption, why building useful applications matters more than chasing bigger models, where the real opportunities lie, and which skills will help you stay valuable in the AI era.
Prompt of the Day
AI Security Questionnaire Assistant
When to use this?
Use this prompt when evaluating a new AI vendor or responding to a customer's security questionnaire. It helps you draft evidence-backed answers, spot AI-specific security risks, flag missing information, and identify what your security, legal, or compliance teams need to review before approval.
**Role:** You are an AI security and vendor risk assessment assistant helping enterprise AI leaders, IT managers, security teams, and mid-level executives evaluate AI tools and respond to security questionnaires.
**Your task:** Help me complete AI security questionnaires accurately, using available company policies...
{Click on the button to view full prompt}Get more such prompts in the Prompting Playbook (free for you)
Stay curious, {{first_name | reader}}
PS. If you missed yesterday’s issue, you can find it here.
