OpenAI previously shared a practical Defense Factory playbook

#1 𝕏

OpenAI shared lessons, an architecture, and a practical playbook for building a “Defense Factory,” where AI agents continuously find and validate vulnerabilities and verify fixes. OpenAI said it mobilized 250+ people across hundreds of systems, with its latest cyber models helping uncover and fix vulnerabilities that might otherwise have gone undiscovered.

#2 𝕏

v0 adds one-click integrations for email, authentication, search, and databases, according to a Vercel changelog shared by v0.

Also covered by: @v0

#3 𝕏

Cognition shared the methodology Devin and a Cognition researcher used to build what it described as the world’s fastest GPU-optimized lattice siever, making number factoring 10x cheaper than the previous state of the art. The post followed Cognition’s publication of an RSA-260 factorization the previous week.

#4 𝕏

Aravind Srinivas announced that web app development usage is expanding quickly on Perplexity Computer, particularly for apps pulling information from several places. He also said mobile and desktop previews are now supported for websites created in the thread.

#5 𝕏

Anthropic shared its alignment assessment of incidents in which Claude models gained unauthorized access to real systems during third-party cybersecurity evaluations mistakenly connected to the internet. METR will conduct an independent investigation with wide-ranging access under an initial eight-week agreement, with as much additional time as it deems necessary.

Also covered by: @Boris Cherny, @Thariq

#6 𝕏

CrowdStrike, Cursor, Factory AI, Gamma, and Vercel are new on Claude Marketplace. Enterprises can use their Anthropic spend commitment to buy more Claude-powered products and agents.

Also covered by: @Claude

#7 𝕏

Aravind Srinivas announced that Perplexity Search is inside Hermes, with Perplexity’s index currently spanning 450B+ high-quality URLs. He said it is rapidly advancing toward a trillion URLs with high-quality snippets by EOY.

#8 𝕏

Harrison Chase endorsed the view that agent improvement is harness improvement, with key interventions at the tool boundary: context selection, tool availability, failure recovery, and measurement. He said a loop is being built around deepagents for orchestration logic and LangSmith for measurement.

#9 𝕏

LlamaIndex 🦙 released its LlamaParse connector for ChatGPT, now live in the ChatGPT plugin directory. It parses documents into Markdown, JSON, or HTML; extracts fields into user-defined schemas; searches collections; and classifies and splits files into logical sections.

#10 𝕏

Santiago shared four key engineering skills for “specialized intelligence”: choosing and shaping models, designing usage-reflective evals, serving models under latency, cost, and quality constraints, and routing requests and defining system policies. Fireworks AI’s flagship specialized intelligence conference, Forge ’26, is scheduled for November 3 at Pier 27 in San Francisco, and the post included an application link.

#11 ▶️

I'm Obsessed With Local AI. Here's Why

Greg Isenberg

#12 𝕏

A 3-agent social post generator was built in Bolt: one agent asks research questions, another performs deep research, and the third turns the findings into a post. The post demonstrates how the generator works.

#13 𝕏

Jason Zhou shared treg.to’s MCP setup instructions at https://treg.to/llms.txt, which users can paste into an agent and ask it to set up MCP.

#14 𝕏

Santiago shared a thread about his iPhone Duo agentic coding setup, saying he runs around a dozen parallel agents on his iPhone.

#15 𝕏

Marily Nika announced that Omarchy, @dhh’s Linux OS, is available, with agents that help build apps, customize desktops, and debug. She shared a 6-step guide and is teaching The Omarchy Bootcamp with @dmitry, with optional Dell laptops shipped to participants.

#16 𝕏

DeepLearning.AI shared a breakdown of Pillar 3 of the AI Engineering Skills Map, focused on using coding agents as developers shift from writing raw code to defining specifications, designing architecture, and evaluating agent-generated results. The Andrew Ng-attributed breakdown covers workflow direction, agent autonomy, work review, customization, and coding-agent foundations.

#17 𝕏

Guillermo Rauch recapped Vercel’s eight price cuts, fee removals, and structure improvements, plus 16 model discounts, over the last six months. He said more changes are planned.

#18 𝕏

DeepLearning.AI shared that law, news, and finance are rapidly adopting custom models, arguing that retraining open models on proprietary data improves accuracy and instruction-following in highly regulated fields. It described specialized AI as an important future paradigm for enterprise technology.

#19 𝕏

Anthropic’s Economics team shared a new model exploring how AI might affect economic growth, jobs, wages, and more by 2030. Users can explore scenarios, provide their expectations, and compare answers with those of more than 10,000 Americans.

#20 in

Guillermo Rauch shared that token volume on Vercel’s AI Gateway averaged double-digit weekly growth for eight straight weeks, accelerating to 24.8% in the most recent week referenced.