OpenAI announces Zero Data Retention for frontier models

Today's top 20 insights for PM Builders, ranked by relevance from Blogs, X, YouTube, and LinkedIn.

OpenAI announces Zero Data Retention for frontier models

#1 đź“ť OpenAI News

Offering Zero Data Retention for frontier models - OpenAI announces offering zero data retention for frontier models, committing to not retain user data for those models and clarifying how this impacts customers and data handling. The post outlines the company's privacy-focused approach for frontier model interactions.

Also covered by: @OpenAI, @OpenAI, @Sam Altman

#2 đť•Ź

Cursor announced that it can now monitor pull requests, watch a Slack thread, and run scheduled tasks. Cloud agents automatically subscribe to pull requests they create and drive them to completion.

#3 đť•Ź

Mustafa Suleyman announced that MAI-Image-2.5 is ranked #1 on the Artificial Analysis leaderboard for image editing.

#4 đť•Ź

Logan Kilpatrick announced that Google AI Studio now supports GitHub repository imports and bi-directional push/pull synchronization. A new UI also supports force pushes and merges.

#5 đť•Ź

Qwen shared that Qwen3.8-27B ranked as the #1 open-weight model on Harvey’s Legal Agent benchmark, describing it as capable of professional tasks while remaining small enough to run locally.

#8 đť•Ź

Philipp Schmid shared that Gemini 3.7 Flash ranked first on Artificial Analysis’s new AA-AnalystAgent, which covers 80 real-world quantitative analysis tasks across 14 business and scientific domains. It achieved 60.0% pass^5, 70.5% pass@1, and 77.5% pass@5, with the fastest reported speed at 1.32 seconds per task and a middle-ranked average cost of $0.54 per task.

#10 đť•Ź

Madhu Guru shared an eval strategy for AI products: define a rubric, use the best available measurement process to establish a trustworthy quality frontier, then reduce costs through automation, smaller judge models, sampling, and deterministic checks where relevant.

#11 đť•Ź

Santiago shared an unnamed automated AI agent security testing tool for chat, code, and phone-based voice agents that runs 10,000+ jailbreaks and adversarial attacks. It requires only a URL or phone number and continuously retests agents as they change.

#12 ▶️

Biggest Unlock for AI Agents in 2026: Skillsmaxxing

Greg Isenberg

Remy stores company agent skills as Markdown SOPs in a GitHub repository, converts the repository into department-specific plugins with JSON files, and installs those plugins in Claude Code and Codex with auto-update enabled.

  • In Claude Code, users run /plugin, select “Add marketplace,” paste the GitHub repository URL, and install only the required department plugins, such as newsletter or finance skills.
  • Each skill is a .md file with a name and description that helps the agent select it; examples include Notion formatting, AI with Remy brand voice, Resend email formatting, proposal generation, and a YouTube workflow that chains title, thumbnail, and description skills.
  • Remy keeps a separate personal “Remy Skills” GitHub repository for skills such as inbox triage and a morning brief after Claude deleted his local Claude folder containing 150 skills and more than 500 hours of work approximately two months earlier.

#13 đť•Ź

Santiago shared Open Bot, describing it as a free, open-source bot that works with any harness or model and can run anywhere. Users can assign it tasks, which it performs in a real browser using its own credentials and files plus user-authorized tools.

#14 ▶️

Why Chatto.run Is Built to Be Extended

SyntaxGTM

Chatto is an open-source, self-hosted work-oriented chat application with paid company-focused hosting, built through a workflow that shifted from manual coding to AI coding agents.

  • Chatto reached 2,000 GitHub stars within a week after its Hacker News launch, which held the number-one spot for a full day.
  • Hendrik began building Chatto manually with tools such as Copilot code completions, then shifted after the Anthropic 4.5-generation models arrived around October–November.
  • Since approximately February, Hendrik says he has not written code directly; AI agents perform most implementation, while Chatto remains open source and can be self-hosted as the free option.

#15 đť•Ź

Cursor announced that subagents can run on their own virtual machines, each with an isolated project copy. They can test a parent agent’s changes in a fresh environment or swarm independent fixes.

#16 in

Claire Vo recapped a How I AI mini-episode covering Grok Bot, multi-agent versus single-agent identity, whether Cursor Origin could rival GitHub, and Grok 4.6’s design sense. She also discussed why she is keeping OpenClaw for now.

#17 đť•Ź

Guillermo Rauch announced that fx.sh has free GLM-5.2 from @blackboxai for two weeks and 50% off @ChatGPT Sol for a month when using â–˛ AI Gateway. He said usage was much higher than anticipated and that more GLM capacity is being added.

#18 đť•Ź

Sundar Pichai announced more learning features being released that day and highlighted free Google AI subscriptions for college students in 140+ countries.

#19 đť•Ź

Harrison Chase announced a new YouTube playlist titled “Managed Deep Agents,” offering an overview and in-depth coverage of core concepts. It launched with six videos: Intro, Conceptual Overview, Quickstart, Instructions and Context Hub, Skills, and Tools.

#20 đť•Ź

Harrison Chase announced LangChain’s upcoming “Towards Automating Eval and Environment Engineering” webinar, featuring @Vtrivedy10, @willccbb, and @AEllisBloor. The discussion will explore automating more of the agent improvement loop, particularly evaluation and environment engineering.

Get tomorrow's brief first

Join AI product managers receiving the latest brief before it reaches the public archive.

Subscribe free