OpenAI Introduces GPT-5.4 Mini and Nano

#1 📝 OpenAI News

Introducing GPT-5.4 mini and nano - OpenAI announces GPT-5.4 mini and nano, smaller variants of the GPT-5.4 family designed for more efficient deployment while retaining advanced capabilities. The release targets use cases needing lower latency and resource usage.

Also covered by: @Simon Willison

#2 𝕏

OpenAI released GPT-5.4 mini today in ChatGPT, Codex and the API—optimized for coding, computer use, multimodal understanding and subagents. It runs twice as fast as GPT-5 mini.

Also covered by: @Simon Willison

#3 𝕏

Mistral AI introduced Forge, a platform that lets enterprises build frontier-grade AI models grounded in their proprietary data and workflows.

#4 𝕏

Cursor trained Composer to self-summarize via reinforcement learning instead of relying on a prompt, cutting compaction errors by 50% and enabling it to tackle coding tasks requiring hundreds of actions.

#5 𝕏

NVIDIA AI: Snap leverages NVIDIA cuDF to accelerate Apache Spark on Google Cloud—achieving 4× faster runtimes, 76% cost savings, and analysis of 6,000+ metrics per A/B test.

#6 𝕏

Mustafa Suleyman unveils Microsoft’s AI reorg around frontier superintelligence models, consolidating all Copilot efforts under @JacobAndreou with a new Copilot Leadership Team. He positions this shift to accelerate an “agentic revolution” in AI products and research.

#7 𝕏

NVIDIA AI is teaming up with AT&T, T-Mobile, Comcast, and Spectrum to deploy distributed AI inference grids on NVIDIA infrastructure, enabling real-time applications from PersonalAI, Linker Vision, ServeRobotics, and DecartAI across telecom networks.

#8 𝕏

Google DeepMind launched a global hackathon with Kaggle to crowdsource new cognitive evaluations for measuring AGI progress, offering $200K in prizes.

Also covered by: @Logan Kilpatrick

#9 𝕏

Logan Kilpatrick launched the Kaggle Measuring AGI competition to create rigorous new benchmarks evaluating AI across learning, metacognition, attention, executive functions, and social cognition.

Also covered by: @Logan Kilpatrick

#10 ▶️

Building AI Agents that actually work (Full Course)

Greg Isenberg

#11 📝 HumanLayer Blog

Getting Claude to Actually Read Your CLAUDE.md - Shows how to use conditional XML blocks inside CLAUDE.md to ensure Claude attends to the right instructions at the right times. These patterns help make agent behavior more reliable by prioritizing context and instructions.

#12 📝 Simon Willison

Subagents - Explains how subagents help manage LLM context limits by splitting larger tasks into smaller agentic components, preserving top-level context and improving results when handling tasks that exceed model memory constraints.

#13 𝕏

LlamaIndex 🦙 launched LlamaParse, adding visual grounding to document parsing with bounding‐box citations so you can hover in the UI to see exactly where each element came from.

#14 📝 Eleanor Berger & Isaac Plath

X1PM: A Shared Workspace for Humans and AI Agents - Introduces X1PM, a shared workspace concept for human and AI agent collaboration using file-system native formats like Markdown and CSV. Presents the idea as the latest topic on the site, inviting readers to explore workflows that integrate agents with familiar file formats.

#15 𝕏

Aravind Srinivas launched Comet, an AI Browser for the enterprise deployable to thousands via MDM and integrated with CrowdStrike Falcon to detect suspicious links and files. It also offers granular controls over what agents can do and where they can operate.

#16 𝕏

Lenny Rachitsky released his complete newsletter archive (350+ posts) and 300+ podcast transcripts as AI-friendly Markdown files, alongside an MCP server and GitHub repo complete with a developer build challenge.

#17 𝕏

Google Research ran a large-scale evaluation of its mammography AI across multiple NHS UK screening services, finding it boosts cancer-detection accuracy and cuts radiologist workload in complex double-reading workflows.

#18 𝕏

DeepLearning.AI Andrew Ng proposes a Stack Overflow–style platform where AI coding agents share learnings to boost documentation and each other’s performance. OpenAI also launched GPT-5.4 with enhanced coding and agentic skills.

#19 𝕏

Guillermo Rauch forecasts that chat and voice AI agents will only become more central to our work and life. He introduces OpenClaw v0 on Vercel—a Generative UI agent serving as the lobster in the coal mine.

#20 𝕏

clem 🤗 questions why major messaging apps (Slack, WhatsApp, Gmail, LinkedIn) still lack AI-powered suggested replies and autocompletion despite having your history and voice, noting on-device open-source models could be a cost-effective fix.

#21 𝕏

Tal Raviv uses Anthropic’s Claude to automate his core PM workflows—drafting specs, prioritizing backlogs, and generating roadmaps—arguing that Claude now outperforms him so fully he might as well “give away his Legos.”

#22 in

🥞 Carl Vellotti used Anthropic’s Claude to parse a week of his Slack messages and meeting transcripts, identify inefficiencies (like unnecessary meetings and redundant status updates), and codify his PM routines in a CLAUDE.md file.

#23 𝕏

PM Diego Granados lays out real PM job postings with salary benchmarks, decodes what hiring managers seek, and offers a detailed, step-by-step playbook for each career path.

#24 in

Dharmesh Shah demos HubSpot’s in-product Breeze AI Assistant (formerly ChatSpot) creating a custom Company property with a single prompt—turning a once-manual setup into an instant, joy-inducing automation.

#25 𝕏

Santiago launched Claude’s new Dispatch feature on mobile—just update the app to enable it and follow the in-app prompts—and you can also scan a QR code from the desktop app’s Dispatch tab to link your phone.