“DeepLearning.AI shared AREX, a new agentic research model and harness that verifies answers requirement by requirement and researches remaining gaps, reporting 82.5% on BrowseComp and 82.0 F1 on WideSearch-en.”

DeepLearning.AI shared AREX, a new agentic research model and harness that verifies answers requirement by requirement and researches remaining gaps, reporting 82.5% on BrowseComp and 82.0 F1 on WideSearch-en. The harness alone delivered up to +10 points, while a fine-tuned 4B model beat an untuned 35B model on 5 of 6 benchmarks.

“DeepLearning.AI shared its free course “Building AI Assistants with On-Device Memory,” developed in partnership with Qdrant and taught by Dylan Couzon.”

#13 𝕏 DeepLearning.AI shared its free course “Building AI Assistants with On-Device Memory,” developed in partnership with Qdrant and taught by Dylan Couzon. It covers building assistants with on-device memory that requires no network connection or remote server.

“#5 𝕏 DeepLearning.AI recapped Google’s Gemini 3.8 Live models, which listen, reason, and respond in one system while accepting image and video input.”

#5 𝕏 DeepLearning.AI recapped Google’s Gemini 3.8 Live models, which listen, reason, and respond in one system while accepting image and video input. The Extended Thinking version ranks first on Artificial Analysis’ Speech to Speech Index, while the standard version ranks second in blind live conversations judged by people and costs $0.84 per hour of input audio—the index’s lowest price.

“DeepLearning.AI shared that, according to Anthropic, GLM-5.3 solves 12% of ExploitBench exploit tasks versus Claude Mythos’s 14%, while $20.40 in tokens found a recent Google Chrome security exploit.”

#9 𝕏 DeepLearning.AI shared that, according to Anthropic, GLM-5.3 solves 12% of ExploitBench exploit tasks versus Claude Mythos’s 14%, while $20.40 in tokens found a recent Google Chrome security exploit. In this week’s letter, Andrew Ng argues these cybersecurity risks are an engineering problem and that defenders hold the long-term edge.

“The Data Engineering Professional Certificate is now on DeepLearning.AI Deeplearning.ai Now available on DeepLearning.AI, the Data Engineering Professional Certificate was created in partnership with DeepLearning.AI and Amazon Web Services (AWS).”

GenAI PM Daily - Anthropic announces faster, lower-cost Claude Sonnet 5.5 September 29, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 20 insights for PM Builders, ranked by relevance from Blogs, X, YouTube, and LinkedIn. Anthropic announces faster, lower-cost Claude Sonnet 5.5 #18 ▶️ The Data Engineering Professional Certificate is now on DeepLearning.AI Deeplearning.ai Now available on DeepLearning.AI, the Data Engineering Professional Certificate was created in partnership with DeepLearning.AI and Amazon Web Services (AWS). It provides hands-on training across the data engineering lifecycle and cloud data-system architecture for beginners without prior cloud data-systems experience and professionals in data engineering, analytics, data science, or software engineering.

“DeepLearning.AI shared Andrew Ng’s explanation of why AI engineering tactics should adapt to a project’s stage for speed and reliability.”

#7 𝕏 DeepLearning.AI shared Andrew Ng’s explanation of why AI engineering tactics should adapt to a project’s stage for speed and reliability. The letter in The Batch covers evaluation pipelines and metrics, scalable software architecture, and product feedback loops.

“DeepLearning.AI recapped Meta AI’s use of dedicated memory agents alongside primary action agents to combat context rot by keeping structured notes and injecting timely reminders.”

#8 DeepLearning.AI recapped Meta AI’s use of dedicated memory agents alongside primary action agents to combat context rot by keeping structured notes and injecting timely reminders. The approach raised Claude Sonnet 4.5 benchmark performance from 37.6% to 45.9%.

“Building AI Assistants with On-Device Memory Deeplearning.ai Building AI Assistants with On-Device Memory uses Qdrant Edge to build an offline assistant that converts text, voice, and image experiences into embeddings and stores them locally in vector memory for retrieval, filtering, deletion, and object recognition.”

#14 ▶️ Building AI Assistants with On-Device Memory Deeplearning.ai Building AI Assistants with On-Device Memory uses Qdrant Edge to build an offline assistant that converts text, voice, and image experiences into embeddings and stores them locally in vector memory for retrieval, filtering, deletion, and object recognition. The course is built in partnership with Qdrant and taught by Dylan Couzon, Developer Experience Engineer at Qdrant. The local memory system stores notes as vectors with context, performs semantic search, filters memories with metadata, and deletes memories when the system needs to forget them using the open-source Qdrant Edge vector search engine. The assistant combines text notes, voice recordings, and photos; it can retrieve a voice note about an appointment and recognize a new object from stored photo examples without retraining a model, running on Mac, Windows, or Linux without sending memories to a remote service.

“#6 ▶️ Full Course: Spec-Driven Development with Coding Agents Deeplearning.ai Spec-driven development uses a versioned markdown constitution and per-feature specs to direct coding agents through planning, implementation, validation, and replanning on separate Git branches.”

GenAI PM Daily September 23, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 20 insights for PM Builders, ranked by relevance from Blogs, X, and YouTube. OpenAI News announces GPT-6 Sol and Luna #1 📝 OpenAI News Introducing GPT-6 Sol and Luna - Introduces two new GPT-6 variants, Sol and Luna, highlighting their capabilities and intended use cases. The announcement describes distinctions between the models and how organizations can adopt them. Also covered by: @Cognition , @OpenAI , @Sam Altman , @There's An AI For That #2 𝕏 Claude Opus 5.5 is available today. Also covered by: @Thariq , @Mike Krieger , @Anthropic , @Aravind Srinivas , @Cursor , @v0 , @Claude , @There's An AI For That #3 📝 OpenAI News Better prompt caching for GPT-6 - Announces improvements to prompt caching for GPT-6 to reduce latency and cost for repeat prompts, improving developer and product performance. #4 𝕏 Aravind Srinivas shared research on a post-training approach that teaches the Perplexity Computer agent from real user sessions by combining rejection sampling fine-tuning with hint-guided self-distillation. #5 𝕏 Boris Cherny demonstrated using Opus 5.5 to formally verify the Claude Agent SDK with Lean, saying a couple short prompts produced 16 PRs addressing bugs and race conditions. #6 ▶️ Full Course: Spec-Driven Development with Coding Agents Deeplearning.ai Spec-driven development uses a versioned markdown constitution and per-feature specs to direct coding agents through planning, implementation, validation, and replanning on separate Git branches.

“DeepLearning.AI shared an analysis of Meta’s Muse agent, which assumes the model will be tricked and implements security at the operating-system level.”

#8 𝕏 DeepLearning.AI shared an analysis of Meta’s Muse agent, which assumes the model will be tricked and implements security at the operating-system level. The model never handles real credentials, tools run in isolated Linux containers, and an independent gatekeeper verifies outbound calls.