Alibaba Qwen Updates Qwen3-30B-A3B to 256K-Token Context
Today's curated insights on AI product management from X/Twitter across 60+ expert sources and YouTube channels.
Alibaba Qwen Updates Qwen3-30B-A3B to 256K-Token Context
From X
AI Product Launches & Updates
Study Mode in ChatGPT: OpenAI @OpenAI announced a new interactive learning experience that guides users through problems step-by-step with Socratic questioning, hints, and scaffolded responses.
Cursor 1.3 Release: Cursor AI @cursor_ai rolled out Cursor 1.3, adding collaboration with Agent in the terminal, visibility into context window usage, and up to 25% faster edit speeds.
Qwen3-30B-A3B Update: Alibaba Qwen @Alibaba_Qwen upgraded Qwen3-30B-A3B for enhanced reasoning, coding, math, multilingual support, and long-context understanding up to 256K tokens.
AI Tools & Applications
Align Evals in LangSmith: LangChain AI @LangChainAI introduced Align Evals to simplify building LLM-as-judge evaluators by matching model scores to human preferences.
Gemini CLI Plan Mode: Philipp Schmid @_philschmid added dedicated slash commands for plan-driven development in the Gemini CLI, enabling custom templates for planning and implementing features.
Guidde Documentation Tool: There's An AI For That @theresanaiforit launched Guidde, an AI-powered extension that automatically captures workflows, generates step-by-step knowledge base articles, and even adds AI voiceover.
Product Management Insights & Strategies
LLM Prompt Best Practices: Paweł Huryn @PawelHuryn outlined strategies to define personas, context, and success criteria for LLMs and embed tool descriptions as micro-prompts to guide agent behavior (@PawelHuryn).
Comet Default Search Strategy: Aravind Srinivas @AravSrinivas set Perplexity as the default search engine in Comet browser to drive Chrome user conversion and capture omnibox queries as a growth lever.
Product Trio Efficiency: Teresa Torres @ttorres emphasized that product trios—cross-functional teams of PM, designer, and engineer—deliver superior outcomes through collaborative efficiency.
AI Industry Developments & News
Buildathon AI Competition: DeepLearningAI @DeepLearningAI announced the Buildathon on August 16 in Menlo Park, partnering with @AI_Fund for a rapid AI product-building event featuring a keynote by Andrew Ng.
Anti-AI Bias on TikTok: Claire Vo @clairevo highlighted TikTok as the most anti-AI platform, critiquing its algorithm for suppressing AI innovation discourse.
"How I AI" YouTube Growth: Claire Vo @clairevo celebrated 25,000 subscribers and nearly half a million views on her AI-focused channel, teasing upcoming high-profile guests and episodes.
From YouTube
The end of Microsoft Excel's 40-Year Dominance? (Shortcut AI is INSANE)
Greg Isenberg • July 29, 2025
Greg Isenberg and Nico Christie demo Shortcut, an AI-powered spreadsheet tool embedded in Excel and Google Sheets that automates complex financial modeling—like DCF updates and P&L projections—and custom dashboards in minutes with full data traceability.
Key Takeaways:
- In a live demo, Shortcut updated a multi-page DCF model by downloading and parsing SEC 10-K filings, fixing circular references and extracting historical and forward data through 2029 in under 10 minutes, citing each value back to its PDF source.
- From a simple English prompt, Shortcut generated a multi-sheet monthly utilization and profitability dashboard for a design agency—calculating employee hours, utilization rates, revenue and costs, applying conditional formatting, and benchmarking against industry data—in minutes.
- Shortcut operates as an end-to-end AI agent within Excel and Google Sheets, claiming to perform up to 90% of routine spreadsheet tasks and transforming non-experts into “Excel gurus” while slashing hours of work to minutes.
Handling Missing Data in R | R for Data Analytics Series
Lex Fridman • July 29, 2025
In this video, Lex Fridman demonstrates how to handle missing data in R using the tidyr package from the tidyverse, including converting blank strings to NA and counting missing values. He then shows how to remove rows with drop_na() and impute missing numeric values with zeros or the column mean, as well as fill character NAs with placeholders like "unknown".
Key Takeaways:
- Specify na.strings = "" when reading the CSV to convert blank fields into NA values for accurate missing data detection.
- Use tidyr::drop_na(data, email) to remove rows lacking email addresses, e.g., dropping customer ID 109 (Helen Carter) from the cleaned data frame.
- Impute missing transaction_amount values by assigning zeros or the column mean with is.na(df$transaction_amount) and mean(df$transaction_amount, na.rm = TRUE), and fill missing customer_name with "unknown".