OpenAI Releases GPT-5.1 API with Prompt Caching
Today's curated insights on AI product management from X/Twitter across 60+ expert sources and YouTube channels.
OpenAI Releases GPT-5.1 API with Prompt Caching
From X
AI Product Launches & Updates
SIMA 2: Our most capable AI agent for virtual 3D worlds: Google DeepMind @GoogleDeepMind detailed how SIMA 2 “goes beyond basic instructions” to reason, understand, and act in interactive game environments.
GPT-5.1 API Release: Sam Altman @sama announced GPT-5.1 is now available in the API (same pricing as GPT-5) with codex and codex-mini variants plus 24-hour prompt caching.
Cursor Integrations with GPT-5.1: Cursor AI @cursor_ai shared that GPT-5.1, GPT-5.1 Codex, and GPT-5.1 Codex Mini are now integrated, powering everyday tasks and ambitious coding workflows directly in Cursor.
AI Tools & Applications
Automatic Anomaly Detection: Guillermo Rauch @rauchg demoed Vercel’s automatic anomaly detection, which catches anomalies & triggers AI-led investigations and pull requests for faster incident response.
Fast Feedback Loop: Lenny San @lennysan highlighted how GammaApp enables teams to build prototypes in the morning and test with ~20 real users by afternoon, streamlining validation.
NotebookLM Deep Research: Josh Woodward @joshwoodward announced that Deep Research is coming to NotebookLM, allowing you to generate report summaries and automatically collect sources with a single tap.
Product Management Insights & Strategies
Career Horizon in AI: Andrew Ng @AndrewYNg shared a student’s question, “Too late to contribute to AI?” stressing that AI remains a growing frontier where early-career professionals and PMs can still make a meaningful impact.
Discipline Over “Vibe Coding”: Phil Schmid @_philschmid shared a thread on 6 months of “hardcore” coding agent usage, noting the ironic takeaway: users moved from free-form prompts back to structured context engineering for better results.
AI Industry Developments & News
First AI-Only Cyber Espionage: Anthropic AI @AnthropicAI revealed they disrupted a large-scale, AI-led espionage campaign without human intervention, marking a new era in AI-driven cybersecurity threats.
Self-Driving Transformation: Andrej Karpathy @karpathy argued that self-driving cars will terraform outdoor spaces, reducing parking lots, boosting safety, and reshaping urban design as one of the next major technology shifts.
GPT-5.1 on Perplexity: Arav Srinivas @AravSrinivas announced GPT-5.1 is live on Perplexity Pro & Max and teased that the Kimi K2 model is “coming very soon” for advanced reasoning use cases.
From YouTube
The PyTorch for Deep Learning Professional Certificate is live
Deeplearning.ai • November 13, 2025
This video announces the launch of the PyTorch for Deep Learning Professional Certificate by Deeplearning.ai, detailing a three-course curriculum that guides learners from PyTorch fundamentals to real-world computer vision and NLP projects, then to advanced architectures and deployment.
Key Takeaways:
- The three-course certificate covers PyTorch fundamentals (tensors, neural networks, full ML pipeline), real-world applications in computer vision and NLP (optimization, pre-trained architectures, transfer learning, fine-tuning), and advanced models and deployment techniques.
- In course one, learners build their first neural networks with PyTorch, explore tensor operations, and execute the complete machine learning workflow from data processing to evaluation.
- Course two teaches optimization of PyTorch models, use of pre-trained architectures, and application of transfer learning and fine-tuning for computer vision and natural language processing tasks.
Local AI Surveillance Is Getting SCARY Good (Qwen3-VL)
All About AI • November 13, 2025
The video demonstrates a local AI security system built with the Qwen3-VL model to analyze IP camera snapshots for generic person detection and specialized attribute checks, then trigger hardware responses—including sounding alarms and deploying a drone to intimidate intruders wearing an orange jacket.
Key Takeaways:
- Using Qwen3-VL 2B on a local GeForce RTX 4080, the system processes 640Ă—360 snapshots every 2 seconds to return true/false on person presence with sub-second latency.
- Specialized detection is achieved by chaining boolean queries—first “is there a person?”, then “is anyone wearing an orange jacket?”—to trigger alarms only for specified targets without additional model training.
- A drone is integrated to automatically arm, take off 1 m, spin towards the frame, and land as an intimidation tactic when the model detects a person in an orange jacket.
“Dumbest idea I’ve heard” to $100M ARR: Inside the rise of Gamma | Grant Lee (co-founder)
Lennys Podcast • November 13, 2025
Grant Lee explains how Gamma overcame investor dismissal to become a $2B-valued, $100M ARR AI presentation tool by revamping onboarding, driving micro-influencer growth, and staying lean and profitable.
Key Takeaways:
- After winning Product Hunt awards in 2022 but stalling, Gamma spent 3–4 months rebuilding onboarding to deliver an AI “magic” moment in the first 30 seconds, spiking daily signups from hundreds to over 20,000 in March 2023
- Grant Lee manually onboarded thousands of micro-influencers—investing ~$20K–30K/month across 20–30 niche creators—to ignite authentic word-of-mouth; today over 50% of user acquisition is organic or direct
- Gamma reached $100M ARR in just over two years with a 30-person team, $2B valuation, and profitability by hiring generalists as “player-coach” managers and rapidly testing prototypes via VoicePanel and UserTesting