This is Talking AI—your no-nonsense source for making sense of AI. I'm Omar Shanti, CTO at HatchWorks AI.
Let’s dive into the latest AI breakthroughs, practical applications, and strategies for future growth.
This Week's Insights
Physics-aware AI video with consent-based cameos
A smarter safer coding model
A $14B GPU capacity deal
A prompt-driven video creation at scale.
AI progress stalls and human bottlenecks outweigh breakthroughs
🎥 AI Video Hits Its Inflection Point
This week marks a pivotal moment for generative video. OpenAI has unveiled Sora 2, a major leap in realism and controllability, capable of modeling physics with surprising accuracy and enabling multi-shot storytelling.
It powers the new Sora app, now rolling out in the U.S. and Canada, where users can create, remix, and cameo in scenes with remarkable fidelity.
Importantly, OpenAI is introducing strict safeguards. Individuals can only appear through verified cameos, a policy designed to protect likeness rights while still encouraging playful remix culture.
Meanwhile, YouTube is also betting big on AI. On its 20th anniversary, the platform announced AI-driven tools that let creators generate video directly from prompts.
Together, these moves suggest the camera is no longer the gatekeeper of creation.
The question now: how do we balance unprecedented creative power with authenticity, safety, and trust?
The State of AI 2025 – Q3 Edition is out, and it captures a moment of both progress and friction in enterprise AI.
Adoption is everywhere, but execution remains uneven: while individual usage is soaring, ChatGPT is nearing 700 million weekly active users, enterprise programs often stall, with 95% of pilots failing to show measurable business impact.
This quarter’s analysis highlights why: agents struggle in production without orchestration, models are plateauing while architectures get smarter, and the real bottleneck is people, not technology.
At the same time, we’re seeing breakthroughs in multimodality, context engineering, and invisible UX that redefine how AI is designed, deployed, and experienced.
The takeaway is clear: winning won’t come from chasing the next big model, but from building structures, governance, and teams that harness the tools we already have.
Stay updated with the latest developments in artificial intelligence.
Anthropic launched Claude Sonnet 4.5, calling it their smartest, safest model yet. It codes better, sustains 30-hour workflows, resists prompt attacks, and feels less like a tool and more like a colleague.
CoreWeave locked in a $14 billion, multi-year AI infrastructure deal with Meta to power future GPU workloads. It underscores how critical backend scale is to the modern AI stack.
Microsoft brings “vibe working” to Copilot. Agent Mode and Office Agent unlock true human-agent collaboration, orchestrating multi-step tasks in Excel, Word, and PowerPoint for expert-level results.
Google’s AI Mode now handles visual prompts with multimodal search. Stronger context understanding makes it better for shopping, refining tricky queries, and turning images into actionable insights.
Matt Paige talks with Noca AI co-founders Avishai Gelley and Benny Touboul about making AI-driven automation intuitive.
They demo Noca’s Web App Builder and Flow Builder, show how prompts create real-time automations, and stress human-in-the-loop oversight.
The conversation explores bridging technical gaps, democratizing development, and shaping the future of AI-driven UX.
Our guide makes master data governance practical: define policies, roles, and quality standards; pair governance with MDM execution; start small, show quick wins, then scale.
With clean customer, product, and financial masters, teams cut errors, align reports, and enable AI that leverages proprietary data via RAG, turning messy silos into reliable, self-serve analytics and insights.
