Three New AI Models, Controversial Apps, and a U.S. Project Compared to the Manhattan Project

In this edition: three major players dropped new models. Google may finally be going head-to-head with OpenAI. Governments are entering the chat. And someone launched an app for “talking” to dead relatives (yes, really).

This is Q&AI, our blog series aimed at keeping you in the know on updates in the rapidly evolving world of AI. Sometimes, these will be quick updates on new developments in the field. Sometimes, they’ll be tips on tactics, features, or functionality. 

If you haven’t met me yet, hi: I’m Jen Taylor, CI’s Director of AI Strategy & Implementation, and your (very human) AI BFF. AI is moving at the speed of light, so I’m here to let you know what matters most now. 

Q: Can you give me the “new models” breakdown? 

A: Sure thing – it’s been a lot!

#1: GPT-5.1 (Released 11/12)

Following the massive launch — and equally massive backlash — of GPT-5, OpenAI rolled out GPT-5.1 quietly. The update is meant to follow instructions more precisely.

My take so far:

• It feels similar to GPT-5, maybe slightly more detailed.
• It still can’t reliably pull or process images embedded in PDFs, which it apparently never did well.
• It can no longer chain CustomGPTs together, a quirky “accidental” feature people loved and now miss.
• It’s a refinement, not a breakthrough — and a bit uneven.

#2: Gemini 3 (Released 11/18)

This model has been the standout launch — and reportedly even triggered an internal memo from Sam Altman telling the OpenAI team to “do better” (leaked, unconfirmed).

This matters because Google has always been a quiet frontrunner: enormous reach, enormous infrastructure, very profitable. Gemini 3 might finally put them head-to-head with OpenAI in the market-share race.

Key updates:

• Switch between Fast and Thinking modes (bottom right of the chat box), similar to GPT-5.
• Strong performance across academic and reasoning benchmarks.

The real excitement is around its image model, Nano Banana Pro — easily Google’s boldest creative-AI step in years.

#3: Claude Opus 4.5 (Released 11/24)

I haven’t tested it yet, but Claude Opus 4.5 is marketed as:

• More accurate
• More reliable
• Less likely to show concerning behavior compared to competitors

It also finally fixed a long-standing frustration: conversation limits in long chats. (If you’ve ever had Claude abruptly “forget” half the conversation, this is welcome news.

Q: What was some recent controversy?

A: An app launched that lets people create AI avatars of deceased relatives and “connect” with them.

I found this deeply unsettling — and was relieved it received immediate backlash. Missing someone is human. But an AI simulation of a dead person is not that person, and framing it as a connection feels harmful, not healing.

Q: What is the U.S. Genesis Mission — and why are people comparing it to the Manhattan Project?

A: Launched on 11/24, the Genesis Mission is a massive government-backed AI initiative leveraging:

• The world’s largest collection of scientific datasets (NASA, NIH, DoD, etc.)
• Extreme-scale supercomputing across 17 U.S. National Labs

The goal: unify siloed scientific data and accelerate breakthroughs with AI.

The timeline is extremely aggressive:

• 60 days: Identify 20+ scientific challenges
• 90 days: Catalog available federal + industry compute
• 270 days: Show initial progress on at least one challenge

The scale, speed, and ambition are why people are drawing Manhattan Project comparisons — though the goals here are constructive, not destructive.

Your Friend,
Jen

Have a question for a future edition? Submit it here!