Howdy, it’s Barsee.

Happy Wednesday, AI family, and welcome to another AI Valley edition.

Today’s climb through the Valley reveals:

• The US Military just launched a custom Gemini chatbot

• OpenAI is prepping new image models to fix past issues

• New clues suggest GPT-5.2 is around the corner

• Google and OpenAI are fighting off ad rumors

• Meta is reportedly ditching open source for its next big model

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

WISPR FLOW

What if your keyboard retired?

AI has changed how we think, but not how we communicate. Wispr Flow bridges that gap, letting you speak naturally while it formats and edits in real time. It understands context, tone, and corrections, so "actually, make that 6 p.m." updates itself on the spot.

It's not just for writing faster; it's for freeing up the mental space that typing steals. Once you start using Flow, the keyboard starts to feel... optional.

Get Wispr Flow → Write at the speed of thought.

_{*This is sponsored}

THROUGH THE VALLEY

1/ The US Military just launched a custom Gemini chatbot

The Pentagon is bringing generative AI to the workforce with a new platform built on Google’s Gemini. It’s part of a $200 million deal to help military staff with unclassified tasks like research, formatting documents, and analyzing data. The tool runs on a secure system called GenAI.mil, and importantly, the military says its usage won’t be used to train Google’s public models. With over 3 million personnel getting access, this represents one of the largest enterprise deployments of Gemini so far.

2/ OpenAI is prepping new image models to fix past issues

OpenAI is getting ready to release its next generation of image models alongside the GPT-5.2 update. The new models, likely called Image-2 and Image-2-mini, have been spotted in benchmarks under the codenames Chestnut and Hazelnut. The big news here is better detail and color, and specifically a fix for the yellow tint that annoyed a lot of users in the previous version. Early tests show they are closing the quality gap with Google’s Nano Banana 2. While there’s no official date yet, code inside ChatGPT suggests the launch is coordinated and coming soon.

3/ New clues suggest GPT-5.2 is around the corner

Rumors about OpenAI’s GPT-5.2 are heating up, with speculation bouncing between a December 9 and December 11 release. While nothing is official yet, the strongest evidence was found inside Notion. A new model labeled "Olive Oil Cake" appeared there with an OpenAI logo and a unique ID. Since Notion often tests unreleased features early, this suggests the model is already in trials. If it drops this week, we might actually see it surface in partner apps like Notion or Microsoft Copilot before it hits ChatGPT directly.

4/ Google and OpenAI are fighting off ad rumors

Both companies are under pressure to make money from their expensive models, but they are denying recent reports about ads. AdWeek claimed Google would put ads in the Gemini chatbot by 2026, but a Google VP quickly shot that down, stating there are no such plans. OpenAI had a similar scare when users saw what looked like chat-based ads for Target and Peloton. OpenAI removed the messages and called it a mistake, insisting they aren't testing ads. Still, with data center costs skyrocketing, everyone is watching to see how long they can hold off on monetization.

4/ Meta is reportedly ditching open source for its next big model

Meta started 2025 betting big on open source, but it looks like that strategy is shifting. After facing internal setbacks and lower-than-expected interest from developers, the company is pivoting to a proprietary model codenamed "Avocado," expected in early 2026. Mark Zuckerberg is aggressively hiring top talent to right the ship, bringing in leaders from Scale AI and GitHub. The pressure is on because competitors like OpenAI and Google are racing ahead, while Meta’s own products haven't quite gained the same momentum.

TRENDING TOOLS

• Now on Public > Turn any idea into an investable index with AI. Generated Assets are like your very own ETF with infinite possibilities *

• Open Scouts > Open-source web monitoring that alerts you the moment anything changes, from ticket drops to restocks to news updates

• Normal > A dedicated branding partner built to elevate your team’s identity and design

• Devstral 2 > An agentic coding system that explores codebases, edits multiple files, and builds engineering agents. It scores 72.2 percent on SWE-bench Verified

• Relace Search > Finds the exact files and code you need up to four times faster than Claude 4.5 Sonnet, using parallel search with multi-step reasoning across four to twelve operations

• Z.ai GLM 4.6V > A new vision-language model available in 106 billion parameters for maximum power and 9 billion parameters for lightweight local use

_{(*) signifies sponsored tool}

THINK PIECES / BRAIN BOOST

• Horses: Horses took two decades to phase out - AI probably won't give us that long

• Alignment is a capability

• I did not expect that the PowerPoint killer would be something called Nano Banana Pro, but that is where its heading

• Harvard and Perplexity researchers conducted the first large‑scale field study of how people use AI agents

• ChatGPT said to near 900M weekly users as Google’s Gemini gains ground

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.