😸 Gemini 3 just dethroned ChatGPT

PLUS: GPT 5.1-Codex-MAX Fox + Palantir = AI newsroom. Klarna cuts staff in half.

Welcome, humans.

Well, you could’ve said this was a bad week to work at OpenAI… until today. Let’s rewind: this week alone, Google launched Gemini 3, and Microsoft dropped $5B on Anthropic to bring Claude to Azure, making it clear that no one's sitting idle while AI models race toward AGI (when AI can do pretty much everything a human can).

We’ll recap Gemini 3 a bit more below, but the TL;Dr is that Gemini 3's Achilles' heel seems to be coding (despite crushing most other benchmarks), while Claude has dominated coding for months (if not all year, depending on who you ask).

And coding isn't just another capability… it's the capability that matters most right now. Why? Because virtually all AI progress flows downstream from software, and software developers are by far the biggest token spenders, whether they're building agents or building with agents.

So OpenAI wasn't about to cede that turf. Enter GPT-5.1-Codex-Max, their new frontier agentic coding model that's laser-focused on taking back the crown. It's faster, more capable, and ~30% more token-efficient than previous versions (more details in the system card).

It's the first OpenAI model with built-in compaction. That means it can work across multiple context windows for hours on project-scale tasks, automatically pruning session history to keep only what matters. Translation? Your coding agent can now persist on complex projects for 24+ hours without losing its train of thought.

Plus, it achieved top scores (77.9%) on SWE-Bench Verified, the gold standard for measuring AI coding chops (ahead of Anthropic AND Gemini). Oh, and it's trained to actually work in Windows environments… finally, some love for the PC crowd!

Available now in Codex for Plus, Pro, Business, and Enterprise users.

Here’s what happened in AI today:

Google released Nano Banana Pro, a new professional AI image tool.
Fox News hired Palantir to build custom AI newsroom tools.
Hugging Face CEO said we're in an “LLM bubble” that might burst next year.
Klarna cut its workforce in half using AI, then boosted remaining staff pay.

Google Just Dropped Nano Banana Pro… And You Can Use It in Adobe Apps Right Now

Google just launched Nano Banana Pro (also known as Gemini 3 Pro Image, but let’s be honest… we’re gonna call it Nano Banana lol), and if you're a creative professional, this one actually matters.

What makes Nano Banana Pro different? Three big upgrades stand out:

It actually nails text in images. Previous models butchered spelling and made fonts look like alphabet soup. Nano Banana Pro generates clean, legible text in multiple languages—perfect for posters, mockups, and designs that need words that don't embarrass you.
It can blend up to 14 images while maintaining consistency of up to 5 people. Turn sketches into photorealistic products. Combine multiple reference images into cohesive compositions. Keep your brand looking consistent across every touchpoint. This is WILD y’all.
It taps into Google Search's knowledge base for factually accurate visuals. Need an infographic about solar energy or a recipe visualization? It pulls real-world info to create context-rich images that actually get the details right.

Nano Banana accurately translating and updating text on a product image

Here’s an example. Look at these pictures of dogs. Can you tell the difference between what’s real and what’s AI? Reddit says they basically can’t anymore.

As for that classic Gemini watermark on today’s Nano Banana images, they're still keeping visible watermarks on free and Pro tier images, but removing them for Ultra subscribers and Google AI Studio users who need clean visuals for professional work.

Are you scared about what this tech will do to misinformation? So is Google. Which is why they are adding a new verification tool to the Gemini app. Upload any image and ask if it was generated by Google AI (or “Is this AI-generated”?), and Gemini will check the new embedded SynthID watermark and tell you. You can try this on the Thursday Trivia below… but make sure you vote BEFORE you get the answer!

Where you'll find it: Adobe made it available day-one in Firefly and Photoshop. In Firefly, Nano Banana Pro powers Text to Image and Firefly Boards for moodboarding. In Photoshop, it's the newest model behind Generative Fill, letting you make realistic edits that you can refine with layers, masks, and selections.

Don’t forget: Adobe’s offering unlimited image generations using Firefly and partner models through December 1 for Creative Cloud Pro and Firefly plan subscribers.

You can also access it in Gemini (obviously lol), or in Google AI Studio to experiment with the model directly (gotta connect your API key though, so this one will cost you; Firefly might be the more affordable way to try it).

Why this matters: Adobe's recent study found 60%+ of creators use multiple AI models for different tasks. Rather than chasing new releases across platforms, Adobe brings top models directly into the apps where you already work. Nano Banana Pro joins models from Black Forest Labs, ElevenLabs, Ideogram, Luma AI, OpenAI, Pika, and Runway.

And as we saw with Microsoft’s big Copilot announcements on Tuesday, Microsoft is doing the same with AI models. This is the new move: maximum value to users at app layer, with AI models used as commodities to be swapped in and out as needed.

Now remember: Nano Banana Pro is built on Gemini 3 Pro, which Google released two days ago and immediately topped the LMArena Leaderboard with a breakthrough 1501 Elo score (overall score across all domains).

Specifically, Gemini 3 crushed benchmarks across the board, including:

37.5% on Humanity's Last Exam (testing PhD-level reasoning), which is high because this one is hard.
91.9% on GPQA Diamond (graduate-level physics, chemistry, and biology questions written by domain experts).
A record 23.4% on MathArena Apex (advanced mathematical problem-solving).
1487 Elo on WebDev Arena (building functional web applications).
76.2% on SWE-bench Verified (fixing real bugs in actual codebases; just below Sonnet 4.5).

Gemini 3 also achieved SOTA on SpatialBench, a lesser discussed benchmark testing spatial reasoning in vision-language models and their ability to understand 3D relationships and positioning in images. This is good for image modeling!

What makes this relevant for creatives? Nano Banana Pro inherits Gemini 3's advanced reasoning and real-world knowledge. That's why it can generate factually accurate infographics, understand complex multi-image compositions, and translate text in images while maintaining context. The breakthrough is genuinely smarter generation backed by Google's most capable reasoning model yet.

This is also building toward’s Google’s vision of a true omni-model that combines the capabilities of all of Google’s top AI models in one. Whether that ends up being a single AI model or more likely, a single system that can mix and match different models’ features and functions, remains to be seen.

FROM OUR PARTNERS

Your AI operating system for real work

Your team wastes 10 hours a week hunting for files, copying data between apps, and building decks from scratch.

That's 520 hours a year per person. Gone.

Sana Agents handles it all—autonomously. Need pricing from your Drive, customer data from your CRM, and a pitch deck? It searches, pulls, analyzes, and builds it across 100+ apps. No coding required.

Fortune 500 teams already see the difference:

10 hours saved per employee each week 62% faster prep time 66% quicker compliance audits

While other AI tools ask you to do the work, Sana's agents actually do it for you. They plan, execute, and learn—like having an assistant who never sleeps.

Start saving 10 hours this week → You can thank us later!

Prompt Tip of the Day

Time to turn ChatGPT into your payroll assistant. Payroll managers spend hours on repetitive tasks—calculating deductions, drafting salary emails, checking compliance. But ChatGPT can handle most of it if you know how to prompt it right.

Gautam Roy just dropped a guide with 50+ prompts specifically for payroll managers, and the trick is simple: be specific. Instead of “create a salary structure,“ try: “Create a salary structure for 50 employees in India, including PF, ESI, and professional tax.“

Key prompts to save:

Salary calculations: “Calculate net pay for an employee with ₹50,000 gross, considering PF (12%), ESI (0.75%), and 2 unpaid leaves.“
Compliance checks: “Summarize latest PF, ESI, and TDS rules for FY 2024-25 with due dates.“
Employee communication: “Write a polite email explaining why net salary is lower this month due to unpaid leaves.“

Pro tip: Never paste real employee data—use placeholders like [Employee Name] or [Salary Amount]. Also, save your best prompts in a Google Doc (or better yet, as custom or project instructions project folder in GPT or Claude) for reuse.

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 600K daily readers here!

*Ideas move fast; typing slows them down. Wispr Flow flips the script by turning your speech into clean, final-draft writing across email, Slack, and docs. Give your hands a break ➜ start flowing for free today.
Antigravity is Google’s free agentic development platform powered by Gemini 3, where agents can autonomously plan and execute tasks across editor, terminal, and browser.
Aardvark is an agentic security researcher powered by GPT-5 that autonomously scans code repositories to find and patch vulnerabilities, achieving 92% detection rates in benchmark testing.
Maxima automates month-end accounting closes by co-writing financial reports with your team—reconciling data, preparing journal entries, and flagging anomalies so accountants review instead of do grunt work (raised $41M).
Agentio connects brands with YouTube creators for sponsored videos using AI to match campaigns, review content drafts, and get better ROI than traditional video ads (raised $40M).

Around the Horn

Fox News hired Palantir to build custom AI newsroom tools over the past year, creating a “digital twin“ of its workflow without giving up its IP—tools like “topic radar“ for story briefings and “text editor“ for style checking now help reporters work faster.
US President Trump warned against AI “overregulation“ on Tuesday, calling for one federal standard instead of 50 state-level regulatory regimes, saying fragmented rules would let China win the AI race.
HuggingFace CEO Clem Delangue said we're in an “LLM bubble“ that might burst next year, arguing the industry's obsessed with building one massive model when specialized, smaller models are the real future.
Klarna cut its workforce nearly in half since 2022 (from 5,527 to 2,907 employees) through attrition and AI replacements, then boosted remaining staff pay by 60%—tech now does the work of 853 full-time employees.

FROM OUR PARTNERS

Gladly’s AI ROI Calculator

How much could AI save your support team? Peak season is here. Most retail and ecommerce teams face the same problem: volume spikes, but headcount doesn't. Instead of hiring temporary staff or burning out your team, there’s a smarter move. Let AI handle the predictable stuff, like answering FAQs, routing tickets, and processing returns, so your people focus on what they do best: building loyalty.

Gladly’s ROI Calculator shows exactly what this looks like for your business: how many tickets AI could resolve, how much that costs, and what that means for your bottom line. Real numbers. Your data.

See your savings