Strap in folks,
Google released Gemini 3. I’ve spent the last 24 hours testing the model & distilling all the release notes and videos.
Here is EVERYTHING you need to know!
📌 TL;DR
Gemini 3 Pro → Google's new model takes the lead (by a mile) in reasoning and visual understanding.
ChatGPT 5.1 → OpenAI's latest update brings faster responses, better reasoning, and improved coding. Solid all-rounder.
ElevenLabs expands → Now does images and video alongside voice. One platform for audio, visual, and multilingual content creation.
Workflow of the Week → Create branded slide decks with Claude in minutes.
Prompt of the Week → Turn your crappy phone selfie into a professional studio headshot using AI.
One small step for Google, one giant leap for AI
Google released Gemini 3 Pro today, and benchmarks show the model ahead in nearly every category, by a huge margin.
Before I get into the specifics, some context: AI has been good enough to write emails, draft social posts, or create ad copy for a while now. We're well past that point. The models aren't getting dramatically better at those basic tasks anymore.
The big leap forward with Gemini 3 is in reasoning, multimodal understanding, and what it can do as an agent.
It's much smarter at hard problems
Gemini 3 Pro scored 37.5% on Humanity's Last Exam (a test designed to stump AI with PhD-level reasoning). The closest competitor, GPT-5.1, hit 26.5%.
On the AIME 2025 math test, Gemini 3 Pro scored a perfect 100% when allowed to use code. For research, complex analysis, or anything needing deep reasoning, this is the model to use now.
It's an LLM with "eyes"
Gemini 3 Pro scored 72.7% on ScreenSpot-Pro, a test that measures how well AI understands what's on a computer screen. Claude Sonnet 4.5 scored 36.2%. GPT-5.1 scored 3.5%.
You can upload screenshots, videos, or screen recordings and ask Gemini questions about them. It'll actually understand what it's looking at and give you useful answers.
This matters for anyone building agents or automations that interact with websites, apps, or visual interfaces.
It's a better agent (and businessman)
On the Vending-Bench 2 test, the AI runs a simulated vending machine business. It has to manage inventory, set prices, order products, and pay expenses over time.
Gemini 3 Pro made $5,478. The closest competitor, Claude Sonnet 4.5, made $3,838.
This shows Gemini is much better at handling long, multi-step projects without losing focus.
Alongside Gemini 3, Google also released Antigravity.
It's an AI-powered coding platform that lets developers build, test, and debug software using autonomous agents that work across your editor, terminal, and browser at the same time.
How to actually think about this
Stop asking "which model is best?" Start asking "best for which workflow?"
Here's a general rule of thumb:
See / Do tasks: Gemini 3
Anything involving screens, video, dashboards, long messy context, or visual analysis.
Write / Talk tasks: Claude / ChatGPT
Persuasive writing, emails, landing pages, brand voice, everyday chat.
What Gemini 3 unlocks for solo businesses
Before Gemini 3, AI mostly worked on text. Now it can actually see. That opens up workflows that were previously impossible:
For marketers and content creators:
"Here are 50 TikToks / Reels. What patterns do you see in the winners vs losers?"
"Find the best hooks in this 40-minute recording and suggest 5 short-form clips."
"Compare high CTR vs low CTR ads. What's visually different?"
For sales and client services:
"Summarize this discovery call and turn it into a 1-page brief."
"List objections, desired outcomes, and language the prospect used."
For solo founders:
"Here's my Notion doc, screenshots of dashboards, and slides. Where do the numbers and narrative disagree?"
"Given this deck + KPI table + MRR chart, what 3 priorities would you set for next quarter?"
Where it fits in your stack
If you just open chat, write a few emails, and ask it some questions, you'll probably think "this isn't that different." The real unlocks show up when you feed it screens, videos, dashboards, PDFs, and mixed media, then ask it to speed up work that used to be off-limits: creative analysis, UI audits, call reviews, long multi-source synthesis.
For writing, I still lean on Claude (but writing is subjective), I suggest you test a few different models and see which outputs you prefer - I’ve heard amazing things about Grok 4.1 and GPT 5.1 thinking.
But for research, problem-solving, visual analysis, or anything involving complex, multi-step tasks, Gemini 3 Pro is the clear leader now.
Gemini 3 Pro is live now in the Gemini app, Google AI Studio, and rolling out across Google products.
GPT-5.1 is here to clean up GPT-5’s mess
GPT 5.1 released without a fuss last week. When GPT-5 first dropped, the response was rough. The tone felt colder, and most people thought it was worse than GPT-4.
The core issue was that GPT-5 followed instructions much more literally (good thing). This is great for advanced users and agent builders with detailed prompts, but everyday users who didn't know how to prompt well got drastically worse results than GPT-4.
GPT-5.1 is OpenAI fixing common complaints. It's faster, more accurate, more conversational, and warmer to talk to. The context window expanded (longer prompts, bigger docs, extended chats), and thinking time improved. Coding performance also got better. Not a massive leap, but a solid cleanup that makes the model feel just that little bit better.
GPT 5.1 is now the default model for all ChatGPT users.
ElevenLabs just became an all-in-one creative suite
ElevenLabs released Image & Video (Beta) today. You can now create images and videos inside ElevenLabs, then add AI voices, music, and sound effects in the same place.
They've integrated top video models like Veo, Sora, Kling, Wan, and Seedance. This is aimed at creators, marketers, and content teams who want to go from idea to final export in one workflow.
Think; product videos, social content, educational materials, etc all doable in a single platform now.
🧰 Create Branded Slide Decks in 2 minutes with Claude (Workflow of the Week)
Pitch decks, client presentations, internal reports. You need them constantly, and they eat hours.
Claude can now build you a full slide deck in minutes, complete with your brand colors, fonts, and messaging, then export it as a .pptx file you can edit in Canva, PowerPoint or Google Slides. Saves 3+ hours every time.
💡 Prompt of the week
Drop this prompt into Google Gemini and select Tools > Create image (which uses their Nano Banana model) to transform any mediocre headshot into a professional studio shot. Perfect for YouTube thumbnails, social media profile pics, etc.
“Create a professional studio-quality headshot based on the provided image. Keep the subject’s pose, facial expression, and overall likeness consistent with the original photo. Apply clean, flattering studio lighting with soft key light and gentle fill light for even skin tones and natural highlights. Use a crisp pure white backdrop with no texture. Enhance clarity, sharpness, and detail while maintaining a natural look—smooth skin slightly, remove blemishes, adjust color balance, and add subtle catchlights in the eyes. Produce a polished, high-resolution portrait suitable for a YouTube thumbnail, with strong contrast and a premium, modern aesthetic.”
💌 Earn free gifts
I’m giving away my personal Claude skills library (playbooks for AI).
Just share this newsletter with 2 friends (who are into AI), and it’s all yours! You currently have {{rp_num_referrals}} referrals, only {{rp_num_referrals_until_next_milestone}} away.
or send them this link: {{rp_refer_url}}
Alright, that's a wrap for this week.
I went solo camping over the weekend. Climbed this massive mountain and caught the sunrise at the top. Absolutely stunning.
The deeper I get into AI and tech, the more my soul craves nature. Finding a healthy balance between the two has been good for me.
Also finished up my dopamine detox this week and had a triple shot coffee after 3 weeks of no caffeine. Sweaty, jittery mess all day. Not my finest moment.
Cheerio chaps!
