- The Intelligent Worker
- Posts
- Gemini AI Beats Pokémon
Gemini AI Beats Pokémon
📘 Playbook on using AI as your personal assistant + 🧠 Microsoft's small models
Hi everyone,
Google's Gemini 2.5 Pro AI has conquered the game Pokémon Blue, sparking conversations about AI's potential.
Microsoft's latest Phi models challenge what we know about small language models by taking their reasoning abilities to the next level.
Also, check out this guide on using AI as your personal assistant!
Let's get right into it.
In this issue:
🤝 In Partnership: The AI guide that streamline success
🤿Deep Dive: Gemini 2.5 Pro conquers Pokémon Blue
🖼️AI Art: Examples of great and trending AI art
🤿Deep Dive: Microsoft's reasoning Phi models
⚒️Tool Snapshots: Tools for AI, no-code, and productivity
📰Top News: News on AI, no-code, and productivity
🤝IN PARTNERSHIP WITH HUBSPOT
Ready to save precious time and let AI do the heavy lifting?
Save time and simplify your unique workflow with HubSpot’s highly anticipated AI Playbook—your guide to smarter processes and effortless productivity.
🤿 DEEP DIVE
Google's Gemini AI Model Claims Victory in Pokémon Challenge Amid Controversy
Intelligence: Google CEO Sundar Pichai announced that their advanced AI model, Gemini 2.5 Pro, has successfully completed the classic video game Pokémon Blue, a claim that has stirred both excitement and debate about the methods used to achieve this milestone.

Gemini 2.5 Pro, developed by Google, has completed the game Pokémon Blue, with the achievement being livestreamed by an independent software engineer named Joel Z, who is not affiliated with Google.
Pichai celebrated the completion on social media, referencing the fun competition among AI models, particularly in the context of previous advancements made by Anthropic’s Claude AI.
Logan Kilpatrick, product lead for Google AI Studio, highlighted Gemini's progress, noting it has claimed its fifth badge in the game, outpacing competitors like Claude.
Despite the achievements, Joel Z cautioned against making direct comparisons between AI models, as they utilize different methodologies and tools for gameplay.
Joel Z revealed that both AI models rely on "agent harnesses" that assist them by providing necessary gameplay context and guiding their decision-making, calling this support a means to improve the AI's reasoning rather than cheating.
🖼️ AI ART
Examples of great and trending AI art
Even heroes need to celebrate. May the fourth be with you!
🤿 DEEP DIVE
Microsoft Unveils Next-Generation Phi Small Language Models Championing Reasoning Abilities
Intelligence: Microsoft has launched three new small language models—Phi-4-reasoning, Phi-4-reasoning-plus, and Phi-4-mini-reasoning—signifying a breakthrough in AI capabilities focused on complex reasoning tasks while remaining efficient and accessible.

Graph source: Microsoft
The new Phi models represent an advanced category of small language models (SLMs), capable of performing intricate reasoning tasks that typically require larger models.
Phi-4-reasoning, featuring 14 billion parameters, competes effectively with bigger models in executing complex reasoning tasks, leveraging sophisticated training techniques like supervised fine-tuning and the generation of high-quality datasets.
Phi-4-reasoning-plus enhances its predecessor with reinforcement learning, utilizing 1.5 times more tokens, which boosts its accuracy even further.
These models have outperformed competitors such as OpenAI's o1-mini and DeepSeek-R1-Distill-Llama-70B on benchmarks, including challenging mathematical problems and science questions required for Ph.D. qualifications.
⚒️ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🧠 RightNow AI 'V2.0' - Enhancing CUDA kernels for optimal performance. Free option available.
📽️ Airpost - Transform product links into engaging video ads instantly. Free to try.
💬 Kotae - 24/7 website lead generation and instant customer support. Free to try.
🕸️ Simular - AI companion making your internet browsing smarter and faster. Free to try.
🧪 Velozity - Swift AI-powered user research and testing tool. Free to try.
⚙️ Synapis - Tailor-made AI models for automation and business scaling. Free to try.
📰 TOP NEWS
News on AI, no-code, automation, and productivity
OpenAI will roll out an opt-in testing phase, clearer update notes, and stricter safety reviews after a recent model update made ChatGPT overly flattering and uncritical.
Sam Altman’s World App now combines crypto payments, social networking, and mini apps, with U.S. debit card support coming soon and a focus on verified human identity through Worldcoin.
Perplexity now lets users forward suspicious messages to a dedicated number for instant fact-checking, offering quick answers with source links in over 20 languages—all without leaving the app.
Microsoft's Azure AI Foundry now offers tools to evaluate, test, and deploy over 1,900 AI models, streamlining development and helping teams manage updates, costs, and innovation in one centralized platform.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?Your feedback is more valuable to us than coffee on a Monday morning! |
What more do you want to see in this newsletter?Please vote |
