ESCOLA AI WEEKLY #30 – Krzysztof Wojewodzic

Hi! This is Krzysztof Wojewodzic from Escola AI 👋 In this issue of AI Weekly, I discuss 8 key events that demonstrate that the AI ​​arms race has entered a personal phase. We're no longer just seeing a battle over model parameters, but a real shift in how we code, invest, and referee sports events. In this episode: 8️⃣ AI at the Olympics – a referee who sees more Artificial intelligence helps referees where the human eye fails (e.g., technique in race walking), but the real "meat" is managing the Olympic village. Optimizing logistics for millions of people and capturing key broadcast moments by analyzing crowd reactions – these are concrete use cases for managing big data. 7️⃣ The "Shorten It to Me with AI" Trap Microsoft warns against Prompt Injection hidden in summary buttons. If your model has persistent memory, clicking on automatic summaries too often can "poison" it and cause it to uncritically recommend specific products. Convenience versus company data security. 6️⃣ Musk's Empire Enters Orbit The merger of Tesla (now a robotics company), XAI, and SpaceX. Elon Musk is preparing servers in space because space and energy are running out on Earth. With a launch cost of $3,700/kg and the power of Starlink, SpaceX is becoming a giant telecommunications company that could consume a market worth trillions of dollars. 5️⃣ Battle of the Billions: Anthropic vs. OpenAI Another round of funding for Cloud Code ($30 billion) and OpenAI. However, the most interesting thing isn't the amounts involved, but... the lack of a handshake between Altmann and Amodei at the summit in India. This personal animosity between the leaders shows how tense and brutal the fight for dominance in the world of LLM is. 4️⃣ Alphabet – a company for 100 years? Google issued 100-year bonds with a massive oversubscription. Investors believe that Alphabet will still be a powerhouse paying interest in a century. This is an important signal for those predicting a quick burst of the AI ​​bubble – long-term capital is betting on a completely different scenario. 3️⃣ OpenClaw and the agentization of Mac Minis Tools allowing agents to "click" on computers for us are becoming a reality. The result? A boom in sales of Mac Minis as cheap, isolated servers for agent work. We are moving from the era of APIs to the era of "agent to interface" – until systems start talking directly, AI will click for us like a monkey, but faster. 2️⃣ Gemini 3.1: A model that knows where you are The new Gemini was the only one that didn't fall for the "car wash test" (logical thinking about a car). But the real breakthrough is multimodality – the model read my GPS, found the area in Google Maps, and generated a photorealistic sign for the gallery I was visiting. Integrating Google data with image generation is starting to crush the competition. 1️⃣ Claude 4.6 and the end of the era of "coding" This isn't just an update, it's a paradigm shift. Top developers declare: "In 2026, we're no longer coding; we're talking to the model." The ability to rewrite half a system in a minute because the model has no emotional attachment to the code radically changes the role of the senior developer. What matters now is what you ask and how you design the architecture, not how efficiently you write tests. In AI, information selection is now the most important competency for a manager. Let us know in the comments whether the vision of a developer who only "talks to the code" is an opportunity or a threat to system security for you? 🎧 You can also watch and listen to AI Weekly on Spotify. Just type Escola AI. Thanks for coming! Krzysiek #AI #ArtificialIntelligence #AIWeekly #EscolaAI #Gemini #Claude #OpenAI #SpaceX #Coding #Business #TechTrends