AI in a Nutshell - Week 46 - GPT-5.1 Arrives, Claude Outsmarts Hackers, and Cursor raised $2.3B
OpenAI releases GPT-5.1 with “Instant” and “Thinking” modes, Claude helps stop the first AI-powered cyberattack, Microsoft launches an AI superfactory, and Cursor raised $2.3B.
Fellow human, it’s week 46. AI hasn’t taken over the world yet. And here’s your lazy-man version of what went down in the AI world in the past week.
Summary for extra lazy people: OpenAI releases GPT-5.1 with “Instant” and “Thinking” modes, Claude helps stop the first AI-powered cyberattack, Microsoft launches an AI superfactory, and Cursor raised $2.3B.
What You Must Know
OpenAI Released GPT-5.1 → GPT-5.1 brings two new modes, Instant for fast replies and Thinking for deeper reasoning. I haven’t tested it yet, I think I’m beginning to develop model fatigue, there’s so much “now even smarter!” I can take in a week, but I’ll probably try it once my curiosity beats my exhaustion.
Chinese Hackers Used Claude Code for an AI Cyberattack → Anthropic caught the first known AI-orchestrated cyber-espionage campaign, allegedly run by Chinese hackers using Claude Code. It’s a chilling preview of what AI can do when misused. On the bright side, Anthropic caught it fast, so at least someone’s paying attention. I’ll read more on it and write a summarised piece later this week.
Cursor Raises $2.3B to Build Its Own AI Model → The coding assistant startup Cursor raised $2.3B at a $29.3B valuation to build its own foundation model for developers. How better can they make an already great software? The future looks interesting.
Anthropic Shares Its Bias-Busting Toolkit → Anthropic has open-sourced a framework to measure the political bias of models. This will be useful for people who argue with ChatGPT about politics 😅
Baidu Unveils ERNIE 5.0 and AI Agents → China’s Baidu dropped ERNIE 5.0, a multimodal model that can process text, voice, and video, plus a suite of AI agents built for offices, classrooms, and factories.
What’s Good to Know
ChatGPT Group Chats Now In Asia-Pacific → OpenAI is testing group chats in Japan, New Zealand, South Korea, and Taiwan. You can now rope your friends or coworkers into one conversation with ChatGPT. Not sure how that’ll work, but it sounds interesting.
Wall Street Questions the AI Hardware Gold Rush → As tech lords race toward $1 trillion in AI infrastructure spending, investors are asking: how long before all these shiny GPUs become junk? Could the AI hardware boom be the next crypto mining rig story? I doubt.
OpenAI and Microsoft Partner with U.S. Attorneys General on AI Child Safety → The two companies are teaming up with U.S. state AGs to tackle AI-related child safety issues, like deepfakes and grooming risks.
NotebookLM Now Has ‘Deep Research’ Mode → Google’s NotebookLM can now handle more file types and dig deeper into sources, basically giving users a mini research assistant.
”Writing Might Be AI’s Worst Use Case” - Sitepoint → SitePoint argues that using AI for writing is counter-productive: you spend more time prompting and rewriting than drafting, and lose the ability to express your own thoughts clearly. I kinda agree, in my opinion, the best use case for AI in writing is proofreading.
Money Moves This Week
Sim.ai raised $7 million in Series A funding to expand its AI agent platform.
Dialog raised €3.74 million in seed funding to build AI shopping agents for ecoms.
Netic raised $23 million in Series B funding to scale its AI-driven infrastructure.
Obello raised $8.5 million in seed funding to enhance its AI creative software.
Lative raised $7.5 million in seed funding to develop smarter AI sales planning.
Kaizen Corner - What the Heck Is “AI Inference”?
I wrote about this last week, but here’s the short, brain-friendly version.
“Inference” is basically the moment when an AI uses what it has already learned to give you an answer.
Think of it like this: training is when the AI studies, and inference is when it finally writes the exam.
So when you type a prompt into ChatGPT, it’s not “learning” anything new. It’s just looking at what it already knows, doing some abracadadra math, and spitting out its best guess.
If you want a longer breakdown with an analogy, read it here.
Meme of the Week
That’s your week in AI.
If you learned something, tell a friend. And if you didn’t, blame yourself :)
Until next Sunday,
Kay - your fellow human.
P.S. If this email lands in spam, that’s your inbox trying to stop you from staying plugged in. Fix it.