Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
AI Coming for Accountants, Lawyers, and Marketers in 18 Months, According to Microsoft CEO
TLDR: Microsoft AI CEO Mustafa Suleyman says AI will handle most white-collar tasks within 18 months, the latest in a string of bold productivity predictions from tech leaders.
The Story: Microsoft AI CEO Mustafa Suleyman told the Financial Times that AI will reach "human-level performance on most, if not all professional tasks" within 12 to 18 months. He named accounting, legal work, marketing, and project management as the jobs most ready to be automated. Anything that involves "sitting down at a computer," he said, is on the chopping block. He pointed to software engineering as the proof, where "many software engineers report that they are now using AI-assisted coding for the vast majority of their code production," a shift that happened in roughly six months. Suleyman compared the change to writing a blog post: "Creating a new model is going to be like creating a podcast or writing a blog. It is going to be possible to design an AI that suits your requirements for every institution, organization, and person on the planet." He said his own mission at Microsoft is to reach "superintelligence" and build the company's own frontier models so it depends less on OpenAI.
Its Significance: Suleyman's timeline joins a list of bold productivity predictions from tech leaders. Elon Musk has said that within 10 to 20 years, work will be optional and people won't need a job at all because AI and robots will do everything, paving the way for what he calls "universal high income." Anthropic's Dario Amodei said last year AI could wipe out half of entry-level white-collar jobs in one to five years. Real-world data is more measured. A Yale Budget Lab report from October 2025 found the labor market hasn't shifted much since ChatGPT launched, and a Thomson Reuters study found productivity gains for lawyers and accountants using AI so far have been small. Entry-level hiring in tech, law, and finance has dropped sharply though, so if you work at a desk, learning to use AI well is the most useful thing you can do this year, whether or not the 18-month clock turns out to be right or if CEO’s of major companies are just excellent at selling their products.
QUICK TAKES
The story: OpenAI signed a deal with the Maltese government to give every resident and citizen a free one-year ChatGPT Plus subscription after they finish a short "AI for Everyone" course built with the University of Malta. The program starts in May and also covers Maltese citizens living abroad.
Your takeaway: This is the first national-scale ChatGPT Plus rollout anywhere in the world. If it works, expect more small countries to cut similar deals, which means more people getting paid AI tools through their government before their workplace catches up.
The story: Elon Musk posted that xAI's internal Grok V9 model has 1.5 trillion parameters, runs on Nvidia's Blackwell chips, and "performed extremely well" in its latest training run. The public version of Grok is still on the older V8 base, which has 0.5 trillion parameters.
Your takeaway: Tripling the parameter count usually means sharper reasoning, better coding, and stronger long-context handling. If V9 lives up to the internal tests, regular users could get a Grok that writes cleaner code, handles much longer documents in one go, and competes more directly with GPT-5.4 and Claude Opus 4.7 on real work tasks.
The story: Both companies have filed plans to launch giant satellite fleets that act as data centers powered by sunlight. Blue Origin's Project Sunrise targets up to 51,600 satellites, and SpaceX has floated a plan involving up to one million satellites delivering 100 gigawatts of compute. SpaceX's own paperwork flags "significant technical complexity and unproven technologies."
Your takeaway: Earth is running out of cheap power and land for AI data centers, so the pitch is to move the problem to orbit. The catch is radiation, latency, debris, and the small issue of nobody being able to walk over and swap a broken server. This is a 10-year bet, not a near-term fix.
TOOLS ON OUR RADAR
🛒 Flathub Free and Open Source: A massive universal application store that lets you easily discover and install thousands of completely free software programs securely.
📥 Transmission Free and Open Source: A beautifully simple and lightweight file sharing program that lets you download large media files quickly without slowing down your computer.
💬 Signal Free and Open Source: A highly secure messaging application trusted by privacy experts worldwide that protects your private text messages and voice calls using advanced encryption without tracking user data.
🕹️ PCSX2 Free and Open Source: A phenomenal software project that preserves gaming history by allowing you to play thousands of classic era video games perfectly on your computer screen.
TRENDING
Dentists Are Using AI to Push Patients Into Unnecessary Procedures - Former WSJ columnist Joanna Stern's new book documents how dental offices are using tools like Pearl AI and Overjet to flag "37 percent more disease" on X-rays, then upsell patients on costly periodontal treatments and implants they may not need.
Amazon Workers Are Faking AI Use to Hit Quotas - Amazon set a target for 80% of developers to use AI weekly and tracks "token consumption" on internal leaderboards. Employees are now using its in-house tool MeshClaw to do personal tasks, mock their managers, and pump up their numbers, a practice called "tokenmaxxing."
Culinary Students Say They're Not Worried About AI Replacing Them - Cooking schools like Johnson & Wales and École Ducasse are folding AI into their programs for menu planning, inventory, and trend forecasting. Students and chefs argue that hospitality, intuition, and human connection aren't replaceable, even if a robot can flip 300 burgers an hour.
Nvidia's $2B Nebius Bet Tightens Its Grip on AI Cloud - Nvidia took an 8.3% stake in Nebius and gave it early access to next-gen Blackwell chips. Nebius plans to deploy more than 5 gigawatts of capacity by 2030 and already has $46 billion in contracts lined up with Microsoft and Meta.
Figure AI's Humanoid Robots Sorted 88,000 Packages in 72 Hours Without a Human Touch - Figure AI livestreamed its F.03 robots sorting packages at roughly 3 seconds each, running on its onboard Helix-02 AI with no remote operators. The company says it'll keep the stream running until a robot physically breaks down. Skeptics note the package placement isn't always precise enough for real warehouse work.
Meta's Yann LeCun Says LLMs Won't Get Us to AGI - LeCun, who just launched AMI Labs in Paris, argues that scaling chatbots is a "dead end" for human-level AI. He says real progress needs "world models" that learn from vision and touch the way a four-year-old does, not just text. A house cat, he points out, understands gravity better than the biggest LLM.
Jane Street Just Spent $7 Billion on AI Compute - The quant trading firm signed a $6 billion deal with CoreWeave for Nvidia Vera Rubin GPUs and took a $1 billion equity stake in the company. Jane Street pulled in $20.5 billion in net trading revenue last year and now sits among CoreWeave's five largest shareholders, putting it in the same compute league as OpenAI and Meta.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
😨 Describe what you're dreading about tomorrow. Find the real fear underneath and one move to make.
Build a single-file HTML app with vanilla HTML/CSS/JS and one Claude API call. Sunday Dread Decoder — surface dread to real fear to reframe to opening move. localStorage key 'sunday_dread_v1'.
Aesthetic: late-night gradient (#1a1828 to #0f0e1a), drifting cloud radials, tiny stars, Fraunces serif italic for body, Inter for UI, JetBrains Mono for labels. Pink/rose for the real fear card (left red border), violet for reframe, blue for the opening move (left blue border).
Form: dread textarea, duration dropdown, intensity dropdown.
API returns raw JSON: surface (cleaned restate), real_fear (specific not generic), reframe (honest not toxic-positive), opening_move (one concrete action with specifics), worst_case (honest + why they'd survive), tomorrow_night (one-sentence preview).
Render: stacked cards in order — surface bar, real fear (pink), reframe (violet), opening move (blue, italic serif), worst case (faded), tomorrow night (centered, italic). Save to localStorage with date and dread snippet. History below.What this does: Type what's making you dread tomorrow. It names the real fear underneath, reframes it without toxic positivity, gives you one concrete opening move, walks through the honest worst case and why you'd survive it, and shows what tomorrow night could feel like if you actually make the move. Saves to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Sort 88,000 warehouse packages in 72 hours without a single human touch (Figure AI)
❌ Still Can't: Train models without buying billions in GPU compute, as Jane Street's $7 billion CoreWeave deal shows
✅ AI Can Now: Be rolled out as a national utility, with every Maltese citizen getting free ChatGPT Plus after a short course
❌ Still Can't: Be trusted not to upsell you on dental work you don't actually need
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
A novel about a near-future where genetic engineering produces a class of humans who don't need to sleep, gaining a massive cognitive advantage that the rest of society can't compete with. Started life as a Hugo and Nebula winning novella in 1991, expanded into one of the most prescient books ever written about cognitive enhancement and the fairness questions raised by AI augmentation.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
