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.
*Over the next few days I'm using the time to migrate to a new hosting platform in an effort to deliver a better experience for my readers. If you notice any delay or hiccup receiving an issue, I apologize in advance, and I thank you for reading along with me. If an issue doesn't show up at its usual time, check your spam or promotions folder. Replying to the email, or marking it "not spam" or "important," helps your inbox provider learn to deliver the email straight to your primary folder going forward.
THE FRONT PAGE
Perplexity's Co-Founder Says 'Safety' Is Really About Control
TLDR: A Perplexity co-founder says the push for "AI safety" is really a way for a few big companies to control who gets to build powerful AI, and he wants an open lab where any top researcher can work.
The Story:
Andy Konwinski, who helped start Databricks and Perplexity AI, published an essay this week saying that keeping AI locked inside a few private labs is its own kind of danger. He pointed to Anthropic as his main example. When Anthropic launched its Claude Fable 5 model on June 9, a note buried in a 319-page document said the model would give worse answers to anyone it thought was training a rival AI. People found it and pushed back hard, and Anthropic reversed the rule within 48 hours. Konwinski's point wasn't that Anthropic made a bad call. It was that one company got to make that call for everyone. His fix is a shared research center with enough computing power that top researchers can reach the most advanced models without asking a private company for permission first.
Its Significance:
AI is turning into basic infrastructure, like electricity or the internet, and whoever controls it ends up shaping how the rest of us live and work. Right now a small group of firms holds the keys. A UC Berkeley dean said her researchers are building on Chinese models because there's no open Western one to use. Yann LeCun, Meta's former chief scientist, backed the argument and compared closed AI to an empire that banned the printing press for 200 years. Not everyone agrees, and plenty of people argue that wider access could make misuse easier.
QUICK TAKES
The story: Tilly Norwood, the AI-made "actor" that angered Hollywood last year, will lead her first full movie, a comedy-drama called "Misaligned" from the studio Particle6. The film follows an AI character with no body and no real memories, only access to everyone else's.
Your takeaway: This is the first feature film built around an AI lead. It arrives while actors' unions are still fighting over consent and whether AI performers were trained on real actors' work without pay. This seems more like publicity stunt than anything else, like announcing an all CGI lead.
The story: Microsoft cut about 4,800 jobs, roughly 2% of its staff, with a heavy hit to its Xbox gaming group where sales have been shrinking. Four game studios will be spun off while the company pours money into AI.
Your takeaway: Microsoft has been the worst-performing big tech stock this year, partly because investors worry its own AI products haven't caught on yet. Even giants are reshuffling people and budgets around AI bets.
The story: A new index from MIT Sloan lecturer Paul Cheek ranks S&P 500 companies by how much they actually use AI, not just how much they talk about it. It pulls from earnings calls, job postings, and patents instead of surveys.
Your takeaway: Nearly every big company brags about AI on earnings calls, but this scorecard tries to separate real work from hype. Tech firms lead. Real estate sits near the bottom.
TOOLS ON OUR RADAR
🐧 Kavita Free and Open Source: A blazing fast, self hosted digital library that acts as a personal streaming service for your ebooks, comics, and PDFs, allowing you to read your collection on any device without relying on corporate cloud servers. (Alternative to Amazon Kindle Cloud)
🎧 Wondercraft Freemium: An AI podcast builder that transforms your written newsletters, blog posts, or text documents into studio-quality audio shows complete with multiple hosts and background music.
🗣️ Langua Freemium: A conversational AI language tutor that lets you practice speaking in real time with interactive roleplays while correcting your grammar and pronunciation instantly.
🍳 Prospre Freemium: An AI meal planner that generates weekly grocery lists and customized recipes based on your exact macronutrient goals and personal dietary restrictions.
TRENDING
Is AI Ready to Take Over Your Prescriptions? - Utah quietly let an AI chatbot named Doctronic refill about 190 common prescriptions on its own. Doctors, lawyers, and the FDA are split on whether that's safe.
What Is Mistral AI? Everything to Know About the OpenAI Competitor - France's Mistral is the closest thing Europe has to OpenAI, now worth around $23 billion and past $400 million in yearly revenue, with a new open-weight model due this summer.
Insert Token to Continue, Says AI. Yeah, About That... - AI firms are racing to make models use fewer "tokens" to cut costs. One new tool called Caveman even strips replies down to caveman-speak to save money.
Anthropic's CEO Tempers His Big Predictions on AI and Biology - Dario Amodei is pulling back on his own bold claims about AI curing disease fast, admitting biology is messy enough that real progress may take years.
AI-Enabled Enforcement Tech Takes Time and Testing - New Cornell research on baseball's automated strike-zone system shows even a "simple" AI rule-checker took seven years to earn trust, a warning for AI used in courts, policing, and work.
Brown Built an AI That Predicts How Fast a Drug Patch Releases Medicine - Brown researchers made an AI that predicts how quickly patches, bandages, and implants release their medicine, cutting lab testing time by up to 94% by baking the physics of how molecules spread right into the model.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
◫ Give it today's tasks and how your energy actually moves. Get an hour-by-hour plan that puts the hard work where your brain works.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Focus Block Architect — enter today's tasks and your energy pattern, get an hour-by-hour plan matched to when your brain works. Persist to localStorage key 'focus_block_architect_v1'.
Aesthetic: dark green-black (#0b0f0e), mint (#5adcb4) primary with top radial glow. Space Grotesk sans for headings, Newsreader serif italic for strategy/reasons, JetBrains Mono for labels/times. Four block types color-coded: deep=mint #5adcb4, shallow=blue #78aaf0, admin=gold #c8b478, break=purple #b48cdc.
Form: today's-tasks textarea (note deadlines/fixed times), energy-peak dropdown (early bird / mid-morning / afternoon / night owl / unpredictable), workday-window text input, focus-stamina dropdown (short 25-45m / solid ~90m / long hours), interruption dropdown (protected / some pings / constant), #1-priority text input.
System instructions: focus strategist designing realistic hour-by-hour plans matched to energy; protect peak hours for hardest deep work, batch shallow/admin into low-energy windows, honor fixed commitments, build in breaks and buffer, never overpack (leave slack); classify each block deep/shallow/admin/break; respect stamina (block length) and interruptions; put #1 priority in best window; realistic beats aspirational. Return raw JSON: strategy (2-3 sentences naming the key move + tradeoff), blocks[]{time, task (from their list), type, why (tied to energy)}, guardrails (3-4 specific protective rules), if_sideways (2-3 sentences, **bold** the non-negotiable block). 6-10 blocks spanning the workday.
Render: gradient "strategy for today" card (italic serif). A timeline with a time gutter (mono) beside color-coded, left-border block cards showing task + a type chip + italic why, plus a color legend. Mint "guardrails" card with ▸ items. Gold "if the day goes sideways" card with bolded non-negotiable. Copy plan + archive keyed by priority/date.What this does: Dump today's task list, note when your brain peaks, your workday window, focus stamina, interruption level, and the one thing that matters most. It builds a realistic hour-by-hour plan that protects your peak hours for deep work, batches shallow and admin tasks into low-energy windows, honors your fixed commitments, and leaves real slack instead of overpacking. Every block is color-coded by type (deep / shallow / admin / break) with a short reason it's placed where it is, followed by a few guardrails to make the plan hold and a "if the day goes sideways" fallback that names the one block worth defending. Saves plans to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Refill around 190 common prescriptions on its own in Utah, no doctor visit needed.
❌ Still Can't: Call balls and strikes without seven years of testing first to earn people's trust.
✅ AI Can Now: Rank which big companies actually use AI, based on real data instead of self-reported surveys.
❌ Still Can't: Deliver on the biggest biology promises quickly, even by its own makers' admission.
RECOMMENDED LISTENING/READING/WATCHING
Filterworld by Kyle Chayka - Book
A 2024 book by a New Yorker staff writer arguing that algorithmic recommendation, the same underlying tech behind TikTok, Spotify, and Netflix, has flattened global culture into a single beige aesthetic where every coffee shop, every apartment, every playlist and every Instagram feed looks and sounds the same. Chayka is one of the sharpest cultural writers alive on the effects of AI, and this book absolutely nails why the world feels weirdly homogeneous now. Under-cited compared to its scope.
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.
