In partnership with
Big Pharma's $240B White Flag Is One Startup's Ticket
Big Pharma spent decades and billions trying to solve osteoarthritis, a $500B market they’ve never cracked.
Thankfully, Cytonics figured out why they keep failing: joints are attacked by multiple culprits at once, and Big Pharma only ever went after one at a time.
So Cytonics discovered a way to get them all, creating the first therapy with the potential to actually address the root cause of osteoarthritis at the molecular level. It’s already proven across 10,000+ patients. Now, they’re pushing toward FDA approval on a 200% more potent version that can be manufactured at scale.
The first human safety trial is already complete with zero adverse events. If approved, the more than 500M osteoarthritis patients worldwide could have their long-needed solution.
Big Pharma created this opening. Now Cytonics is prepared to seize it.
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 Is Already in the Courtroom and Your Judge Might Not Be Writing Their Own Rulings
TLDR: More than 60% of federal judges are now using AI tools in their work, raising big questions about who is really writing the rulings that decide people's lives.
The Story:
A new survey of federal judges found that more than 60% have used at least one AI tool in their work, with legal research as the top use. Some judges use AI to summarize cases, build timelines, and even draft questions for attorneys before hearings. Texas-based federal judge Xavier Rodriguez, for example, uses AI to identify key players in a case and map out events, describing these tasks as "relatively risk free" because no final decision is being made. But not everyone agrees. A string of recent incidents showed what can go wrong: two federal judges approved AI-generated court orders that contained fake case citations, made-up quotes, and references to people who were never part of the case. Both judges blamed staff members for using AI without proper checks in place. Senate Judiciary Chairman Chuck Grassley went public with the errors and called on the entire federal judiciary to build formal rules around AI use. The Administrative Office of the U.S. Courts is now reviewing how AI is being used across courtrooms nationwide.
Its Significance:
If AI is helping shape the rulings that decide custody battles, criminal sentences, and civil lawsuits, most people have no idea it's happening. Right now, there are no nationwide rules requiring judges to disclose when AI played a role in their decisions. That's very different from what lawyers face: attorneys who use AI incorrectly have already been fined and sanctioned. The concern isn't that AI can't help with routine tasks. It's that the line between "routine help" and "real judgment" is blurry, and errors that make it into official court orders are hard to take back. Some courts are starting to act, with a few judges requiring attorneys to certify how AI was used in filings. But most courtrooms still have no written policy at all. What happens in a courtroom shapes real lives, and right now, the rules around AI in those rooms are mostly being made up on the fly.
QUICK TAKES
The story: Business Insider reviewed internal documents from a project called Stagecraft, run through data-labeling startup Handshake AI, in which 3,000 to 4,000 freelancers are being paid at least $50 an hour to help train ChatGPT on niche occupations including commercial flying, animal husbandry, music composition, and agriculture. Contractors are asked to build a professional persona and create tasks that reflect real work in their field, with three rounds of review before the data reaches OpenAI.
Your takeaway: One contractor summed it up plainly: "We all were aware that we were basically training AI to replace us." This is how AI gets good at specialized jobs: not by magic, but by paying actual experts to teach it, one occupation at a time.
The story: Anthropic accidentally exposed over 500,000 lines of internal source code for its Claude Code tool after a debug file was mistakenly bundled into a public software update. When the company tried to pull copies off GitHub, its copyright takedown notices hit roughly 8,100 repositories, including many that had nothing to do with the leak, like innocent forks of Anthropic's own public code.
Your takeaway: Anthropic quickly walked back most of the takedowns, but the code was already everywhere. For a company preparing for a potential IPO, leaking your own source code and then accidentally nuking thousands of developers' projects in one week is not a great look.
The story: A new Lumina Foundation-Gallup survey of nearly 4,000 college students found that close to half have considered changing their major because of concerns about what AI could do to their career. Already, 16% say they've actually switched, with men more likely to have done so than women (21% vs. 12%).
Your takeaway: AI anxiety is no longer something students are just talking about in the abstract. It's showing up in real decisions about what to study, and that is going to reshape which programs grow, which shrink, and what a college degree looks like in the next five years.
TOOLS ON OUR RADAR
🐧 DeerFlow Free and Open Source: A highly flexible automation harness that breaks down your complex instructions into smaller tasks for parallel processing and code execution in a secure self hostable environment. (Alternative to LangGraph)
📝 Prodini Freemium: A specialized artificial intelligence product manager that analyzes your existing data and project history to instantly generate professional requirement documents and surface potential technical risks.
✉️ Stamp Paid: A reimagined email and calendar workspace that acts as your personal artificial intelligence secretary by prioritizing your inbox and drafting replies in your specific voice while managing your schedule.
📊 UTM Mind Freemium: A specialized marketing governance platform that uses conversational agents to help your team build and manage standardized tracking links for all your digital advertising campaigns.
TRENDING
Cognichip Raises $60M to Use AI to Design Computer Chips — Startup Cognichip just raised $60 million, bringing its total to $93 million, to build AI that helps engineers design the chips that power AI. Chip design currently takes three to five years and costs hundreds of millions of dollars. Cognichip says its approach can cut that timeline by more than half and costs by over 75%, using a model trained specifically on chip design data rather than a general-purpose AI. Intel CEO Lip-Bu Tan is joining the board.
What Anthropic's Leaked Code Revealed About Its Own Development — Before the takedown notices flew, thousands of developers pored over Anthropic's leaked source code and found a lot. Among the discoveries: an always-on background AI agent that can act without being asked, an "undercover" mode designed to hide the fact that Claude is contributing to public code repositories, and a mood tracker that measures a coder's frustration level. Anthropic says the code leaked due to human error and that no customer data was exposed.
Shake Shack Is Using AI to Run Its Restaurants Better — Shake Shack launched "Project Catalyst," a tech overhaul aimed at helping it grow to 1,500 locations. The plan includes new point-of-sale and kitchen display systems, its first-ever loyalty program, and AI tools that give managers real-time alerts about what's happening in their restaurant, from bottlenecks in the kitchen to shifts in customer traffic. The AI layer is designed to run quietly in the background so team members don't even need to think about it.
AI Can Now Predict Research Trends Two to Three Years Before They Happen — Researchers at Germany's Karlsruhe Institute of Technology trained an AI to analyze materials science publications and map relationships between concepts, then used it to identify research directions that turned out to gain traction two to three years later. The results, published in Nature Machine Intelligence, suggest AI could help scientists figure out where to focus before the field gets there on its own. The approach matters because the volume of scientific papers is growing so fast that researchers can't keep up with their own fields anymore.
ChatGPT May Help You Learn Faster, But You'll Remember Less — A study of 120 university students found that those who used ChatGPT for an assignment scored 11% higher right after the task than peers who used traditional methods. But when both groups were surprise-tested 45 days later, the ChatGPT group scored noticeably lower: an average of 5.75 out of 10 versus 6.85 for the traditional group. The researcher's conclusion: using AI to find and organize information reduces the mental effort that helps things stick in long-term memory.
Reasoning Failures Are Keeping AI From Reaching Human-Level Intelligence — Researchers and AI critics are pushing back on claims that current AI models are close to human-level thinking, pointing to consistent failures in basic reasoning tasks. Despite scoring high on academic benchmarks, today's AI systems still regularly fail on problems that require common sense, context awareness, or multi-step logic that humans handle without effort. The argument is that building a "digital mind" requires something more fundamental than scaling up the current approach.
Slack's New AI Can Coordinate Multiple AI Agents at Once — Salesforce has added agent orchestration to Slack, meaning the AI built into Slack can now manage and direct multiple specialized AI agents working on the same task at the same time. Instead of one AI trying to handle everything, different agents can take on different parts of a job, like pulling data, drafting a response, and routing a decision, all inside one conversation thread. Salesforce is positioning this as a way for teams to run more complex automated workflows without leaving Slack.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🃏 Build a flashcard study app with flip animation and deck management
Build a flashcard study app. Let me create multiple decks with custom names and colors. Each card has a question and answer that flips with an animation when I click it. Track which cards I've marked as "got it" vs "still learning". Show a progress bar per deck. Include a study mode that only shows unlearned cards, and let me add new cards to any deck. Use a dark purple color scheme. React and Babel, all in one HTML file.What this does:
This one has a satisfying card flip animation built in. Create decks for anything you’re studying, add questions and answers, and tap to flip. The “Got it” and “Still learning” buttons sort cards automatically so you focus on what you don’t know yet. Works great for languages, vocabulary, or anything you need to memorize.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Predict which scientific research topics will become hot fields two to three years before most researchers notice them.
❌ Still Can't: Reliably reason through multi-step logic problems the way humans do, which is why researchers say true human-level intelligence remains out of reach.
✅ AI Can Now: Help judges prepare for hearings by summarizing cases, building timelines, and drafting questions for attorneys.
❌ Still Can't: Be trusted to draft court documents without human review, as fake citations and invented facts have already made it into real federal court orders.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Based on a series of Ken Liu short stories, an animated series about what happens when human consciousness gets uploaded to the internet. The animation is understated but the ideas aren't, covering corporate control of digital minds, what identity means without a body, and whether an uploaded person is still a person at all. AMC cancelled it after one season then quietly released the second, which actually sticks the landing. Genuinely one of the most thoughtful explorations of AI consciousness in any medium, animated or otherwise. The fact that it flew completely under the radar says more about TV trends than the show itself.
Your next great hire lives in Slack.
Viktor is an AI coworker that connects to your tools and ships real work. Ask Viktor to pull a report, build a client dashboard, or source 200 leads matching your ICP. Most teams hand over half their ops within a week.
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.
