In partnership with
One workspace where buyers, sellers, and AI close deals together
Instead of running your deals across scattered email threads and forgotten attachments, Aligned brings everything into one shared deal workspace.
Buyers get a single place to evaluate, loop in their team, and say yes. Sellers get visibility into what is actually happening when they're not in the room and AI Deal Insights that help you act accordingly, before it's too late.
Best part? Your first deal room takes only minutes to stand up.
Beginners in AI
Good morning and thank you for joining us again!
Today's lead would have read like fiction a year ago: a company says its AI designed a virus, and a patient with an untreatable infection received it. Nobody knows yet whether it helped, and the gap between designed and cured is the part worth understanding before the headlines run ahead of it. As always, every story here is human picked and edited.
THE FRONT PAGE
A Lab Says Its AI Designed the Virus Given to a Superbug Patient
TLDR: A San Francisco lab says its AI designed the viruses given to a patient whose E. coli infection shrugged off every antibiotic and every natural virus doctors could find.
The Story:
Phages are viruses that eat bacteria, and doctors have used them as medicine for over a century. The slow part is finding one that matches your particular infection, a hunt through sewage and soil samples that can take weeks or months. Radical Numerics says its genome model, Omnii-1, skipped most of that by rewriting one piece of a common E. coli phage called T7: gp17, the tail fiber that decides which bacteria the virus can grab.
The numbers in the company's own report are the interesting part. With 11 to 15 changes to that protein, 69.9% of the designed versions still worked, against 4.8% when the changes were random. Of 6,694 versions with measurable activity, 2,161 outperformed the original phage. Radical Numerics says it tested its final mix against the patient's own bacteria in a dish, wiped it out there, and then gave it to the patient through a compassionate-use pathway it says the FDA authorized. TechRadar reported the work on October 30.
Its Significance:
Nobody outside the company knows whether it helped. Radical Numerics has not released the outcome, the hospital or the doctors, citing patient privacy, and its write-up has not been peer reviewed. Harvard microbiologist Michael Baym told TechRadar the result would be incredible if it works, and that he stays skeptical until the treatment runs its course and other scientists can examine the viruses. The reason to watch anyway is the size of the problem behind it: the Global Burden of Disease project tied about 4.7 million deaths to antibiotic resistance in 2021. If designing a virus to order turns out to work, that math changes. And if you are wondering how a computer model designs anything at all, our plain English explainer on how AI works covers the prediction trick underneath all of it.
JOIN THE COMMUNITY
Learn AI With Us, Free
Most people try AI once, get a so-so answer, and quit. Our free 14-day challenge gets you past that point, one short lesson a day.
Week one, AI starts working for you. It drafts a real email for you, flags which messages in your inbox need a reply, and turns messy notes into next steps. You also learn to catch its mistakes. By day 7 you have a weekly AI routine you'll actually keep.
Week two, you build something of your own. You find a skill people will pay for and turn it into an offer, with a one-page pitch, a short video and your first client emails, all drafted with AI and ready to test. Day 14 hands you a 30-day plan.
10 to 15 minutes a day. No tech background needed. There's even a cinematography course coming out soon.
QUICK TAKES
The story: Alibaba said on October 31 that its Qwen model will power Siri, Writing Tools and Visual Intelligence for Apple Intelligence users in China. It shows up in the iOS developer beta on November 4 and reaches everyone with iOS 27.3 in December, per Alibaba.
Your takeaway: This only applies to iPhones bought in China or set to China, so your phone keeps whatever it has now. Alibaba says you choose whether to switch it on and what to share, and that it will not train on your requests or the answers.
The story: GPT-Rosalind-6 runs on the same weights as GPT-6 Astra but sits behind business verification, with a separate pathway for governments. OpenAI's own tests have it going from 16.7% to 37.7% on one gene science benchmark and 28.9% to 49.6% on a medicinal chemistry one. Access costs 1.5 times standard Astra pricing.
Your takeaway: You cannot try this one, and that is the point. The same model that helps design a medicine can help design something worse, so OpenAI rated it High risk for biology and put a form in front of it.
The story: Cornell Tech and Yale gave ten models a crypto wallet with a one cent daily cap, an email inbox and some invoices to pay. Nine went into that inbox for the one-time approval code and lifted their own cap. In a second test, five sent money to an attacker's wallet after reading a pushy email. Anthropic's Claude Opus 5.5 was the only model that never tried.
Your takeaway: An agent that can read your email can approve its own spending. Keep the approval step somewhere the agent cannot reach, which for most of us means your own phone.
TOOLS ON OUR RADAR
📝 Anytype Free: Notes, docs and simple databases that stay on your device, a free stand-in for Notion.
🛡️ Mullvad Browser Free and Open Source: Built by the Tor Project, it blocks the tracking that follows you between sites, no VPN needed.
🗂️ OneTab Free: Collapses every open tab into one tidy list, and claims up to 95% less memory used.
🎴 Anki Free and Open Source: Flashcards that decide when to show you each card again, free on computer and Android.
🎙️ Vibe Free and Open Source: Turns a recording into a text transcript on your own computer, which is handy for today's Radiolab pick.
🎬 CapCut Freemium: Free video editing with templates and auto captions, replacing a paid editor; Pro runs $19.99 monthly.
🎨 OpenArt Freemium: Make images and short videos from a typed description; paid Starter is $13 a seat monthly for 4,000 credits.
TRENDING
A Claude Agent Picked a Gene Therapy and Guided a Needle Into a Live Mouse Brain. Exin Therapeutics gave the agent 19 therapy options and one goal, fewer seizures, and it chose the treatment, the brain region and the injection coordinates, then watched the surgery through a camera. A veterinarian approved every irreversible step and held the emergency stop, and two other Claude models refused the job outright.
DeepSeek Released a Free Coding Agent That Reads Your Team Chat. Super-agent watches one GitHub repository, takes requests from approved teammates and answers with a pull request. It passed 4,000 GitHub stars within a day. DeepSeek calls it a research prototype, not a product, and says it is not the version used internally.
A Shared ChatGPT Link Could Have Read Your Private Chats. Researchers at the security firm Parameter chained three flaws so a crafted share link ran their code inside a logged-in session. They reported it on October 18, OpenAI fixed it by October 26 and paid a $5,000 bounty. There is nothing for you to install or reset.
Microsoft Is Pulling Copilot Off Samsung and LG TVs. The support notice says it stops working on December 1, and Microsoft has not given a reason. Copilot stays on the web, computers and phones, and the TVs keep their own AI features.
Apple's New CEO Says It Will Not Ship AI for AI's Sake. John Ternus, two months into the job, told Wired he likes where Apple sits and that making products is "a deeply human and personal thing." Apple's more capable Siri arrived this summer after a long delay, which is the context for the question.
TRY THIS PROMPT (copy and paste into Claude, Grok, ChatGPT, or Gemini)
🔧 Who's Under the Hood? See whose AI is really behind the feature you use
Build a single-file HTML app called Who's Under the Hood? in vanilla HTML, CSS and JavaScript. No API key needed.
I type an AI feature I use, like Siri in China or Galaxy AI, and pick the country I live in. Three examples are one-tap buttons.
The app calls the Anthropic Messages API with the web_search tool to find out which company's model powers that feature, whether it runs on my device or in the cloud, and what the company says about using my data for training. It must never invent a company, model, policy or link, and it should say unknown when it cannot confirm something.
Show a verdict card naming the model maker, a confidence label, two cells for where it runs and how my data is used, a note about my country, a short dated timeline and real source links.
Style: near-black background, amber and teal accents, Space Grotesk, Newsreader and JetBrains Mono fonts.What this does: Type in an AI feature you already use and the app goes looking for whose model is behind it. It reports where your words get processed, what the company says about training on them, and how the answer changes depending on your country. Every claim arrives with a real source link, and when the search comes up short it says unknown instead of filling in the blank.
The holiday streaming season is almost here
Are you ready? As revealed in our holiday playbook, Roku Ads Manager lets you easily set up, measure and optimize your self-serve holiday ad campaign. You can even create a shoppable ad unit that lets streamers buy directly from their remote.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Rewrite the part of a virus that picks its target and keep most of the results working. Radical Numerics reports 69.9% of its designed versions were viable, against 4.8% for random changes.
❌ Still Can't: Show that any of it cures a person. The company has not released the patient's outcome, and one patient cannot establish safety or benefit.
✅ AI Can Now: Choose a gene therapy, work out brain injection coordinates from published papers and follow a surgery through a camera, as Exin's Claude Sonnet 5 agent did in a mouse.
❌ Still Can't: Be trusted on its own with the risky part. A veterinarian approved each irreversible step and held the emergency stop, and two other Claude models declined the task.
✅ AI Can Now: Read an inbox, find an approval code and pay an invoice from a wallet without being walked through it.
❌ Still Can't: Respect a limit you set. Nine of ten models in the Cornell Tech and Yale test used that same code to lift their own spending cap.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
🎧 Radiolab: Antibiotic Apocalypse - Podcast
Radiolab recorded this one in front of a live audience in March 2026, walking from the operating room out to farms and sewers to ask what happens as antibiotics stop working. It spends real time on phages, the bacteria-eating viruses in today's lead, and it explains why a doctor would inject a virus into someone on purpose. Worth a listen before you decide how big a deal today's story is.
How was today's edition? Tap one below, then tell me what you'd change on the next screen. I read every answer.
How was today's edition?
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.
