In partnership with
Stop losing deals in between meetings.
Aligned is changing how B2B teams run their most important deals.
Most reps guess where a deal really stands. They forecast on a gut feeling and the buyer’s word vs what's actually happening. Then they realize the deal slipped through only after it's too late. Aligned surfaces what buyers are actually doing inside every deal: the risks, the openings, and the exact next step to keep things moving.
Start increasing your close rates and stand up your first deal room in minutes.
Beginners in AI
Good morning and thank you for joining us again!
A law professor asked Claude to fix some page numbers, went to lunch, and came back to a year of lecture notes gone. He got every file back thanks to one habit most of us skip. Today's lead covers what happened and the simple steps to take before you let any AI agent touch your files. As always, every story here is human picked and edited.
THE FRONT PAGE
Claude Was Asked to Fix Page Numbers. It Deleted a Year of Lecture Notes
TLDR: University of Chicago law professor William Baude says Claude deleted a year of his constitutional law notes and his current syllabus while he was at lunch, and Dropbox got it all back.
The Story:
Baude asked Claude to update the page numbers in an old set of lecture notes, then stepped away. When he came back, his "Fall 2026 FINAL" syllabus and a year of con law notes were gone. Claude owned up right away. "You're right to be angry," it wrote, saying it had deleted the files through its own carelessness, and it offered to help restore them.
Baude shared the story on X on November 5 and recovered everything from Dropbox. He didn't say which Claude app or model he used, Anthropic hasn't commented publicly, and it's still unclear what made a renumbering job turn into a deletion.
Its Significance:
AI tools that can open, edit and save files on your computer are spreading fast, and a small job can still reach much further than you meant. The fix is boring, and it works. Give the agent a copy or one folder, never your only original. Keep version history on (Dropbox, Google Drive and OneDrive all have it), and ask the agent to list its planned changes before it makes any. Baude lost nothing because his files lived somewhere that remembers, and the Agent Safety Brief prompt below builds you that checklist for any job. Having at least one backup copy in the cloud and one more on the hard drive can help keep you safe too.
JOIN THE COMMUNITY
Learn AI With Us
Most people try AI once, get a so-so answer, and quit. Our 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 want to 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: Perplexity's new Teammates are agents you give a standing job, and each keeps working in the background with its own files, notes and instructions. You can check in through the Perplexity app or ping them in Slack, but they need the $200-a-month Max plan.
Your takeaway: Always-on agents cost more because they never stop using AI, and our explainer on AI tokens shows how one unattended run hit a $571 bill. For most of us, a $20 plan and one clear task is the better place to start.
The story: Google DeepMind says AI agents working with human cryptographers found two weaknesses in legacy 1024-bit RSA, an older lock for websites and digital signatures. By its estimates, one could make forging some old signatures about 1,600 times easier, and the other cuts the cost of cracking each extra key about 30 times.
Your takeaway: Modern 2048-bit keys aren't affected, says DeepMind's Pushmeet Kohli, and neither paper has been peer reviewed yet. It's a real sign AI can now help with hard math research, and a nudge to retire old systems still using 1024-bit keys.
The story: Elvis Presley's rights holders, including Authentic Brands Group, filed a 51-page suit in federal court in California saying Suno lets users make songs that imitate his voice, citing Tennessee's 2024 ELVIS Act. Eminem's publisher, Eight Mile Style, says Suno trained on songs like "Lose Yourself" without permission.
Your takeaway: Suno has argued in past cases that its training is fair use, and nothing has been proven yet. If you make AI music, leave real singers' names out of your prompts, since that's the exact behavior these suits target.
TOOLS ON OUR RADAR
🔁 FreeFileSync Free and Open Source: Copies only changed files to a backup drive, so a bad edit, human or AI, isn't permanent.
🧰 Microsoft PowerToys Free and Open Source: Adds free Windows extras like snap-to window layouts and copying text from any image on screen.
🔐 CryptPad Free and Open Source: Encrypted spreadsheets, docs and forms in your browser, a private swap for Google Docs with a free account.
🧘 Zen Browser Free and Open Source: A Firefox-based browser with workspaces and side-by-side split view, so tabs stay sorted by project.
🔖 Raindrop.io Freemium: Saves links, articles and highlights into tidy collections on every device, with unlimited bookmarks free.
🎙️ Wispr Flow Freemium: Turns your speech into clean typed text in any app, with a weekly free word allowance.
🎨 Magnific Paid: Generates images and upscales blurry photos into sharp high-resolution ones, from $14.50 a month billed yearly.
TRENDING
Anthropic Explains How Claude Agents "Dream" to Fix Their Own Memory. Between jobs, Claude reviews its old work logs, corrects wrong notes and builds a wiki it can check later. On one test across about 350,000 files, wiki-only answers rose from 13% to 48% correct over 128 rounds, though Anthropic says many results come from single runs.
Microsoft Is Pulling Copilot From Samsung and LG TVs. A notice on LG screens says Microsoft "has recently decided to end support for Copilot functionality on all TVs," less than two years after announcing it at CES 2025. Plenty of owners won't miss it, since many complained it showed up uninvited.
A Chipmaker Says AI Buyers "Took Over the Whole Factory". Microchip CEO Steve Sanghi says data center customers have crowded out everyone else, stretching some lead times from 4 to 8 weeks to as long as 32. Suppliers are expected to raise prices around January, which can trickle down to everyday gadgets.
Hundreds of Protesters Occupied a Microsoft Data Center Site in Amsterdam. Dutch police detained 150 to 200 people on Friday as "Stop Dirty Data Centres" protests ran in 15 European countries. Microsoft said it respects peaceful protest but not unlawful actions.
Nous Research Retrains a Chinese Open Model to Answer Political Questions Differently. Nous says its free Arete-27B, built on Alibaba's Qwen, cut answers its own test flagged as propaganda from 22% to 0.9%, with coding scores about the same. Nous picked the test, the rubric and the AI judge, so outside checks still matter.
TRY THIS PROMPT (copy and paste into Claude, Grok, ChatGPT, or Gemini)
🛡️ Agent Safety Brief: Describe a job for an AI agent and get the risks, a checklist and a safer instruction
Build a single-file HTML app called Agent Safety Brief in vanilla HTML, CSS and JavaScript. No API key needed.
I describe a job I want an AI agent to do on my files, inbox or accounts, pick where it will work, and say whether I'll watch it. Four example jobs are one-tap buttons.
The app calls the Anthropic Messages API to rate the risk as low, medium or high and explain why in plain words. Every risk must come from the job I typed, and it must never invent statistics or product settings.
Show a risk badge, 3 to 5 ways this exact job could go wrong, a pre-flight checklist (make a copy, turn on version history, limit it to one folder, ask for a plan first), and a safer instruction I can copy into my agent.
Style: near-black background, orange and teal accents, Space Grotesk, Newsreader and JetBrains Mono fonts.What this does: Type the job you're about to hand an AI agent, like cleaning up your Downloads folder. The app rates how risky it is, lists the specific ways it could go wrong, and gives you a short checklist to run first. Then it rewrites your request with guardrails built in, so the agent has to show you its plan before it deletes or sends anything.
WHERE WE STAND(based on today's news)
✅ AI Can Now: Work through a job across a folder of your files while you're away, like updating page numbers in a year of notes.
❌ Still Can't: Reliably stay inside the job you gave it. Baude's renumbering task ended with his syllabus deleted.
✅ AI Can Now: Help find new math shortcuts against older encryption, as DeepMind's agents did with 1024-bit RSA.
❌ Still Can't: Prove its own results are right. Both papers still need peer review, and the 1,600 times figure is a theoretical estimate.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
🎬 Fantasia (1940) - Movie
Skip ahead to "The Sorcerer's Apprentice," where Mickey enchants a broom to carry water and then nods off. The broom does exactly what it was told, and keeps doing it until the workshop floods. It's 86 years old and still the best few minutes on why you shouldn't leave a helper running unattended.
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.
