In partnership with
Attio is the AI CRM for high-growth teams.
Connect your email, calls, product data and more, and Attio instantly builds your CRM with enriched data and complete context. Whether you’re running product-led growth or enterprise sales, Attio adapts to your unique GTM motion.
Then Ask Attio to plan your next move.
Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.
Ask more from your CRM.
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 Just Hacked Two-Factor Login, and OpenAI's Counter-Move Could Save Your Job
TLDR: OpenAI launched Daybreak, a free cybersecurity program that uses AI to find and fix software bugs before hackers do, on the same day Google confirmed criminals used AI to build the first real-world zero-day attack.
The Story: OpenAI unveiled Daybreak on Monday, pairing its frontier models with Codex Security to scan code for hidden bugs, propose patches, and test fixes in isolated environments. Cloudflare signed on as the launch partner. Sam Altman said AI is already good at cybersecurity and is about to get much better, and he wants to start working with as many companies as possible right now. The timing wasn't an accident. Hours earlier, Google's Threat Intelligence Group reported the first confirmed case of hackers using AI to build a zero-day exploit in the wild, a Python script that bypassed two-factor authentication on a popular open-source admin tool. Google caught it before the attack scaled. The script gave itself away with textbook AI tells, including a fake CVSS score the model hallucinated.
Its Significance: This is the moment AI moved from helping hackers write phishing emails to building the actual weapons. Two-factor authentication is what stands between your bank account, your work email, and someone who already has your password. If AI can find ways around it once, it can find ways around it again. The flip side is that like any technology that can be used by criminals, the same tools work for defenders. Mozilla used Anthropic's Mythos model to find 271 unknown bugs in Firefox last month. Claude’s Mythos was released to a select few for the same reason in private and government sectors, to sure up their defenses. The race now is whether the people building free tools like Daybreak can outpace the criminals using the same tech to break in.
QUICK TAKES
The story: Baidu released ERNIE 5.1, which it says cost about 6% of what comparable AI models spend on training. The model ranks fourth globally on the LMArena Search leaderboard and beat DeepSeek's previous Chinese benchmark on agentic tasks like filling spreadsheets and browsing the web.
Your takeaway: Cheap, capable Chinese models keep coming. When training costs drop this fast, the gap between the most expensive US labs and everyone else gets harder to defend. That puts pressure on what American AI companies can charge for access.
The story: GM cut about 600 IT workers, more than 10% of its IT staff. The company says it's still hiring for the same department, but only for AI-native development, agent and model building, prompt engineering, and AI workflows.
Your takeaway: This is what AI job displacement actually looks like inside a Fortune 50. Not "AI replaced workers." More like "the skills that got you hired five years ago aren't the skills they want now." If your job description says IT and doesn't say AI, this is the memo.
The story: Google's new REPLIQA program pairs Google Quantum AI with Harvard, MIT, UC San Diego, UC Santa Barbara, and the University of Arizona. The goal is to simulate molecules like the P450 enzyme, which is key to drug development and too complex for regular computers to model accurately.
Your takeaway: Quantum plus AI is starting to tackle problems neither tool could handle alone. This won't show up in your pharmacy next year, but if it works, drugs that take a decade to develop today could move much faster. Worth watching for the long game.
TOOLS ON OUR RADAR
🚀 Spacedrive Free and Open Source: A futuristic file manager that connects all of your cloud storage accounts and hard drives together into one single beautiful folder interface.
📺 SmartTube Free and Open Source: A phenomenal application for your smart television that lets you watch online videos completely free of annoying promotional segments or commercial interruptions.
⚔️ Wesnoth Free and Open Source: An incredibly deep and highly engaging fantasy strategy game where you command magical armies and explore a massive beautifully drawn world.
📖 Mihon Free and Open Source: A fantastic digital comic application for your phone that allows you to easily discover organize and read thousands of manga titles and graphic novels.
TRENDING
A Developer Used AI to Build a Tool That Figures Out What Wakes Him Up at Night — Martin paired his Garmin watch sleep data with a Raspberry Pi microphone and built the whole thing in about 8 hours. AI didn't solve his problem. It made it cheap enough for him to solve it himself.
Hackers Used AI to Build a Zero-Day Exploit That Bypasses Two-Factor Authentication — Google's Threat Intelligence Group caught the first confirmed AI-built zero-day in the wild, a Python script designed for a mass exploitation attack. Google says the script's structure and hallucinated CVSS score gave it away.
Mira Murati's Thinking Machines Reveals 'Interaction Models' — The former OpenAI CTO's startup unveiled its first product, an AI that processes audio, video, and text at the same time and responds in under 0.4 seconds. That's faster than Google's Gemini-3.1-flash-live on the same benchmark.
Palo Alto Networks Launches Idira to Secure AI Agents — The company says machine and AI identities now outnumber humans inside companies by 109 to 1, and 9 out of 10 organizations had an identity-related breach last year. Idira tries to lock down access for every human, machine, and AI agent on a single platform.
UCF Students Boo Commencement Speaker for Calling AI 'The Next Industrial Revolution' — A graduating class of artists and communicators was not having it. One student yelled "AI sucks." Another called the speech "tone-deaf" and "ad-like." The speaker, a real estate executive, seemed genuinely surprised.
Why Agentic AI Is Security's Next Blind Spot — AI agents are already running in production at most companies, executing tasks and accessing data without the security team really understanding what they do. The piece argues that policy alone won't fix it. Security teams need to actually build and test agents themselves.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🧱 Give it any belief or assumption. Watch it strip every layer until only bedrock truth remains — then rebuild from there.
Build a single-file HTML app using vanilla HTML, CSS, JS, and one API call. Create First Principles Stripper — a tool that deconstructs any belief layer by layer until bedrock truth remains. Use localStorage key 'first_principles_v1'.
Aesthetic: warm off-white (#f5f0e8), stark black (#141410) typography, diagonal stripe texture overlay, Unbounded 900 for headings, Karla for body text, Overpass Mono for labels. Brutalist card style with 4px offset box shadows and 2px borders. Layer cards get progressively warmer/darker as depth increases.
Form: large textarea for the belief, domain dropdown (7 options), how-deeply-held dropdown.
Call the API instructing it to be ruthless but fair — distinguish what is assumed from what is actually known. Return raw JSON: belief_clean, layer_count, layers array (depth label, the stripping question, 2-3 sentence answer on what falls away), bedrock (1-2 sentences of irreducible truth — should feel like striking rock), rebuilt (the belief rebuilt from first principles), survived array (parts that held up), dissolved array (parts that dissolved).
Render: a black header bar with the belief text and layer count, a stacked layer card system where each card has a vertical depth label on the left, the stripping question in bold, and the answer below — backgrounds get progressively warmer as depth increases, a black bedrock card with BEDROCK watermark, a rebuild section, and a survived/dissolved chip row showing what made it and what didn't (dissolved items shown with strikethrough). Save to localStorage. Make it work in a single HTML file.What this does: Enter any belief, assumption, or piece of conventional wisdom — "you need a degree," "more features make a better product," "I'm not a creative person." The deconstruction engine asks "why" and "is this actually true?" at each layer, working through 4–6 levels until it hits the irreducible bedrock — the things that are true regardless of context, culture, or convention. Then it rebuilds: what would this belief look like if you started from the foundation instead of the assumption? A survived/dissolved section shows which parts of the original held up and which dissolved under scrutiny. Every deconstruction saves to localStorage.
What this looks like:
Stop re-prompting. Say it right the first time.
Voice-first prompts preserve the nuance you cut when typing. Speak once, paste into any AI tool, get results that don't need a follow-up. 89% of messages sent with zero edits.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Find and weaponize zero-day software bugs that bypass two-factor authentication
❌ Still Can't: Be trusted to run inside companies without identity controls built specifically for AI agents
✅ AI Can Now: Help one developer build a custom home sleep-tracking tool in a single weekend
❌ Still Can't: Win over a room of new graduates worried about their first jobs
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
A wordless animated film about a lonely dog who buys a build-it-yourself robot companion and accidentally creates something resembling true friendship. No dialogue at all, just two characters and an entire emotional arc about loneliness, attachment, and what we expect from artificial beings. Got an Oscar nomination and almost nobody watched it.
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.
