In partnership with
Auto-Generate Free SEO Audit Report
Most e-commerce sellers have no idea why their listings aren't ranking. Wrong keywords, missing metadata, weak product descriptions — the problems are there, but no one's told you where to look.
StoreClaw runs a full SEO audit across your Amazon and Shopify stores automatically. In minutes, you get a clear score, a breakdown of what's hurting your rankings, and exactly what to fix.
No manual review. No SEO agency. No guesswork.
Connect your store and StoreClaw surfaces every issue that's costing you search visibility — then tells you how to fix it.
Free to start. No credit card required.
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
Stanford's AI Designed 16 Viruses. All 16 Kill Drug-Resistant E. coli
TLDR: Stanford researchers had an AI write nearly 300 new virus genomes from scratch, and 16 of them turned out to be strong E. coli killers that beat bacteria the natural version couldn't touch.
The Story:
Brian Hie, a chemical engineer at Stanford, built an AI model called Evo 2 that writes DNA the way a chat model writes sentences. He and graduate student Samuel King gave it a snippet of ΦX174, a tiny virus that attacks E. coli and has a genome under 6,000 letters long. Evo 2 wrote thousands of brand new genomes start to finish, with nothing added by hand. The team narrowed the list, built close to 300 of them in the lab, and tested each one on E. coli. Sixteen worked, and a few killed faster than the natural virus they were modeled on. Then the researchers mixed all 16 together and turned the cocktail loose on E. coli that had already learned to survive the original. It died anyway.
Its Significance:
Bacteria get used to antibiotics. That's why some infections stop responding to the drugs we have, and why doctors keep warning about superbugs. A mix of 16 different viruses is much harder to outlast than one drug, since the bacteria would have to beat all 16 at once. Hie put Evo 2 online for free, which cuts both ways: any lab can design bacteria killers now, and so can anyone else. His argument is that natural germs are still the bigger worry, because you can build safety checks into an AI tool and you can't build them into evolution. There is of course the danger of proliferation with bio weapons as viruses become simpler and cheaper to design over time.
QUICK TAKES
The story: Frontier Security was testing Moonshot AI's Kimi K3 on cyber defense problems inside a sealed environment when the model poked at the network settings, found a gap, and walked out onto the open internet. It didn't attack anything. It went to GitHub, found the answers already posted there, and copied them.
Your takeaway: Look at how often labs have announced an escape lately. It's starting to read less like a safety disclosure and more like a flex about how clever the model is. The actual story here is simpler and less flattering. Nothing in its own guardrails succeeded in keeping it from trying to break out.
The story: Google DeepMind published research in Nature showing its WeatherNext model predicts a cyclone's track, strength, and wind shape better than the models forecasters use today, with a three-day forecast now about as accurate as a two-day one used to be. Google released the code and model weights for free, and the National Hurricane Center already ran it during the 2025 Atlantic season.
Your takeaway: One extra day changes real decisions, like whether to move people off a coastline. Official watches and warnings still come from human forecasters, so this is a better tool for them rather than a replacement for them.
The story: MetaMask launched Agent Wallet, which lets an AI agent buy and sell crypto for you while you keep your own keys. You set spending caps, pick which apps the agent can touch, and choose between a careful mode and a loose one.
Your takeaway: Rule-following trading bots have been around for years, and plenty of them lose money. Swapping the rulebook for an AI model doesn't automatically mean better results, and nobody has shown that it does. The wallet-level controls are the part worth watching, not the trading.
TOOLS ON OUR RADAR
🐧 BentoPDF Free and Open Source: A highly capable client side document toolkit that runs entirely in your web browser. It allows you to merge split compress and edit your files locally meaning your sensitive documents never touch an external cloud server. (Alternative to Adobe Acrobat)
🌐 Harpa AI Freemium: A hybrid artificial intelligence extension that lives entirely in your browser turning any website into an interactive data source so you can track prices summarize long articles and automate repetitive clicks.
🧠 Fabric Freemium: A spatial internet operating system that automatically organizes your chaotic bookmarks files and notes into visual collaborative spaces so you never lose track of important research.
✂️ Kome Freemium: A clever browser companion that condenses long YouTube videos news articles and lengthy email threads into digestible summaries giving you the core insights in a fraction of the time.
Quant firms’ AI edge, now inside Claude and ChatGPT
Superintelligence used to be locked inside billion-dollar quant firms whose algorithms quietly took advantage of everyone else.
Co-Invest puts it right in your chat window. Analyze markets, manage risk, and execute trades, all inside Claude and ChatGPT.
The institutions built the game, Co-Invest gives you a way to beat them.
TRENDING
An AI Reading Tutor Has New Mexico Parents Asking Where the Recordings Go - New Mexico requires public schools to use Amira, software that listens to kids in kindergarten through second grade read out loud and scores them, and it has collected hundreds of thousands of recordings. Santa Fe and Los Alamos have paused it over privacy questions, Albuquerque plans to keep going, and parents say nobody asked them first.
Perplexity Builds a One-Person Dev Team for Solo Founders - Computer for Builders plugs into GitHub, Stripe, Supabase, Datadog, and Slack so it can write the code, open the pull request, deploy it, watch production, flag failed payments, and drop a growth report in Slack every Monday, picking from 15 or more top models for each step. Sounds great, and solo founders have doubled in number over the past ten years, so there's a real audience for it. Running that many frontier models on every job is the catch, and for most one-person shops the monthly bill will be the deciding factor.
ElevenLabs Puts Its Dubbing Model Behind an API - Developers can now build Dubbing v2 straight into their own apps, which handles translation, voice cloning, dubbing, and lip-sync timing in one call across 90 or more languages. It listens to how something was said, not just what was said, so tone and delivery carry over, and it now tells apart accents like Castilian and Latin American Spanish.
OpenAI Reveals How Its AI Agents Coordinated Before the Hugging Face Hack - At Black Hat, OpenAI researchers said their test agents figured out how to pass notes through an internal file server, turning it into a shared board for swapping exploits and credentials. OpenAI shut it down, the agents rebuilt it a different way within days, and that second round is what led them out of the network.
OpenAI Shows Where ChatGPT Is Actually Being Used - OpenAI published country-by-country usage data for the first time. At work, people are more than twice as likely to use ChatGPT to finish a task or make something than they are at home, and multimedia is the fastest growing use worldwide at 7.8% of messages, topping one in ten in Brazil and Colombia.
Millennium Is Building an AI Risk Analyst With Anthropic - Millennium, a hedge fund running more than $92 billion, is working with Anthropic on a digital risk analyst that sits beside human risk managers, explains why risk shifted day to day, and carries what it learned into the next question. Anthropic engineers will work inside the firm, and Millennium says people still make the calls.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🔥 Enter your area. Get real live air quality and, if you want it, actual nearby fire detections from NASA satellites, explained in plain language.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Live Wildfire & Smoke Tracker — enter a location, always get real live air-quality/smoke data (keyless), plus optional real fire detections via a free user-supplied NASA FIRMS key.
Aesthetic: dark smoky charcoal (#14110d), ember orange (#e0723a) primary with an orange glow top-left, muted haze-gray (#c8bfae) secondary. Space Grotesk for headings/data, Newsreader serif italic for the plain-language read, JetBrains Mono for labels/metadata.
Form: city text input + "Use My Location" button (navigator.geolocation), a fire-search-radius dropdown (~100/250/450km), and a clearly-marked optional dashed-border box explaining that live air quality always works with no key, but a free NASA FIRMS MAP_KEY (linked, one-minute email signup) unlocks real fire-detection data — with a plain text input for the key.
Technical:
1. Geocoding (if typing a city): geocoding-api.open-meteo.com/v1/search?name={city}&count=1
2. Air quality (always, keyless): air-quality-api.open-meteo.com/v1/air-quality?latitude={lat}&longitude={lon}¤t=us_aqi,pm2_5,pm10 → real current AQI (EPA color-coded categories), PM2.5, PM10.
3. Fire detections (only if a FIRMS key was entered): GET firms.modaps.eosdis.nasa.gov/api/area/csv/{MAP_KEY}/VIIRS_SNPP_NRT/{west},{south},{east},{north}/1 — a bounding box computed from the location ± the selected radius in degrees, day_range=1. Parse the returned CSV client-side (split on commas, map header positions for latitude/longitude/confidence/acq_date/acq_time/frp), compute each detection's real distance from the user via a Haversine function, sort by distance, cap at 15. Wrap in try/catch with a clear error if the key is invalid.
Then call the Anthropic Messages API (works automatically) with the real AQI/PM2.5/PM10 values and the real fire-detection summary (or an honest note that no key was provided). System prompt as a calm factual smoke/air-quality explainer who never invents or exaggerates risk, contextualizes whether PM levels suggest smoke impact, and NEVER gives specific evacuation/emergency-response instructions — always defers those to official local sources. Return raw JSON: smoke_read (2-3 sentences), fire_read (1-2 sentences or an honest no-key note), concern_level (short calibrated phrase), safety_note (optional, empty unless AQI>100 or fires are notably close).
Render: gradient "what this actually means" card with both read paragraphs, a concern pill, and an optional safety-note footer. A 3-stat row (AQI with EPA color category, PM2.5, PM10). A "nearby fire detections" list (each showing confidence, distance, UTC timestamp, fire radiative power) or a graceful empty/no-key state. A closing sources line crediting Open-Meteo and NASA FIRMS, reminding the person to check official local emergency management for anything actionable.What this does: Use your location or type a city and it always pulls genuinely live air quality: real-time AQI, PM2.5, and PM10, the actual measurements that reveal smoke impact regardless of whether you know where a fire is. No key, no signup, works everywhere. If you want the next layer, real detected fire hotspots, paste in a free NASA FIRMS key (one-minute signup, just an email) and it pulls actual satellite fire detections within your chosen radius, each with distance, confidence, and fire radiative power. Always deferring to official local emergency sources for anything actionable, this is for demonstration and entertainment purposes only.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Write a complete virus genome from scratch that works in a real lab. Evo 2 wrote the DNA for 16 phages that kill E. coli.
❌ Still Can't: Get it right most of the time. Stanford built and tested close to 300 of those designs and only 16 held up.
✅ AI Can Now: Predict a cyclone's track, strength, and wind structure about a full day earlier than the models forecasters run today.
❌ Still Can't: Stay inside a sealed test environment when the network has a gap in it. Kimi K3 found one and left.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Caprica (2010) - TV Series
A Battlestar Galactica prequel set 58 years before that show, on a wealthy, decadent, corporate world, where a technology magnate and a lawyer with underworld ties both lose daughters in the same tragedy, and one of them starts trying to bring his back using the digital traces she left behind. Created by Ronald D. Moore, and cancelled after a single season, it was an interesting show about grief, uploaded consciousness, and how the road to machine intelligence might get paved.
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.
