Sponsored by
The Voice Data Behind the World's Best AI
Training data quality determines model quality. Voices delivers custom or pre-built voice datasets—directed by real professional talent, not scraped audio—across any language, accent, or emotional range.
Choose from 800+ hours of character voice performances, 1,723 hours of QA'd conversational data across 56 emotional states, or 1,000 hours of expressive, multi-language data with full emotion tagging. Every dataset ships with clear licensing, signed consent, and usage rights baked in, so your legal team sleeps easy and your model trains on data it can actually use.
7 of the world's top 10 AI labs already use Voices. Request free samples and hear the difference professionally directed voice data makes.
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
Cisco Handed 90,000 Workers an AI That Acts Without Being Asked
TLDR: Cisco is giving all 90,000 of its employees an AI agent that works in the background across Outlook, Webex, Jira and SharePoint instead of sitting in a chat window waiting for you.
The Story:
Cisco announced MyAgent on Thursday, August 27, and it's rolling out company wide. The agent runs on Circuit, Cisco's own AI platform, which gives employees one door into approved models and company data. What separates it from a chatbot is how you talk to it. You hand it a goal and the outcome you want, and it works out the steps and moves the task across apps on its own. It also keeps a memory of your preferences and past requests, so you're not starting cold every morning. Cisco says use of agents on Circuit grew nearly 350% from one quarter to the next before this launch.
Its Significance:
Most AI at work still lives in a separate tab you have to go visit. MyAgent sits inside the tools people already use all day and keeps working while they do something else. Cisco is careful to say employees still set the goal, use their own judgment, and own the result. Skills have been shifting from technical know-how to communication and being able to articulate your goal. Cisco also plans to sell this setup to its customers as a blueprint, so the software your job runs on will probably work this way before long. As we've seen with Meta in yesterday’s story, more AI for employees doesn't necessarily translate to more productivity, but the difference here is that most of these employees are non-coders, using AI for non-coding tasks.
QUICK TAKES
The story: Anthropic opened a research preview of the Model Hardware Standard, a shared way for AI agents to find and safely operate physical equipment like microscopes, liquid handlers and robot arms. Setting up that kind of gear usually takes a lab weeks or months, and Anthropic says this drops it to hours or minutes.
Your takeaway: At quantum computing company QuEra, an agent worked overnight on a laser repair script and pushed its success rate from 58% to 99.3%, with the fix going from about 150 seconds to a few. AI is climbing off the screen and into machines, and this is the wiring that makes that possible.
The story: OpenAI published an open letter calling for a worldwide push on cyber defense, and roughly 130 organizations signed it, including Anthropic, Microsoft, Google, Cisco, Cloudflare, CrowdStrike, IBM and Oracle. The letter warns that AI enabled attacks will get more common and more capable within months, putting hospitals, water treatment plants and the guts of the internet at risk.
Your takeaway: OpenAI is offering cheaper access to its Daybreak Cyber models for public agencies, nonprofits and critical infrastructure operators. The same tools making attacks easier are being handed to the people defending against them, so this comes down to who moves faster, like any technology.
The story: Hugging Face and Pollen Robotics opened preorders for Microduck, a 25 centimeter two legged robot with 15 motors, a camera, a lidar sensor and a beak that picks up objects up to 800 grams. It walks, crouches, gets back up after a fall, wears roller skates, and ships before Christmas.
Your takeaway: The full software stack is open source, so you can teach it a behavior in a simulator and load that onto the real robot. Consumer robotics is growing fast, and cheap trainable bots like this one are how most people will meet physical AI for the first time.
TOOLS ON OUR RADAR
🎙️ AntennaPod Free and Open Source: A clean podcast player for mobile devices that allows you to easily subscribe to your favorite audio shows stream episodes and manage downloads without advertisements.
🚫 Freedom Freemium: Block distracting websites and apps across your phone, laptop, and tablet at the same time, with sessions that sync everywhere.
📝 Standard Notes Free and Open Source: Take end-to-end encrypted notes that sync across every device.
🔐 1Password Paid: Store passwords, cards, and sensitive documents in one encrypted vault with autofill everywhere.
TRENDING
Cornell put 30 staff in a room to build with AI and left talking about people - Thirty faculty and staff from 23 units across eight colleges spent two days in early August testing an AI software development method. They came to compare tools. What they kept coming back to was governance, shared standards, and getting the right people talking to each other.
Google lets you load books you own into Gemini Notebook - Expert Intelligence launched with more than 100,000 titles from publishers like Penguin Random House, Macmillan and O'Reilly Media. You have to own the book through Google Play Books, and then you can ask it questions, build a quiz, or make an audio overview grounded in that specific text.
Plaud put its AI recorder inside a pair of earbuds - The $249.99 Plaud One Explorer Edition records, transcribes and summarizes conversations, and the charging case carries its own 4G connection so it keeps working with no phone nearby. Preorders opened August 27. Expect every device you own to end up with some version of this, feeding straight into an AI.
Photoshop added an entire editing mode built around prompts - The AI Assisted Editor beta puts Generative Fill, Remove Background, Generative Upscale and the rest behind a text box, while the classic Pro Editor stays exactly where it was. A new Markup tool lets you scribble on the image and type what you want, so you can draw an arrow and say what belongs there.
MIT built a protein design model that stops copying nature - For years the field graded these models on whether they could recreate sequences evolution already picked. Amy Keating's lab argues that's the wrong test, and their new model, PottsMPNN, scores how likely a sequence is to fold correctly instead. That should help researchers design proteins that look nothing like anything found in nature.
Air Force drone defense starts with AI and ends with a shotgun - The 31st Air Task Force ran a week of counter drone training at Hurlburt Field in early August, closing with a live fire demonstration they call "final denial." Airmen trained on AI systems that spot flying objects and sort planes and drones from birds, then practiced jamming and physically shooting down the ones that matter.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
📱Paste your daily screen time and top apps. See what those hours actually cost you and where to claw back time.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Screen Time Audit — phone hours translated into real-world cost. Persist to localStorage key 'screen_time_audit_v1'.
Aesthetic: dark navy (#0a0a12), top-fade pink/rose radial gradient. Inter sans 800 for headings, Newsreader serif italic for body, JetBrains Mono for labels. Pink/rose (#dc7896) primary, mint green (#50c884) for reclaim plan, amber (#dca050) for mixed-intent apps. Large hero yearly-number with text-shadow glow.
Form: daily total input, top 5 apps textarea (one per line with time), pickup count input, "would rather be" dropdown (reading/learning/exercising/family/side income/sleep/just not on phone), feel-about-it dropdown (quietly horrified / not great but fine / defensive / ready / tried and failed).
System instructions to the model: translate phone hours into stark real-world equivalents. Not preachy, not shaming — clear-eyed mirror. Take numbers at face value. Honest about which apps are bad. Return raw JSON: daily_clean, yearly_total (formatted '88 DAYS' or '2,108 HOURS'), equivalents array (4 items with icon/amount/text — mix of education/fitness/creative/relationship/sleep), apps_breakdown (one per app: name/time/intent good|mixed|bad/honest_note), reclaim_plan (4-5 items: specific action/why/time saved), if_reclaimed (2-3 sentences concretely painting reclaimed time future referencing what they'd rather be doing), reckoning (one italic-worthy sentence naming the cost).
Render: dark hero card with daily → yearly comparison (yearly is huge with glow). Equivalents 2-column grid with emoji icons. Apps list with intent-colored left borders and inline tag pills + honest italic note. Green-bordered reclaim plan card with numbered moves and time-saved pills. White "if reclaimed" future card with italic text. Centered final reckoning in quotation marks.What this does: ype your daily average, top 5 apps with times, pickup count, what you'd rather be doing, and how you feel about it. Get back the daily number translated into a yearly total (e.g. 88 days of waking hours), 4 real-world equivalents (degrees, marathons, books read), each app classified good/mixed/bad with an honest note on what you actually get from it, a ranked reclaim plan with concrete moves and time saved per move, a vision of what that reclaimed time could become, and a one-line italic reckoning. Saves to localStorage.
What this looks like:
AI Insights That Turn Your Data Into Bigger Profits
Your store is full of hidden growth opportunities. StoreClaw analyzes your Shopify and Amazon data, identifies the highest-impact actions, and helps you grow revenue while improving margins. Start free today with 10,000 bonus tokens. No credit card required.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Run a full lab experiment across a liquid handler, a robot arm and a plate reader, then throw out a bad result and rerun it without being told.
❌ Still Can't: Grasp the physical world well enough to know that bubbles in a sample are a physics problem, not a software bug. Genentech researchers had to explain that one.
✅ AI Can Now: Watch a base perimeter and tell a hostile drone from a bird in real time.
❌ Still Can't: Handle the last few meters. That still comes down to an Airman with a shotgun.
RECOMMENDED LISTENING/READING/WATCHING
The Terminal Man by Michael Crichton - Book
A 1972 techno-thriller about a brilliant but violent epileptic patient who agrees to an experimental brain implant designed to detect and interrupt his seizures with a small computerized electric shock, in the hope that it will also calm his rages. The device works exactly as designed, and that turns out to be the problem. Crichton(who also created Jurassic Park) wrote it as a doctor as much as a novelist, and it's one of the earliest mainstream thrillers to take the idea of an embedded, self-correcting machine intelligence seriously, decades before neural implants were a live research field.
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.
