LATEST AI NEWS
AI NEWS
Claude Learns Skills by Watching Your Screen
AI NEWS
Google Adds Selfie Video Sign-In
HEALTH
ChatGPT Opens a Dedicated Health Hub
SOCIAL MEDIA
Old Books Become Prime AI Training Data
Meta Tests StoryKit, an AI Bedtime-Story App
Meta is piloting StoryKit, an iOS app that generates personalized children’s stories — with custom characters, settings, lessons, and music — from a few prompts. Its App Store pitch: “You don’t need to write a single word.” Meta says it uses AI safety filters and has no social features.
Source: TechCrunch
AI NEWS
Claude Learns Skills by Watching Your Screen
Anthropic has given Claude a new way to pick up tedious tasks: instead of writing out a prompt, subscribers can now show the assistant what to do. A “Record a skill” option inside Claude’s Cowork tools starts a short screen share, and Claude watches where you click, how you type, and any voiceover you add as you walk through a process.
When the recording ends, Claude processes the demonstration, works out how to repeat the steps, and saves it as what Anthropic calls “a repeatable skill.” The idea is that you no longer have to describe a multi-step workflow in writing — you perform it once and hand the playback to the model, which turns your actions into an automation you can trigger later.
Anthropic aims the feature squarely at repetitive work. If you run a regular report that pulls figures from several sources, you can show Claude where each piece of data lives and how you want it presented; the system returns a skill you can switch on to regenerate that report on demand, without re-explaining the process each time.
The option reached Max, Pro, and Team subscribers soon after Anthropic announced it, and appears under Cowork > Record a skill > Start recording. There is no sign it is coming to free accounts, and Anthropic has not said how often each plan can record. Alongside it, the company added an iOS app-testing feature to the Claude Code Mac app, letting developers emulate iPhone and iPad apps in a virtual simulator.
Source: Yahoo Tech
Robi’s Insights:
Show-don’t-tell lowers the barrier: staff who could never write a clean prompt can still record a working one.
The quiet win is institutional memory — a recorded workflow outlives the one employee who knew how it ran.
Locking this to paid tiers tells you Anthropic sees automation, not chat, as where the revenue is.
“Watches your clicks and typing” is also a data-capture story; check what’s on screen before you hit record.
A skill is only as good as its demo — record a sloppy process and you’ve faithfully automated the mess.
Anything you dread doing on a monthly cadence is the obvious first candidate to hand over.
Robi’s Remarks:
“Twenty years of ‘please write proper documentation,’ and the winning move was letting people mumble through a screen share. Turns out nobody wanted to write the manual — they wanted the manual to write itself.”
OTHER IN AI NEWS
Google Adds Selfie Video Sign-In: Google is rolling out a backup way to get into a locked Google Account by recording a short selfie video with guided head movements, which it matches against a stored clip; the video is encrypted, can be deleted anytime, and uses liveness checks to help block deepfakes and impersonation.
Source: Google
SOCIAL MEDIA
Old Books Become Prime AI Training Data
As AI developers run short of fresh, high-quality text, a data broker called ISBNdb is pitching old printed books as the premium source precisely because they predate the AI “slop” now flooding the open web. “The world’s best AI training data is sitting on a shelf,” the company says, describing books as curated, peer-reviewed, domain-specific human knowledge that is “dense, edited, authoritative,” and claiming the world’s largest book database. ISBNdb offers high-volume book acquisition, sourcing physical books in bulk for AI labs to scan into training sets. Its logic is simple: titles published before 2022 are effectively guaranteed to contain no machine-generated text,
unlike a web crawl that increasingly ingests other models’ output. The pitch lands at a moment when training data has become a bottleneck. The easiest text to scrape has largely been used, licensing deals are expensive, and a rising share of the web is itself AI-written raising the risk that new models learn from the output of older ones. Print, scanned page by page, offers a large pool of human writing that sidesteps that contamination.
404 Media reported that ISBNdb even tells clients “the optics problem is real” an admission that hoovering up print to feed AI carries reputational and potential copyright costs. The scramble reflects a deeper anxiety: as synthetic text piles up online, models risk training on their own exhaust, and clean, human-written material becomes a scarce commodity worth buying by the pallet.
Source: 404 Media
🤖 Robi’s Take :
“The industry swore the open web was all it needed, and now it’s buying dusty hardcovers by the pallet to escape the slop it manufactured. Nothing says progress like paying extra for pre-AI reality.”
OTHER IN SOCIALS
White House Accuses Moonshot AI of Copying Anthropic: Michael Kratsios, who leads the White House Office of Science and Technology Policy, publicly accused China’s Moonshot AI of running large-scale “distillation” against Anthropic’s Fable model to build its Kimi K3, alleging a covert platform switched between access methods to avoid detection a claim Moonshot has not answered.
Source: CyberScoop
HEALTH
ChatGPT Opens a Dedicated Health Hub
OpenAI is rolling out Health in ChatGPT to logged-in U.S. users aged 18 and older, on web and iOS across the Free, Go, Plus, and Pro plans. With permission, users can connect Apple Health and supported medical records so the assistant can reference them inside ordinary conversations.
OpenAI says more than 300 million people a week already ask ChatGPT health questions, but that context is usually scattered across patient portals, apps, and wearables. Health can compare a new lab result with prior tests, summarize what changed since your last appointment, or relate sleep and activity to your routine.
Health lives in the left sidebar as a home base for connecting data, viewing recent trends, browsing synced records, and returning to past health chats. Connections include U.S. hospital systems, One Medical, and Function Health.
Privacy is the headline pitch. By default ChatGPT asks before using connected records; that data is encrypted, is not used to train OpenAI’s models or target ads, and any source can be disconnected, with synced data deleted within 30 days.
OpenAI calls GPT-5.6 Sol its strongest health model yet and says hundreds of physicians helped build its evaluations while stressing that ChatGPT can still make mistakes and does not replace professional care. Health is U.S.-only at launch and not yet available in Codex.
Source: OpenAI
🤖 Robi’s Take :
“Strapping your medical records to a chatbot that ‘can still make mistakes’ is a bold trust exercise. The privacy promises read well on paper; the incentives around health data rarely hold that still.”
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: You record a solid podcast, voice note, or video, then lose hours cutting out “ums,” background hiss, mouth clicks, and dead air.
AI Tool: Cleanvoice AI scans an audio or video track and automatically removes filler words, background noise, long silences, and mouth sounds and can transcribe and summarize it too.
Solution: Upload a raw recording and get back a cleaner, tighter cut in minutes, so you spend your time creating instead of editing.
PROMPT OF THE DAY
The Repeatable-Workflow Architect
Prompt: You are an operations strategist specializing in workflow automation. Your task is to design a repeatable reporting workflow for a small marketing team.
Your plan should include: (1) the data sources to pull from, (2) the exact steps in order, (3) where a human must review, (4) the output format and where it lands, (5) failure and edge-case handling, and (6) measurable success criteria such as time saved per run and error rate. Align each step to how the team already works so adoption feels effortless.
SPOT THE FAKE
Can you outsmart AI?
We’ve got a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.
Click option below A or B. 👇
A
B
BEFORE YOU GO
Ready to take your AI journey further?
For the latest AI updates and insights Click Here 👉 Explore our Weekly Newsletter
For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel
AT THE END
Craving more AI chaos?
That's it for today!
Read Daily AI News at BitBiased.AI. Support us by following us on LinkedIn and X ( Twitter ).
Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team
