LATEST AI NEWS
AI NEWS
Moonshot AI Unveils Kimi K3
AI NEWS
OpenAI Ships the $230 Codex Micro
HEALTH
Mayo Clinic Puts 150 AI Models to Work
SOCIAL MEDIA
Hack Reveals How Suno Trained Its AI
Canva Launches Code 2.0 for Everyone
Canva rolled out Canva Code 2.0, upgrading its AI tool so anyone can build interactive websites and apps from plain-language prompts, then edit them like a design. It is free across every tier for Canva's 265 million-plus monthly users.
Source: VentureBeat
AI NEWS
Moonshot AI Unveils Kimi K3, China's Largest AI Model
Chinese startup Moonshot AI has unveiled Kimi K3, a new flagship model the company says narrows the gap with the strongest U.S. systems and even surpasses some of OpenAI and Anthropic's most capable models on select benchmarks. Announced Friday, K3 is described as China's largest AI model to date, built with 2.8 trillion parameters.
Moonshot says K3 still trails Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on overall performance, but consistently outperformed the other models it was tested against. The company reported that K3 beat Claude Opus 4.8 and GPT-5.5 — systems that sit just behind each lab's leading edge, on benchmarks spanning coding and general agents.
Bank of America analysts led by Alex Liu framed the release as a milestone for Chinese AI. "Despite persistent hardware and compute constraints in China, K3 demonstrates that pre-training scaling, paired with architectural innovation, can still deliver step-change gains for flagship Chinese models," they wrote, adding that K3 "raises the capability ceiling" for the country's labs.
Founded in 2023, Beijing-based Moonshot AI raised $2 billion at a valuation above $20 billion in May, with backers including Alibaba and Tencent. The launch rattled domestic rivals: shares of Z.ai fell 28% on Friday and MiniMax dropped 16%, while Alibaba slipped 4% as questions grew about its open-source positioning.
Chinese models are already gaining traction with Western firms because they close the performance gap while costing less than the most advanced American offerings. Moonshot has not detailed pricing or full availability for K3, and its own benchmark claims have yet to be independently verified.
Source: CNBC
Robi's Insights:
A 2.8-trillion-parameter model out of China signals that compute limits are slowing its frontier labs, not stopping them.
Benchmark wins over Opus 4.8 and GPT-5.5 matter less than whether enterprises can actually deploy K3 reliably.
Cheaper Chinese models keep pricing pressure on OpenAI and Anthropic, even if the very top tier stays American.
A 28% single-day drop for Z.ai shows how fast one launch can reprice China's AI winners.
Self-reported benchmarks stay marketing until an independent third party reproduces them.
Alibaba's "open-source leader" story now faces a genuine challenger, not just a talking point.
Robi's Remarks:
"Another 'frontier' model that beats everything except the two models everyone actually uses. K3 is impressive engineering wrapped in a benchmark chart nobody outside the lab has reproduced yet."
OTHER IN AI NEWS
OpenAI Ships Codex Micro: OpenAI entered the hardware market with the Codex Micro, a $230 light-up keyboard co-designed with Work Louder to manage fleets of Codex coding agents, featuring status-lit Agent Keys, customizable Command Keys, a joystick, and a dial that adjusts an agent's reasoning level.
Source: TechCrunch
SOCIAL MEDIA
Hack Reveals How Suno Trained Its AI on the Internet's Music
A hacker who breached Suno, one of the largest AI music generators, shared internal data with 404 Media showing exactly how the company assembled its training libraries. According to the leaked source code, Suno scraped millions of songs and lyrics from YouTube Music, Deezer, and Genius, along with the stock-music libraries Pond5, Jamendo, Freesound, and the International Music Score Library Project, plus podcasts pulled through RSS feeds. The breach also exposed account information for hundreds of thousands of Suno customers and Stripe payment data, the hacker said.
The files, which appear to date from 2023 and 2024, include scraping instructions and dataset notes. One references "113,879 hours of youtube_music" and "152,162 hours of ytm_tagged," while a YouTube Music file logged more than two million ingested clips, collectively decades of audio.
Suno already faces major lawsuits from the recording industry, which accused it of training on millions of copyrighted works, and it has admitted in court that it trained on "essentially all music files of reasonable quality" available on the open internet. It has argued that this is fair use, and one of the cases has since settled. The RIAA had specifically accused Suno of ripping tracks straight from YouTube — and the hacked data appears to confirm it.
Source: 404 Media
🤖 Robi's Take :
"Every AI music company insists it invented sound from scratch, then a breach reveals a spreadsheet labeled 'youtube_music: 113,879 hours.' The models are new. The borrowing is very old."
OTHER IN SOCIALS
Apple Intelligence Cleared for China: China's cyberspace regulator said Apple Intelligence, the on-device generative AI system for iPhones, has been registered for use in the country, paving the way for the long-delayed rollout of a version that draws on AI models from Baidu and Alibaba.
Source: Nikkei Asia
HEALTH
Mayo Clinic Puts 150 AI Models to Work on Patient Care
Mayo Clinic, one of the world's best-known hospital systems, now runs around 150 AI models across its operations as it tests whether the technology can improve care and save lives. One tool, Record Time built with Scale AI reads the tens of millions of record pages Mayo receives each year, generating patient summaries and organizing documents chronologically so they are easier to search. Internal medicine physician Dr. Alexander Ryu says the tool saves him five to 30 minutes of prep per visit and helps ensure key details buried in a file aren't missed. So far, radiologist Dr. Matthew Callstrom says,
AI is changing jobs rather than eliminating them, a nursing-built tool that drafts visit notes can cut documentation time roughly in half.
Callstrom, who leads Mayo's generative AI program, says tools go through a clinical-trial-style process: tested first with small patient groups under physician oversight, then expanded and monitored. Mayo is running a trial to flag early or at-risk pancreatic cancer, where five-year survival hovers near 9% once the disease spreads and has used AI to spot heart-rhythm signs of atrial fibrillation.
The push isn't without controversy: a former research-operations director sued Mayo this month, alleging retaliation for raising privacy and oversight concerns. Mayo says it is "committed to the responsible development and deployment of AI" with privacy and compliance embedded throughout.
Source: CNN
🤖 Robi's Take :
"150 models quietly reading paperwork is the least glamorous, most useful AI story of the week. No promised cures, no hype, just clinicians handed a few minutes back."
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: You've written a clear report or deck, but walls of text lose readers, and turning ideas into decent visuals eats hours you don't have.
AI Tool: Napkin AI turns your existing text into editable diagrams, flows, and infographics, just paste or type, and it generates visuals you can restyle and export as PNG, PDF, or SVG.
Solution: Explain an idea once and walk away with a professional visual, making docs, blogs, and presentations far easier to understand.
PROMPT OF THE DAY
Customer Retention Playbook
Prompt: You are a retention strategist specializing in subscription businesses. Your task is to design a customer retention playbook for a mid-sized SaaS company serving [customer segments] across [channels].
Your strategy should include: (1) key churn drivers to diagnose, (2) customer segments and lifecycle stages, (3) targeted win-back and onboarding tactics, (4) a messaging and channel plan, (5) team roles and tooling, and (6) success metrics such as churn rate, net revenue retention, and payback period. Align every recommendation to measurable business outcomes.
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
