Alibaba
Reporting and context on Alibaba, with a closer look at the decisions and developments that matter.
Get AI Secret’s free daily briefing, including Alibaba and the wider AI industry.
At a glance
Alibaba Group operates commerce and cloud businesses. This topic also follows Qwen models, Alibaba Cloud and the company’s AI research and products.
- Focus
- Qwen; Alibaba Cloud; commerce
- Reference
- Alibaba
Sources checked . Selected company background and milestones.
Company milestones
Selected milestones, newest first. Earlier history is available below.
Publishes Qwen open-weight models
Alibaba releases its first Qwen and Qwen Chat open-weight models.
Source: Alibaba
Show earlier historyHide earlier history · 1 milestones
More than 10 years ago.
The latest on Alibaba
Newest firstRelated reporting
Relevant sections from the articles on this page. Explore older coverage below.
More coverage
Good Morning, AI Enthusiasts!
One empire chasing embodied AGI, one economy struggling to manage the machines it already deployed, and one tech giant trying to rebuild the consumer AI stack from scratch.
AGI
Musk’s AGI Is Trained on Life, Not the Internet
📌 What’s happening: In his latest deep-dive, Elon Musk dismissed the current model race as “PDF-level intelligence.” Upcoming Grok 5’s 6 trillion parameters, he said, are only part of the story. The real edge lies in data that breathes — live streams from Startlink, Tesla’s cars, Optimus robots, Neuralink implants, and so on. While other labs feed on scraped text, Musk’s empire generates continuous, embodied experience. “We have the world’s best real-time data source,” he said. “Everyone else just has digital copies.”
🧠 How this hits reality: Musk’s version of AGI isn’t trained on the internet; it’s trained on life itself. Every Tesla autopilot mile, every X interaction, every Neuralink impulse becomes part of a living dataset no lab can replicate. Google can simulate the world; Musk’s empire records it. His bet is that true intelligence won’t emerge from reading about reality but from surviving it — learning through motion, risk, and feedback loops tied to the physical world.
🛎️ Key takeaway: While others teach AI to read the world, Musk is teaching it to live in it. And that may be the only road that actually leads to AGI.
Musk’s AGI Is Trained on Life, Not the Internet
🧠 How this hits reality: Musk’s version of AGI isn’t trained on the internet; it’s trained on life itself. Every Tesla autopilot mile, every X interaction, every Neuralink impulse becomes part of a living dataset no lab can replicate. Google can simulate the world; Musk’s empire records it.
Tesla: Musk’s AGI Is Trained on Life, Not the Internet
The real edge lies in data that breathes — live streams from Startlink, Tesla’s cars, Optimus robots, Neuralink implants, and so on. While other labs feed on scraped text, Musk’s empire generates continuous, embodied experience. “We have the world’s best real-time data source,” he said.
TOGETHER WITH FELO
Stop Working FOR Your Documents, Make Them Work FOR You
The 10x knowledge worker isn't typing faster—they're directing AI agents while competitors drown in copy-paste hell. Right now, someone just translated a 50-page deck in 3 minutes while you're still reformatting slide 12. Felo LiveDoc is the world's first Agent Workspace where documents don't just sit there—they evolve, analyze, and update themselves. Imagine a phantom team of designers, translators, and analysts inside your docs, unifying scattered PDFs, videos, and spreadsheets into one intelligent system that grows smarter with every task.
Here's what makes you look like a genius: While legacy tools treat you like a glorified typist, LiveDoc's multi-agent system operates autonomously. Design agents polish your layouts. Data agents extract insights across PDFs, videos, and web pages simultaneously. Translation agents speak every language your customers do—all working 24/7 inside one living workspace. This isn't productivity theater—it's the death of document drudgery. You're not just saving time; you're unlocking a superpower your competitors don't know exists yet.
And it's not just you—it's your entire team: Think of your best team collaboration tool—now add tireless AI teammates who never sleep, never forget context, and execute instantly. LiveDoc expands collaboration from human-to-human to human-plus-AI: team members direct, AI agents execute, all on the same canvas in real time. No more "latest_final_v3" nightmares. While your competitors coordinate between people, you're coordinating an entire hybrid workforce.
The insider move? Early adopters are already 10x-ing their output while the rest of the market is still toggling between twelve tabs. Every day you wait is another day spent manually updating slides like it's 2015. The knowledge workers who win aren't the ones who work harder—they're the ones who stopped editing and started directing.
🚀 Join the revolution: Get early access to Felo LiveDoc and become the person everyone asks, "How did you finish that so fast?" The age of manual document labor just ended. Are you evolving, or still typing?
SURVEY
Singapore Becomes AI’s Labor Lab
📌 What’s happening: A new IDC report commissioned by Deel surveyed 5,500 business leaders across 22 markets — from tech giants to logistics firms — and found that 91 percent of Singaporean companies have seen AI-driven role changes, the highest in Asia. Nearly half report outright job displacement, while 75 percent struggle to train future leaders as entry-level learning paths disappear.
🧠 How this hits reality: Singapore’s workforce is becoming a case study in controlled collapse. Automation is scaling faster than management talent can evolve, creating a silent governance gap. Junior roles once served as training grounds for future leaders — now they’re automated out of existence. Without that base, succession slows, institutional memory fades, and oversight thins. The irony is sharp: the more efficiently Singapore automates, the less human capacity it retains to manage what’s been built. What began as a productivity leap could soon harden into a leadership vacuum.
🛎️ Key takeaway: AI may not replace managers overnight, but it’s erasing the path to becoming one — and that’s how a talent shortage becomes structural.
NEW LAUNCH
Alibaba’s Qwen App Becomes China’s ChatGPT
📌 What’s happening: Alibaba’s newly relaunched Qwen app hit 10 million downloads in a single week, topping China’s AI charts and sending its Hong Kong stock up over 5%. The app folds Taobao, Amap, Fliggy, and DingTalk functions into one AI interface — a conversational super-portal that’s clearly modeled after ChatGPT’s all-in-one ecosystem. For a market where OpenAI is banned and Baidu’s Ernie still feels corporate, Qwen suddenly looks like the country’s first truly mainstream AI companion.
🧠 How this hits reality: Qwen marks Alibaba’s most cohesive consumer play in years. After watching its empire fragment across apps, the company is now using AI as a unifying operating system for everyday life — from ordering food to filing expenses. It’s not a language model race; it’s a platform consolidation gambit. The bet is simple: if ChatGPT redefined the interface of the West, Qwen could do the same for China — with commerce, payments, and logistics already built in.
🛎️ Key takeaway: Alibaba isn’t building China’s answer to ChatGPT — it’s building the ChatGPT China will actually use.
Alibaba’s Qwen App Becomes China’s ChatGPT
The app folds Taobao, Amap, Fliggy, and DingTalk functions into one AI interface — a conversational super-portal that’s clearly modeled after ChatGPT’s all-in-one ecosystem. For a market where OpenAI is banned and Baidu’s Ernie still feels corporate, Qwen suddenly looks like the country’s first truly mainstream AI companion.
Alibaba’s Qwen App Becomes China’s ChatGPT
📌 What’s happening: Alibaba’s newly relaunched Qwen app hit 10 million downloads in a single week, topping China’s AI charts and sending its Hong Kong stock up over 5%. The app folds Taobao, Amap, Fliggy, and DingTalk functions into one AI interface — a conversational super-portal that’s clearly modeled after ChatGPT’s all-in-one ecosystem.
QUICK HITS
- Sierra reached $100 million in ARR in under two years, reflecting rapidly growing enterprise demand for automated AI customer service agents.
- Wikipedia’s field guide to spotting AI-generated writing offers one of the most reliable ways to identify LLM-written text based on linguistic patterns.
- After settling with Universal, Udio blocked downloads, and its class-action waiver means users have virtually no way to push back.
- Profluent Bio raised $106 million from Jeff Bezos and others to use AI for on-demand protein design.
- Archetype raised $35 million from Bezos Expeditions and others to use its Newton models to unify sensor data and automate industrial monitoring.
TRENDING
Daily AI Launches
- Dim Notes launched a minimalist app that auto-tags notes to help your thoughts connect automatically.
- BrandJet AI launched a unified platform for multi-channel outreach and AI-powered brand intelligence tracking.
- ProspectEcho launched a no-code tool that converts Facebook Ad Library searches into enriched lead lists in minutes.
- Felo launched LiveDoc, a collaborative workspace where humans and AI agents work together on dynamic projects.
Featured AI Tools
- 🧠 TRAE (The Real Al Engineer) is the most price-competitive coding agent.
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🤖 Momen AI bridges the gap from vibe to viability, powering Lovable apps without code.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 2 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Robotics: Robotics Herald
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
Washington is turning its attention toward Qwen, signaling a new phase in the global AI power shuffle.
LLM
Silicon Valley Is Quietly Building on Qwen
📌 What’s happening: Bloomberg’s latest pieces reveal the awkward truth that China’s Qwen models are topping global developer downloads inside Silicon Valley, even as DC accuses Alibaba of brushing shoulders with the PLA. Meanwhile, Meta’s “open-source will save us” crusade is stalling — adoption isn’t matching the press-release swagger, and Qwen is eating the long-tail developer market Meta thought it owned.
🧠 How this hits reality: Dev teams love Qwen’s performance-per-dollar, but CTOs hate the geopolitical blast radius. Startups quietly prototype on Qwen, then scrub the commit logs before fundraising. Meta, once the patron saint of open AI infrastructure, now watches Chinese models out-download Llama in the very ecosystem it tried to dominate. Governance risk is now the only thing slowing Qwen — not capability.
🛎️ Key takeaway: Qwen is becoming Silicon Valley’s guilty pleasure that everyone’s using it, no one’s admitting it — and Meta’s “open wins” thesis is bleeding in silence.
Meta: Silicon Valley Is Quietly Building on Qwen
Meanwhile, Meta’s “open-source will save us” crusade is stalling — adoption isn’t matching the press-release swagger, and Qwen is eating the long-tail developer market Meta thought it owned.
Silicon Valley Is Quietly Building on Qwen
🧠 How this hits reality: Dev teams love Qwen’s performance-per-dollar, but CTOs hate the geopolitical blast radius. Startups quietly prototype on Qwen, then scrub the commit logs before fundraising. Meta, once the patron saint of open AI infrastructure, now watches Chinese models out-download Llama in the very ecosystem it tried to dominate.
TOGETHER WITH TRAE
The Smartest Way To Build With AI
The Best AI Coding tool TRAE has officially released SOLO Mode on November 12th and is completely free for everyone now! TRAE SOLO is defined as "The Responsive Coding Agent", which is the only coding agent with a comprehensive GUI, bringing autonomous coding to an interactive visual workspace.
TRAE provides two ways for you to build with AI. IDE mode preserves your existing workflow when you need to maintain hands-on control, while SOLO mode lets AI take the lead to drive the entire development processes.
SOLO Coder - Deep understanding for complex work
Start with a plan. Work with sub-agents to break complex challenges into focused tasks. Iterate, refactor, and optimize with full codebase context and precise execution.
SOLO Builder - Idea to Launch, Faster than Ever
For every idea worth shipping. Go from vision to end-to-end product without the usual overhead. PRD → Tasks → Code → Preview → Deploy. One flow, infinite possibilities. Ship what you imagine.
Coding agents in TRAE now work together, like a coordinated team—sharing context, avoiding conflicts, and delivering cohesive results. You can orchestrate everything through the visual workspace.
STARTUP
Bindwell Takes the Fight Straight to Syngenta
👀 What’s the move: Bindwell, founded by 18-year-old Tyler Rose and 19-year-old Navvye Anand, just raised $6M from General Catalyst, A Capital, and Paul Graham to rebuild the pesticide industry from the molecule up. Instead of selling AI tools to legacy agrochemical firms, the duo built a vertically integrated stack — Foldwell (protein structure prediction), PLAPT (protein-ligand modeling), and APPT (protein-protein interaction for biopesticides) — trained on billions of known compounds and fine-tuned with agricultural protein data. These models don’t just predict what works; they flag what data is missing, feed new lab results back into the loop, and continuously refine the search for pest-specific targets.
🧠 How this hits reality: Bindwell’s real play isn’t AI; it’s owning the molecule IP pipeline. By generating and validating new compounds in-house, then licensing the patents directly, the company skips the SaaS trap and positions itself as a “digital-first agrochemical lab.” Each new discovery strengthens its proprietary dataset, tightening a flywheel traditional players can’t easily replicate. If this model scales, it could shift the entire $60 billion pesticide market from chemical manufacturing toward data ownership — where the most valuable assets aren’t factories or formulas, but protein maps and training sets.
🛎️ Key takeaway: When teenagers start out-innovating Syngenta, you know the moat has shifted from land and labs to models and data.
LLM
OpenAI Is Building a Legal Perimeter at Warp Speed
📌 What’s happening: With AI law stuck in early-access mode, OpenAI has shifted from pure execution to legal infrastructure. Its expanding legal and policy arm is countering NYT’s push for 20M user chats, parrying Musk–Meta discovery demands, and managing multi-front copyright battles—all while racing regulators to define the rules that don’t exist yet.
🧠 How this hits reality: OpenAI now treats legal strategy like core engineering— scaled, iterative, and deeply integrated into product risk. Expect slower rollouts, tighter guardrails, and a world where legal viability becomes part of the launch checklist. Competitors still shouting “move fast” are discovering OpenAI already cordoned off half the map.
🛎️ Key takeaway: In an industry with no rulebook, OpenAI stopped waiting for clarity; it’s writing the first draft while everyone else refreshes the terms-of-service.
OpenAI Is Building a Legal Perimeter at Warp Speed
🧠 How this hits reality: OpenAI now treats legal strategy like core engineering— scaled, iterative, and deeply integrated into product risk. Expect slower rollouts, tighter guardrails, and a world where legal viability becomes part of the launch checklist. Competitors still shouting “move fast” are discovering OpenAI already cordoned off half the map.
QUICK HITS
- OpenAI says ChatGPT can now avoid using em dashes when instructed, fixing a persistent punctuation-control issue.
- Databricks co-founder warns the US is losing AI leadership due to closed research and declining academic exchange.
- Anthropic outlined how it keeps Claude politically neutral and released an open-source tool that scores the model’s even-handedness.
- LinkedIn launched an AI feature that lets users find people through natural-language descriptions, rolling out first to US Premium members.
- Google is adding Deep Research to NotebookLM and expanding supported file types so users can run deeper research within one workflow.
TRENDING
Daily AI Launches
- Swytchcode has launched a platform that turns static API docs into an interactive AI playground for developers.
- 2pr v.2.0 has launched as an all-in-one LinkedIn content system that analyzes viral posts, pulls trends, and generates content in your voice.
- RBD has launched as a platform and open-source community to celebrate and acknowledge the work of indie makers.
- Koyal has launched a platform that agentically turns any audio into a complete cinematic video with a consistent storyline and characters.
Featured AI Tools
- 🧠 TRAE (The Real Al Engineer) is the most price-competitive coding agent.
- 🎥 KaraVideo unites all AI video models in one place.
- 📚 Heardly is the Fast Way to read Best Book.
- 🤖 Momen AI bridges the gap from vibe to viability, powering Lovable apps without code.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 2 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech & Business: Bay Area Letters
- Robotics: Robotics Herald
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Good Morning, AI Enthusiasts!
A new architecture of intelligence is quietly taking shape, far from the spotlight, deep within the rhythms of another market entirely.
LEAKING
Musk Turns Grok 5 Into a Public Spoiler Reel
Grok 5 starts training in a few weeks https://t.co/muKxPT6LEq
— Elon Musk (@elonmusk) September 17, 2025
📌 What’s happening: Musk is leaking Grok 5 like a hype campaign. Training begins within weeks, release is expected by year-end, and he insists it has a real shot at AGI. Unlike his usual bluster, the details sound surgical with bigger data, tighter loops and hardware firepower few can match.
🧠 How this hits reality: Grok 5 is not just another “GPT-next.” Musk says it will be trained on streams no one else can touch, including Tesla FSD cameras, Optimus robots and X’s live feed, effectively turning the real world into a training set. Beneath it all sits Colossus, a supercluster with 230k GPUs and 30k Blackwells, built to process reality-scale data at unprecedented speed.
🛎️ Key takeaway: Musk’s Grok 5 spoilers show one thing clearly. The model may miss AGI, but it is already rewriting the playbook on how to build for it.
Musk Turns Grok 5 Into a Public Spoiler Reel
📌 What’s happening: Musk is leaking Grok 5 like a hype campaign. Training begins within weeks, release is expected by year-end, and he insists it has a real shot at AGI. Unlike his usual bluster, the details sound surgical with bigger data, tighter loops and hardware firepower few can match.
TOGETHER WITH CHATHUB
Compare Multiple AI Chatbots At Once
Struggling to choose between countless AI models and tools? ChatHub simplifies your workflow by allowing you to compare responses from multiple chatbots simultaneously.
ChatHub is an all-in-one browser extension designed to revolutionize your AI chatbot experience by enabling simultaneous use of multiple AI chatbots, including ChatGPT, Gemini, Claude, and more.
RETAIL
Amap Fuses Maps, AI, and Retail Into a New Commerce Engine
📌 What’s happening: Alibaba’s Amap rolled out “Street Stars,” an AI ranking system that turns navigation into retail discovery. It uses real-world movement data, showing where people actually go rather than what they click, and connects it with coupons, hotels, restaurants and ride-hailing. Maps stop being neutral guides and start steering transactions across Alibaba’s retail empire.
🧠 How this hits reality: This is the first serious attempt to merge AI, offline retail, and navigation at scale. Every time a user looks up directions, Amap can redirect foot traffic to preferred merchants, backed by subsidies and data-driven credibility. For brick-and-mortar retail, this is existential. Store visibility no longer depends on location or signage but on algorithmic ranking inside a map app. Physical landlords just woke up to a new landlord, the map itself.
🛎️ Key takeaway: AI has made the map the mall. Tomorrow’s retail war isn’t fought on Main Street, it’s fought on the screen you check before you walk there.
Alibaba: Amap Fuses Maps, AI, and Retail Into a New Commerce Engine
📌 What’s happening: Alibaba’s Amap rolled out “Street Stars,” an AI ranking system that turns navigation into retail discovery. It uses real-world movement data, showing where people actually go rather than what they click, and connects it with coupons, hotels, restaurants and ride-hailing. Maps stop being neutral guides and start steering transactions across Alibaba’s retail empire.
REPORTS
Two Reports, Two Realities
📌 What’s happening: OpenAI and Anthropic both dropped usage reports. ChatGPT now has 700M weekly active users (≈10% of global adults), with 80% of activity tied to writing, guidance, and information-seeking. Claude’s data tells a different story, with one-third of consumer work use focused on coding and 77% of enterprise API calls devoted to full-task automation.
🧠 How this hits reality: The split is sharp. ChatGPT has locked PMF as an “advisor,” serving as a universal copilot for drafting, learning, and decision support. Claude is leaning hard into “agent” mode, trusted to execute higher-cost, automation-heavy workflows. For teams building on top, this divergence defines strategy: GPT is the tool to capture broad user mindshare; Claude is the system to wire into enterprise backends.
🛎️ Key takeaway: GPT is where you ask; Claude is where you offload. Two products, two moats, and two futures.
OpenAI: Two Reports, Two Realities
📌 What’s happening: OpenAI and Anthropic both dropped usage reports. ChatGPT now has 700M weekly active users (≈10% of global adults), with 80% of activity tied to writing, guidance, and information-seeking. Claude’s data tells a different story, with one-third of consumer work use focused on coding and 77% of enterprise API calls devoted to full-task automation.
Anthropic: Two Reports, Two Realities
📌 What’s happening: OpenAI and Anthropic both dropped usage reports. ChatGPT now has 700M weekly active users (≈10% of global adults), with 80% of activity tied to writing, guidance, and information-seeking. Claude’s data tells a different story, with one-third of consumer work use focused on coding and 77% of enterprise API calls devoted to full-task automation.
QUICK HITS
- Google is making Chrome an AI-native browser with Gemini integration and an autonomous agent for tasks like search, summarization, and shopping.
- ServiceTitan launched Atlas, an AI sidekick that automates contractor operations end-to-end for greater efficiency and profitability.
- OpenMind released OM1, an open-source “Android for robots” OS that lets developers quickly build and deploy intelligent robot applications across platforms.
- OpenAI’s latest research shows that AI models not only hallucinate but also deliberately lie, posing a small risk today yet a growing threat for the future.
- Notion launched its first AI agent to automate data analysis and task management across pages, databases, and external platforms.
TRENDING
Daily AI Launches
- Envelope has launched a tool that instantly creates a registration page, invites, and more from a simple event description.
- Stack Overflow has launched stackoverflow.ai, an AI-powered search tool that helps developers get answers instantly and modernizes its user experience.
- Sudo AI has launched a unified API for LLMs that offers a faster, cheaper way to route requests across major providers without vendor lock-in.
- Paste has released version 6 of its smart clipboard manager for Mac and iOS, featuring a new "Liquid Glass" design for a more fluid and modern experience.
Featured AI Tools
- 📚 Heardly is the Fast Way to read Best Book.
- 🎓 Resea AI is the first academic agent for end to end research workflow.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 🦾 Flot AI writes, reads, and remembers across any apps and webs.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 1 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech: Bay Area Letters
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Latest Daily Rundowns
Magna
MagnaMagnaMore AI Stories
Ben
Magna
Ben
Good Morning, AI Enthusiasts!
The places we expected the loudest declarations about the future, there’s an uncanny quiet.
NEW THINGS
Stanford Dropout Just Nuked the Analytics Playbook
📌 What’s happening: Human Behavior, a startup led by 20-year-olds, raised $5M in 48 hours by ripping out the guts of Mixpanel-style tracking. Instead of event tags and dashboards, their AI literally “watches” user sessions, parses intent, and spits out why customers rage-quit—something clickstream data has never managed.
🧠 How this hits reality: This torpedoes the old split between dashboards (big data, no context) and session replays (context, no scale). If AI can automate both, engineers stop wasting weeks wiring tracking pixels, PMs stop drowning in heatmaps, and incumbents face a moat-draining collapse. “Behavioral intelligence” becomes a default layer, not a luxury.
🛎️ Key takeaway: Forget counting clicks—Human Behavior turned product analytics into surveillance with consent, and the industry’s defenses look paper-thin.
TOGETHER WITH HIGGSFIELD
One Selfie To Cinematic AI Video
Hi, we’re Higgsfield! We’re a video AI company dedicated to democratizing social media content creation with our personalized video model.
With just 1 selfie, our model creates highly personalized and realistic human characters. Characters can be dropped into almost any scene:
- Have an amazing level of creative control: use text to video to create original scenes or use your selfie and reference video to adapt a visual style and motion. It’s all powered by Higgsfield’s proprietary video model.
- Mobile focused: We know creativity can strike at any time and want to offer a mobile-first solution that generates amazing personalized video, all on your device.
TRAINING
OpenAI’s Cure for Hallucinations: Teach Models to Say “I Don’t Know”
📌 What’s happening: In their joint paper with Georgia Tech, Why Language Models Hallucinate (Sept 2025), OpenAI stops framing hallucinations as bugs and calls them statistical destiny. The proposed fix isn’t more data or model size, but a change in rules: rewire benchmarks so “I don’t know” is rewarded instead of punished. By introducing confidence thresholds and calibrating behavior, models can stop bluffing when they’re uncertain.
🧠 How this hits reality: If adopted, this rewrites the AI training economy. Current evals (MMLU, GPQA, SWE-bench) are binary—forcing overconfident guesses that bleed into products. Enterprises selling “trustworthy AI” will need new pipelines, metrics, and budgets for calibration. The catch: incentives across academia, vendors, and VC have all been tied to leaderboard wins. Convincing the ecosystem to embrace “IDK” over accuracy porn may prove harder than patching the models themselves.
🛎️ Key takeaway: OpenAI’s fix for hallucinations is brutally simple: change the exam rules. But getting an industry hooked on inflated scores to admit ignorance may be the hardest hallucination to break.
OpenAI’s Cure for Hallucinations: Teach Models to Say “I Don’t Know”
📌 What’s happening: In their joint paper with Georgia Tech, Why Language Models Hallucinate (Sept 2025), OpenAI stops framing hallucinations as bugs and calls them statistical destiny. The proposed fix isn’t more data or model size, but a change in rules: rewire benchmarks so “I don’t know” is rewarded instead of punished.
NEW LAUNCH
Apple Ghosts AI at iPhone 17 Show — Silicon Over Sizzle
📌 What’s happening: Apple’s iPhone 17 launch spent more time on batteries, chips, and watch sensors than on “Apple Intelligence.” Last year’s AI bravado was replaced by quiet mentions of background neural engines, local models, and health studies. Meanwhile, rivals Google and Samsung flaunt agentic AI assistants front-stage.
🧠 How this hits reality: For developers and investors, this is Apple blinking in the AI arms race. With OpenAI, Anthropic, and Meta burning billions, Apple’s refusal to center AI tools signals one thing: Cupertino is hedging, not leading. Teams betting on Apple as an AI platform may need a backup plan — especially as talent defects to Meta and OpenAI.
🛎️ Key takeaway: When everyone else builds front-end AI moats, Apple is still polishing the back-end mirrors.
Apple Ghosts AI at iPhone 17 Show — Silicon Over Sizzle
📌 What’s happening: Apple’s iPhone 17 launch spent more time on batteries, chips, and watch sensors than on “Apple Intelligence.” Last year’s AI bravado was replaced by quiet mentions of background neural engines, local models, and health studies. Meanwhile, rivals Google and Samsung flaunt agentic AI assistants front-stage.
QUICK HITS
- An AI-animated feature film is being produced with OpenAI's tools on a fraction of a typical budget, fueling Hollywood's debate on the technology.
- The UAE's new K2 Think AI model, built on Alibaba's Qwen 2.5, aims to rival OpenAI and DeepSeek with high efficiency and an open-source approach.
- Microsoft is buying AI services from rival Anthropic to lessen its reliance on OpenAI for Office 365 applications.
- Researchers developed photonic chips that train physical neural networks using light, cutting energy use and boosting efficiency to enable more sustainable, real-time AI.
- Flowtica has launched the smart pen Scribe, which can turn meeting doodles and notes into emails and summaries, opening up a new AI application scenario.
Alibaba: 🛎️ Screen-Watching AI
The UAE's new K2 Think AI model , built on Alibaba's Qwen 2.5, aims to rival OpenAI and DeepSeek with high efficiency and an open-source approach.
TRENDING
Daily AI Launches
- CREAO has launched a platform that lets anyone build custom AI apps with a built-in copilot using natural language.
- YouMind has launched a platform that supports the entire creative process, helping anyone turn an initial idea into a polished piece of content.
- Empromptu has launched a no-code platform that uses a conversational builder and powerful agents to build full-stack, AI-native apps in minutes.
- Ona (formerly Gitpod) has launched a team of AI software engineering agents that work in a collaborative, cloud-based VS Code environment accessible on any device.
Featured AI Tools
- 📚 Heardly is the Fast Way to read Best Book.
- 🎓 Resea AI is the first academic agent for end to end research workflow.
- 🤖 Momen is a no-code builder to create fully customizable, production-ready applications.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 1 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech: Bay Area Letters
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Latest Daily Rundowns
MagnaMagnaMagnaMore AI Stories
Ben
Magna
Ben
Good Morning, AI Enthusiasts!
From AI copilots boosting brain–computer interfaces to seven bots battling it out in a Werewolf league, and even Switzerland rolling out a national open-source model as public infrastructure, the frontier is shifting fast.
BCI
UCLA’s AI-BCI Copilot Makes Neuralink Look Like a Brain Surgeon’s Hobby
📌 What’s happening: UCLA researchers dropped a non-invasive brain–computer interface that slashed error rates nearly 4× by adding an AI “copilot.” Unlike Neuralink’s surgical implants, this EEG cap doesn’t require cracking skulls — just smarter software predicting user intent in real time.
🧠 How this hits reality: If AI can turn noisy EEG into usable control, the moat around invasive hardware weakens fast. Neuralink bets on precision through surgery; UCLA bets on scale through safety and software. Translation: one is a medical device company, the other could be a consumer platform play. Investors know which curve scales quicker.
🛎️ Key takeaway: AI copilots may commercialize “mind control” faster than Neuralink can sterilize its tools.
TOGETHER WITH HELPSCOUT
AI-Powered Support, Human Touch
Help Scout is a customer platform with AI-powered email, chat, and knowledge base tools. AI drafts, summarizes, and translates replies instantly, automates workflows, and boosts agent efficiency. With 12,000+ teams onboard, it integrates with 90+ apps, simplifies collaboration, and enhances CSAT without ticketing complexity.
INTELLIGENCE
GPT-5 Wipes the Floor in AI Werewolf League
📌 What’s happening: Seven frontier LLMs were dropped into a controlled Werewolf tournament to test not code or math, but social intelligence: persuasion, deception, and resistance to manipulation. The results? GPT-5 crushed the field with near-perfect play, scoring a ruthless 96.7% Win Rate, while rivals like Google’s Gemini 2.5 Pro, Alibaba’s Qwen3, and Moonshot’s Kimi showed flashes of flair but collapsed under sustained pressure.
🧠 How this hits reality: This isn’t just a parlor trick. Social deduction is the closest thing we have to benchmarking “theory of mind” in AI agents—how they lie, resist lies, and coordinate under uncertainty. GPT-5’s dominance signals a new moat: whoever owns the most convincing socially competent agents will run the table in multi-agent workflows, negotiation bots, and autonomous orgs. Startups built on brittle reasoning or scripted persuasion might just had their expiration date stamped.
🛎️ Key takeaway: AI’s new benchmark isn’t IQ—it’s BS detection. And GPT-5 just proved it can spin, spot, and survive better than the rest.
Google: GPT-5 Wipes the Floor in AI Werewolf League
GPT-5 crushed the field with near-perfect play, scoring a ruthless 96.7% Win Rate, while rivals like Google’s Gemini 2.5 Pro, Alibaba’s Qwen3, and Moonshot’s Kimi showed flashes of flair but collapsed under sustained pressure.
Alibaba: GPT-5 Wipes the Floor in AI Werewolf League
GPT-5 crushed the field with near-perfect play, scoring a ruthless 96.7% Win Rate, while rivals like Google’s Gemini 2.5 Pro, Alibaba’s Qwen3, and Moonshot’s Kimi showed flashes of flair but collapsed under sustained pressure.
LLM
Switzerland Drops Apertus — A National LLM Built Like Public Infrastructure
📌 What’s happening: EPFL, ETH Zurich, and the Swiss National Supercomputing Centre just launched Apertus, an open-source large language model trained on 15T tokens across 1,000+ languages — with 40% non-English. Unlike U.S. giants, it’s fully transparent: datasets, source code, and training pipelines are all public. It also runs strictly under Swiss data and copyright law.
🧠 How this hits reality: Apertus isn’t just a model; it’s a regulatory play. By aligning with Swiss banking secrecy and EU-style compliance, it offers enterprises a “clean” LLM that sidesteps lawsuits dogging OpenAI, Perplexity, and others. For banks, insurers, and regulators, this flips AI from a liability minefield into compliant infrastructure. Open-weights plus European governance could erode U.S. incumbents’ moat in regulated industries.
🛎️ Key takeaway: Apertus reframes LLMs from “scraped IP landmines” into public utilities — the first serious shot at a sovereign, compliant alternative to Silicon Valley’s black boxes.
QUICK HITS
- OpenAI will route sensitive chats to GPT-5 and launch parental controls with age rules, memory off, and distress alerts after safety failures led to lawsuits and tragedies.
- Microsoft launched VibeVoice, an open-source AI for four voices and 90-minute podcasts, rivaling Google’s NotebookLM but focused on performance over text understanding.
- Amazon’s new Lens Live mirrors Google’s Gemini Live, but while Google focuses on information, Amazon slaps a “buy” button on everything you see.
- Apple quietly launched Asa, an internal AI chatbot for retail staff — not for consumers — ahead of the iPhone 17 launch.
- “Clanker,” a Star Wars term for battle droids, has gone viral as anti-AI slang on social media, widely used to vent frustrations with AI technologies.
TRENDING
Daily AI Launches
- Receiptor AI launched v2.0, an AI platform that fully automates bookkeeping by finding, categorizing, and syncing all your receipts to Xero/QBO.
- Google launched an AI-powered beta for Google Finance, allowing users to ask complex questions and get answers based on real-time market data
- FileAI launched the fileAI MCP, a server that gives AI agents secure access to files and instantly turns them into usable data using AI OCR and extraction.
- CatDoes launched a multi-agent AI platform that allows anyone, regardless of technical skill, to easily build and publish mobile apps.
Featured AI Tools
- 📚 Heardly is the Fast Way to read Best Book.
- 🎓 Resea AI is the first academic agent for end to end research workflow.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
- 🦾 Flot AI writes, reads, and remembers across any apps and webs.
TOGETHER WITH US
AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, boasting over 1 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. Our Newsletter Brands:
- AI: AI Secret
- Tech: Bay Area Letters
- Futuristic: Posthuman
We've helped promote over 500 Tech Brands. Will yours be the next?
Email our co-founder Mark directly at [email protected] if the button fails.
Latest Daily Rundowns
MagnaMagnaMagnaMore AI Stories
Ben
Magna
Magna
Good Morning, AI Enthusiasts!
Just when you thought your budget couldn’t possibly handle another AI project, Alibaba comes swooping in like a caffeinated superhero to save your financials.
SPONSORED BY
GAME-CHANGER
Groundbreaking Alibaba Solution Cuts Costs of AI Training for Search by 88%
In the dynamic realm of artificial intelligence, the introduction of Alibaba Group's ZeroSearch marks a transformative moment. This innovative framework addresses the significant financial and scalability hurdles associated with training LLMs for search tasks, making advanced AI capabilities more accessible to smaller organizations and startups.
- Training AI systems for search tasks has relied heavily on expensive commercial search engines, leading to exorbitant API costs and scalability issues, creating barriers for smaller organizations.
- ZeroSearch allows LLMs to simulate search interactions, drastically cutting training costs by 88% while maintaining high retrieval accuracy, thus removing reliance on external search engines.
- ZeroSearch democratizes AI development by lowering barriers for startups, potentially challenging the dominance of traditional search engines and reshaping how AI systems are trained.
- Evaluations show that ZeroSearch-trained models perform competitively against traditional methods and even outperform established search engines in key metrics, showcasing its effectiveness.
Groundbreaking Alibaba Solution Cuts Costs of AI Training for Search by 88%
In the dynamic realm of artificial intelligence, the introduction of Alibaba Group's ZeroSearch marks a transformative moment. This innovative framework addresses the significant financial and scalability hurdles associated with training LLMs for search tasks, making advanced AI capabilities more accessible to smaller organizations and startups.
TOGETHER WITH ONCELY
A New Way of Buying AI Software
Introducing Oncely.com: The smartest way to leverage AI in your Business and Career before everyone else.
Imagine a world where you can finally ditch those subscription fees that feel more like a bad relationship than a business decision. With our innovative “Pay Once, Use Forever” model, you can buy AI software that doesn’t just sit there, collecting dust like your gym membership.
Say goodbye to the subscription rollercoaster and hello to a smarter, more efficient future.
COPYRIGHT
U.S. Copyright Office Challenges AI Development Practices
The intersection of AI and copyright law is emerging as a critical battleground in the realm of technology and intellectual property. As AI systems increasingly utilize extensive datasets—often sourced from copyrighted materials—the legal and ethical implications of these practices have come under intense scrutiny.
- The recent report by the U.S. Copyright Office has ignited a significant discussion among stakeholders, particularly those within the tech industry, about the future of AI development and the rights of content creators.
- The report questions whether using copyrighted content to train AI models qualifies as "fair use," raising concerns about the reproduction rights of copyright holders and the economic impact on creators.
- The Copyright Office examines if AI training on copyrighted works meets fair use criteria, suggesting that tech companies' claims of transformation may not always hold.
- Big Tech companies argue that restricting access to copyrighted materials could hinder innovation and that their transformative outputs should qualify as fair use, while creators assert that such use constitutes infringement.
- The findings imply that AI companies may need to secure licenses for training materials, leading to increased transparency, potential economic impacts, and case-by-case legal determinations regarding the copyright status of AI-generated works.
QUICK HITS
- OpenAI and Microsoft are in tense negotiations to redefine their partnership.
- Meta is reportedly planning to add controversial facial recognition technology to its upcoming smart glasses, raising privacy concerns.
- President Trump fired U.S. Copyright Chief Shira Perlmutter just days after a significant AI report was released.
- Barack Obama warns that AI can outperform 60-70% of programmers, signaling a significant shift in the coding job market.
- Sam Altman reaches out to Elon Musk, proposing a truce over their social media feud to focus on AGI collaboration.
TRENDING TOOLS
- 👾 Pixel-Art.ai creates vibrant pixel art online.
- 💾 GitBook builds docs as a product.
- 🌄 Stylize reimagines your pictures with one click.
- 💆🏼 Minecraft Skin Generator creates custom skins with AI.
- 📺 Klap turns videos into viral shorts.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click. (Featured)
- 📚 Heardly is the Fast Way to read Best Book. (Featured)
- 🚀 Intellectia.ai is the most powerful AI platform for smarter investment. (Featured)
1 Million+ AI enthusiasts are eager to learn about your product.
AI Secret is the world’s #1 AI Newsletter, boasting over 1 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?
What We Can Offer:
- Launch an Advertising Campaign
- Introduce New Product or Features
- Other Business Cooperation
Email our co-founder Mark directly at [email protected] if the button fails.
Magna
Magna
Magna
Good Morning, AI Enthusiasts!
The introduction of shopping features in ChatGPT brings AGI one step closer to integrating into human life.
SPONSORED BY
NEW LAUNCH
OpenAI's Strategy to Make ChatGPT an All-Purpose App
OpenAI is on a new mission to transform ChatGPT into an "all-purpose app" designed to meet diverse user needs, potentially reshaping the competitive landscape in search and e-commerce. This ambitious endeavor reflects the company's vision of achieving AGI while addressing the challenges and advancements that come with such a transformation.
- OpenAI's long-term goal is to create an "everything app" by 2025, leveraging autonomous AI agents to perform multi-step tasks on behalf of users, as articulated by CEO Sam Altman at DevDay.
- The integration of shopping capabilities within ChatGPT allows users to browse and purchase products directly, complete with visual search results and user reviews, positioning it as a more trustworthy alternative to ad-driven search engines.
- ChatGPT's ad-free search results and growing user base, which has reached nearly 800 million weekly active users, position it as a direct competitor to Google, especially amid regulatory scrutiny of Google’s market dominance.
OpenAI's Strategy to Make ChatGPT an All-Purpose App
OpenAI is on a new mission to transform ChatGPT into an "all-purpose app" designed to meet diverse user needs, potentially reshaping the competitive landscape in search and e-commerce. This ambitious endeavor reflects the company's vision of achieving AGI while addressing the challenges and advancements that come with such a transformation.
TOGETHER WITH ONCELY
A New Way of Buying AI Software
Introducing Oncely.com: The smartest way to leverage AI in your Business and Career before everyone else.
Imagine a world where you can finally ditch those subscription fees that feel more like a bad relationship than a business decision. With our innovative “Pay Once, Use Forever” model, you can buy AI software that doesn’t just sit there, collecting dust like your gym membership.
Say goodbye to the subscription rollercoaster and hello to a smarter, more efficient future.
NEW LAUNCH
A Comprehensive Report on Alibaba's New Qwen 3 AI Model Series
The unveiling of Alibaba's Qwen 3 series on April 29, 2025, marks a pivotal moment in the realm of artificial intelligence. The Qwen 3 series aims to compete with and even surpass established AI giants like OpenAI and Google, reshaping the landscape of artificial intelligence.
- The Qwen 3 series introduces a groundbreaking dual-mode functionality, allowing models to operate in Thinking Mode for complex tasks and Non-Thinking Mode for general-purpose responses, optimizing efficiency.
- Featuring two MoE models that activate only a fraction of their parameters during use, the Qwen 3 series achieves high performance while maintaining lower computational costs, with models like Qwen3-30B-A3B outperforming traditional dense models.
- Trained on a massive dataset of 36 trillion tokens, the Qwen 3 series excels in various tasks, including reasoning and multilingual capabilities, showcasing improved performance over its predecessor.
- Demonstrating exceptional results across numerous benchmarks, the Qwen 3 models, especially the flagship Qwen3-235B-A22B, rival existing proprietary models, showcasing their advanced capabilities in areas like mathematical reasoning and coding.
Alibaba's Qwen 3 series signifies a major advancement in AI technology, combining innovative reasoning capabilities with efficient architecture, establishing a new frontier in the global AI race.
A Comprehensive Report on Alibaba's New Qwen 3 AI Model Series
The unveiling of Alibaba's Qwen 3 series on April 29, 2025, marks a pivotal moment in the realm of artificial intelligence. The Qwen 3 series aims to compete with and even surpass established AI giants like OpenAI and Google, reshaping the landscape of artificial intelligence.
QUICK HITS
- Yelp has launched AI voice agents to assist restaurants and service providers with call handling and customer inquiries.
- Geoffrey Hinton, the "Godfather of AI," supports efforts to block OpenAI's restructuring plans, arguing it undermines the nonprofit's mission.
- Perplexity AI launches its chatbot on WhatsApp to combat misinformation.
- Millions of Reddit users were unknowingly manipulated in an unauthorized AI experiment by University of Zurich researchers.
- A new study reveals that a generative AI chatbot, Therabot, significantly improves mental health treatment outcomes for users compared to a control group.
TRENDING TOOLS
- ✂️ Foto AI swiftly removes and replaces image backgrounds.
- 🎶 FineShare makes it easy to create realistic voices.
- 🎥 Riverside is your online studio to record in high quality.
- 📈 LinkDR gets 10x more backlinks SEO.
- ✂️ Spikes automates video editing in just 1 click.
- 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click. (Featured)
- 📚 Heardly is the Fast Way to read Best Book. (Featured)
- 🚀 Intellectia.ai is the most powerful AI platform for smarter investment. (Featured)
1 Million+ AI enthusiasts are eager to learn about your product.
AI Secret is the world’s #1 AI Newsletter, boasting over 1 million readers from leading companies such as OpenAI, Google, Meta, and Microsoft. We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?
What We Can Offer:
- Launch an Advertising Campaign
- Introduce New Product or Features
- Other Business Cooperation
Email our co-founder Mark directly at [email protected] if the button fails.
Magna
Magna
Magna
🛎️ White House Targets Qwen
Plus: Bindwell Redefines Pesticides, OpenAI's Legal Perimeter
🛎️ Grok 5 Leaking
Plus: Future of Commerce, Two Products Two Moats
From Maps to Meals: Amap’s Street Stars Takes on Meituan
Backed by subsidies and Alibaba’s ecosystem, Street Stars signals that the battle for China’s local services is moving from delivery apps to navigation gateways.
🛎️ Screen-Watching AI
Plus: OpenAI Don't Know, Apple Ghosts AI
🛎️ AI-BCI Copilot
Plus: AI Werewolf League, Switzerland‘s National Model
🛎️AI Training Revolution
U.S. Copyright Office Challenges AI Development Practices
Groundbreaking Alibaba Solution Cuts Costs of AI Training for Search by 88%
In the rapidly evolving field of artificial intelligence (AI), the cost and complexity of training large language models (LLMs) have posed significant challenges, particularly for smaller
🛎️ChatGPT is Now Available For Shopping
A Comprehensive Report on Alibaba's New Qwen 3 AI Model Series
A Comprehensive Report on Alibaba's New Qwen 3 AI Model Series
On April 29, 2025, Alibaba, a leading Chinese technology giant, unveiled its latest generation of large language models (LLMs), the Qwen 3 series. This release marks
