Daily NewsletterSep 3, 2025

🛎️ AI-BCI Copilot

Plus: AI Werewolf League, Switzerland‘s National Model

By Magna · 6 min read

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.

Try now👉

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.

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: