Daily Newsletter Aug 12, 2025

🛎️ Artificial Jagged Intelligence

More Stories: Claude Gets Memory, Grok 4 Waterloo

By Magna · 7 min read

Good Morning, AI Enthusiasts!

Artificial intelligence is rarely a straight line — more like a serrated blade, catching on triumph and failure alike.

AGI

Google’s “AJI” Era: Sharp Minds, Dull Edges

📌 What’s happening: DeepMind CEO Demis Hassabis says we’re stuck in AJI—artificial jagged intelligence—where models like Gemini can ace Olympiad math but botch high school algebra. The culprit? Inconsistency. Even with DeepThink reasoning boosts, these systems are elite in some domains and embarrassingly brittle in others. Sundar Pichai’s AJI label is now the polite way to say “brilliant idiot.”

🧠 How this hits reality: AJI isn’t a half-step to AGI—it’s a chasm. Closing it means more than shoving GPUs and data at the problem; it requires breakthroughs in reasoning, planning, and memory. For teams betting on near-term AGI, this is a cold shower: your “almost there” model may still hallucinate its way out of a paper bag.

🛎️ Key takeaway: AGI isn’t just “more AJI”—it’s a different beast. And right now, the beast is missing teeth.

TOGETHER WITH MURF

AI Voice Infrastructure that Powers Enterprises

Build audio experiences at scale with Murf's AI voice generator. Our ethically developed text-to-speech model delivers ultra realistic voices through studio, API, and localization tools - streamlining voiceovers for enterprises.

With Murf, the only limit is your imagination. Get ready to make your audience listen, laugh, and lean in.

Try now👉

NEW LAUNCH

Claude’s Memory Goes Selective—And That’s the Point

📌 What’s happening: Anthropic rolled out a “search-and-reference” memory for Claude, letting users pull past chats on demand. It works across devices, keeps projects siloed, and never builds a persistent user profile. Unlike OpenAI’s always-on memory, Claude won’t “remember” unless explicitly asked — no silent data hoarding, no surprise callbacks.

🧠 How this hits reality: For enterprise buyers and compliance teams, Claude’s opt-in recall is a feature, not a bug. It sidesteps privacy backlash, keeps audit trails clean, and reduces the risk of unintentional behavioral profiling. OpenAI’s default-on approach gives richer personalization but also a bigger regulatory attack surface. In a market already twitchy about AI “overfamiliarity,” Anthropic just handed security teams an easy win.

🛎️ Key takeaway: Claude remembers only when told — turning “forgetfulness” into a trust moat OpenAI can’t claim.

LLM

Grok 4’s Chess Loss Is a PR Bloodbath for Musk

📌 What’s happening: While Elon Musk was busy telling Microsoft CEO Satya Nadella on GPT-5 launch day that OpenAI would “eat Microsoft alive,” his own LLM, Grok 4, was being eaten alive — 4–0 — by OpenAI’s o3 in a live-streamed Google Kaggle AI chess showdown. The kicker? Five-time world champion Magnus Carlsen was live on mic, laughing, face-palming, and likening Grok’s blunders to “kids’ games” and club amateurs who only know openings.

🧠 How this hits reality: Forget Kaggle rankings — this was a marketing assassination. In an arena meant to showcase AI prowess, Grok’s collapse gave OpenAI a free highlight reel of dominance, complete with the world’s best chess player laughing at Musk’s flagship model. In a hype war where perception is product, Grok 4 just took a branding loss it can’t spin.

🛎️ Key takeaway: In AI chess, as in AI marketing, one bad night can hand your rival a year’s worth of victory ads.

QUICK HITS

Elon Musk's xAI capitalizes on GPT-5 user backlash by offering free, limited-time access to its Grok-4 model.
Reddit is now blocking the Internet Archive to prevent AI companies from using the Wayback Machine to scrape its content.
Microsoft to integrate GitHub more closely with its CoreAI unit after key executive’s departure.
Commonwealth Bank of Australia has rolled out an AI-powered tool within its app to help customers detect scams in real time.
Recent research suggests that the reasoning abilities of AI models are a 'brittle mirage' that fails when tested with problems outside of their training data.

TRENDING

Daily AI Launches

nFactorial AI launched MasterClass, offering real-time lessons and demos from top figures.
Hyprnote launched AI Notepad, recording meetings locally and generating personalized summaries.
My Juno Health launched AI Doctor, providing personalized health and vitality management.
SuperCraft launched a tool that lets teams design physical products using natural language.

Featured AI Tools

📚 Heardly is the Fast Way to read Best Book.
🪶 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: