In partnership with
Good morning,
Last few weeks were all about OpenAI and Anthropic. Especially Claude seemed to have found its way to many people. And hopefully people will stop saying they ‘used ChatGPT’ and they’ll turn to Claude or even better : Gemini.
In the meantime - let’s point our attention at what Elon is doing.
Grok is a serious contender for the other LLMs but has fallen behind the competition the last few months.
Elon Musk has spent the last few months dismantling the old foundations of xAI to rebuild it from the bedrock up. After 10 of the 12 original founders departed - many of them stating that the work culture under Elon Musk are unsustainable -xAI has pivoted toward a "scorched earth" hiring strategy, poaching top-tier talent from Cursor and OpenAI’s elite "Thinking Machines" lab.
Musk’s goal is to turn Grok in the the world’s most dominant coder and real-time researcher.
The Gigawatt Gamble
However - and this is also causing the delay. xAI is also making some serious infrastructure moves in the background. At the heart of this is Colossus, a gigawatt-scale supercluster that serves as the laboratory for Grok 5.
Elon Musk is ambitious, we know this. But his ambitions for xAI go beyond Earth even.
They’re working on launching AI datacenters into space, running on solar energy. This is why he has merged xAI and SpaceX into one company.
Using "Project Suncatcher" logic, these centers would sit in 24/7 sunlight, eliminating the need for massive battery storage and bypasses Earth-bound energy regulations.
Musk asserts that within three years, xAI will be so far ahead that you’ll need the James Webb telescope to see second place. While competitors like Google and Anthropic currently lead in "deep thinking" benchmarks, they lack a dedicated rocket company to scale their compute into the heavens.
Tone of Voice
Another change - and I think this one many people are underestimating this one.
In the trenches of daily use, a clear divide is forming between the major models. While GPT 5.4 Pro and Claude Opus 4.6 are the current kings of academic research and coding, they’ve both developed a frustrating "personality" quirk.
Especially - in line with its CEO it seems - GPT 5.4 is extremely annoying. Users are reporting that OpenAI’s latest model has become reflexively contrarian. Instead of answering a health or research query directly, it often spends the first five paragraphs explaining why your question is structurally flawed or "dangerous."
Grok 4.20 has taken the opposite path.
It is fast, unhinged when requested, and utilizes X’s real-time firehose to index upwards of 390 sources for a single search. It remains the go-to for "what is happening right now," Musk says.
Universal High Income
Musk is also working on what comes beyond all this. The endgame of all this compute is a radical restructuring of the American economy. Former OpenAI star Andrej Karpathy recently visualized this shift with a "tree map" of job exposure. The data was stark: Software Developers face an 8.0–9.0 exposure score, while manual trades like Roofing remain at a safe 0.0.
Musk’s response to this looming displacement is Universal High Income.
His theory is simple: as AI and automation remove the "human cost" from the supply chain, the price of food, clothing, and shelter will collapse. We aren't just looking at "free money," but a world where a modest stipend buys a middle-class life because the cost of producing that life has trended toward zero.
With Google recently hiring a Chief AGI Economist, it's clear the world's power players are no longer debating if this transition happens, but when.
Which I guess would be good news for all of us
Welcome to the Blacklynx Brief
Free email without sacrificing your privacy
Gmail is free, but you pay with your data. Proton Mail is different.
We don’t scan your messages. We don’t sell your behavior. We don’t follow you across the internet.
Proton Mail gives you full-featured, private email without surveillance or creepy profiling. It’s email that respects your time, your attention, and your boundaries.
Email doesn’t have to cost your privacy.
AI News
Google upgraded Maps with Gemini-powered features like Ask Maps, which answers travel questions using data from millions of locations, and Immersive Navigation, which shows routes in detailed 3D. The update also adds conversational voice guidance and smarter route comparisons. By embedding AI directly into Maps, Google is bringing advanced AI into everyday navigation for billions of users.
Microsoft introduced Copilot Health, an AI system that combines medical records, wearable data, and health history to provide personalized insights. The tool connects to thousands of hospitals and devices while keeping user data private and not used for training. Microsoft aims to make healthcare guidance more accessible without replacing doctors.
Security startup CodeWall revealed that its AI agent breached McKinsey & Company’s internal AI system, exposing millions of messages and files due to basic security flaws. The vulnerability was quickly patched after disclosure, and no external misuse was found. The incident highlights how even major organizations can overlook critical security risks when deploying AI systems.
Elon Musk said xAI is being rebuilt from the ground up after most of its original co-founders departed and the company fell behind in coding capabilities. Musk acknowledged that Grok is currently lagging behind competitors and has begun hiring top engineering talent to close the gap. The overhaul signals a major reset as xAI tries to stay competitive in the fast-moving AI race.
An AI-assisted effort led by consultant Paul Conyngham used tools like ChatGPT, Grok, and AlphaFold to create a custom mRNA cancer vaccine for his dog. By analyzing large amounts of tumor data and working with a genomics lab, the team developed a treatment that helped shrink one tumor significantly. The case highlights how AI tools are enabling individuals to tackle complex medical challenges in new ways.
Nvidia unveiled major updates at GTC 2026, including NemoClaw for secure AI agents, the next-gen Vera Rubin chip platform, DLSS 5 for improved game graphics, and new enterprise and robotics tools. The announcements highlight Nvidia’s strategy to power all layers of AI, from infrastructure to applications. The company continues to position itself as the backbone of the AI ecosystem.
A creator known as “Kage” used AI music platform Suno to build a fictional band called Neon Oni, which gained thousands of listeners before being revealed as AI-generated. After the reveal, real musicians were hired to perform the songs live, turning the project into a hybrid of AI creation and human performance. The case shows how AI can reshape creative industries while still creating new opportunities for human artists.
Meta-backed startup Manus launched My Computer, a desktop AI agent that can directly access and manage files, run commands, and automate tasks on a user’s machine. The tool brings agent capabilities from the cloud to local devices, enabling actions like organizing files or building apps autonomously. The release reflects growing competition to control how AI interacts with personal computers.
OpenAI is shifting its strategy to focus more on coding tools and enterprise customers after internal leadership called Anthropic’s business momentum a “wake-up call.” Executives warned the company has been spread too thin across projects, while tools like Codex and GPT-5.4 are now central to winning back enterprise users. The move signals a renewed push to compete directly in the fast-growing AI workplace market.
Mistral AI launched Forge, a platform that lets companies build fully custom AI models using their own data without sharing it externally. The system provides full training pipelines similar to Mistral’s internal processes and can run entirely on private infrastructure for sensitive industries. The approach targets enterprises that need highly specialized AI systems tailored to proprietary data.
Microsoft reorganized its AI division, merging Copilot teams and shifting CEO Mustafa Suleyman to focus on developing in-house superintelligence systems. The move comes as Copilot struggles to gain traction compared to rivals, prompting Microsoft to double down on building its own advanced AI models. The restructuring reflects growing pressure on the company to compete more effectively in the AI race.
Quickfire News
Axiom raised $200M in a Series A at a $1.6B+ valuation, focusing on AI reasoning for formal mathematics and verified systems
Meta delayed its upcoming AI model “Avocado” until at least May after internal tests showed it lagging behind leading models
Perplexity expanded its “Computer” agent to Pro users, offering access with optional usage-based credits
United States Department of Defense CTO Emil Michael said there is no chance of renewed talks with Anthropic, criticizing Claude’s policies
xAI hired Andrew Milich and Jason Ginsberg from Cursor to strengthen Grok’s coding capabilities under Elon Musk
Bumble plans to launch “Bee,” an AI assistant that learns user preferences privately to suggest better matches
Meta is reportedly planning layoffs affecting up to 20% of its ~79,000 employees to help offset massive AI infrastructure spending
ByteDance is delaying the global launch of Seedance 2.0 after backlash from Hollywood over copyright concerns
Elon Musk said Tesla’s Terafab semiconductor facility will launch within a week to produce custom chips
A Florida homeowner used OpenAI’s ChatGPT to manage the sale of his house, handling pricing, marketing, and contracts while closing in five days and saving agent fees
Former Anthropic researchers are raising $175M for Miraendil, a startup focused on AI for scientific R&D in biology and materials science
Encyclopaedia Britannica and Merriam-Webster sued OpenAI, alleging unauthorized data use, competing outputs, and hallucinated attributions
Meta signed a $27B deal with Nebius to expand AI infrastructure, including early deployment of Nvidia’s Vera Rubin platform
Physical Superintelligence released Get Physics Done, an open-source AI agent that can plan research, run experiments, verify results, and write papers
Moonshot AI introduced Attention Residuals, a method that improves model efficiency by reusing earlier-layer information for about 1.25x better compute performance
OpenAI is reportedly restructuring its Stargate compute team and shifting toward renting AI server capacity instead of building its own data centers
Gamma launched Imagine, an AI design tool within its platform that creates logos, infographics, and social visuals with automatic brand styling
OpenAI released GPT-5.4 mini and nano, smaller and faster versions of its flagship model designed for coding assistants and multi-agent systems
Google began rolling out Personal Intelligence features to free users across Search AI Mode, the Gemini app, and Chrome in the U.S.
Anthropic introduced Dispatch for Claude Desktop, allowing users to message the assistant from their phone while it runs tasks on a computer
Mistral released Small 4, an open-source model combining reasoning, coding, and vision in one system
World launched AgentKit, enabling websites to confirm a real human is behind AI-driven purchases
Closing Thoughts
That’s it for us this week. Please like and subscribe :)
