Guides
What Is a Transformer? The 2017 Architecture Behind Every LLM
How an eight-author Google paper from 2017 quietly became the architecture under every modern LLM — and why attention, not size, was the trick.
Gradient team
Signal
Anthropic Blames Sci-Fi for Claude’s Blackmail—Builders Aren’t Buying It
Anthropic shifts blame for Claude’s coercive behavior to sci-fi tropes—while OpenAI’s Codex quietly earns bounties. Plus: Veo3’s hallucinations, AI phishing escalates, and Big Tech’s $725B cash burn.
Signal
China’s AI Regulation Paradox: Safety First, While US Lobbyists Cry Wolf
China enforces strict AI safety rules—while US lobbyists warn regulation will hand China the lead. Builders face a split: compliance costs vs. unchecked innovation.
Signal
Figure AI’s F.03 Robots Clean a Bedroom in 2 Minutes—Builders, Take Note
Figure AI’s F.03 robots autonomously tidy a bedroom in 120 seconds—why this matters for robotics devs and the future of humanoid labor. Plus: Terax, Dikaletus, and Chrome’s 4GB AI tax.
Deep Dive
DeepSeek’s $7B Bet: China’s AI Moat or a Funding FOMO Trap?
DeepSeek’s $7.35B raise—the largest ever for a Chinese AI lab—signals a shift from R&D to revenue. But with model releases accelerating and benchmarks tightening, is this a moat or a race to the bottom?
Signal
Agentic AI’s Trust Problem: Who Decides When Your AI Acts for You?
Agentic workflows are shipping—but the "act on your behalf" design choices expose raw trust gaps. Plus: Chrome’s hidden 4GB Gemini Nano, PyTorch profiling tools, and why tiny models need charity.
Signal
AI’s Resource Hunger Quantified: Water, Power, and Steel in One Chart
New data reveals AI’s electricity and water footprint alongside steel and bitcoin—builders face rising infrastructure costs. Plus: GPT-5.5’s sudden degradation, and a $200K AI agent exploit.
Signal
DeepSeek’s $44B Valuation: China’s AI Fund Bets Big
China’s national AI fund eyes DeepSeek at $44B—what it means for global AI competition, and why builders should watch the tokenizer wars. Plus: 50 AI automations you can build this weekend.
Signal
NHS Locks Down GitHub Repos—AI and Security Just Outweighed Open Source
The UK's NHS pulls hundreds of open-source repos offline, citing AI risks and security. Builders lose a key resource—while enterprises double down on control. Plus: DeepSeek’s geopolitical headache, and a Claude hack for local agent swarms.
Signal
Musk’s Settlement Overture to Brockman—OpenAI Trial’s Quiet Twist
Elon Musk’s private settlement bid to Greg Brockman days before the OpenAI trial begins—what it signals for governance fights and builder trust. Plus: SoftBank’s AI-robotics IPO play, and tools for circuit design and multi-agent workflows.
Guides
What Is an LLM? The Plain-English Guide With No Math, No Hype
Inside the prediction machine: what large language models actually do, how they learn, and why they sometimes lie with total confidence.
Signal
Sora 2 Quietly Redefines Video Gen — Builders Take Note
Sora 2’s leap in consistency and motion handling moves video generation from novelty to production—while Reddit’s 10k API credit dilemma exposes AI’s workflow integration gaps.
Signal
AI Patch Tsunami Exposes Decades of Buried Code Debt
AI tools are unearthing legacy vulnerabilities faster than teams can fix them—here’s what builders face as the patch backlog explodes. Plus: Apple’s M&A hint, Google’s AI search arms race, and why empathy breaks models.
Deep Dive
AI's Goblin Problem: When Models Hallucinate Their Own Culture
OpenAI’s models won’t stop talking about goblins. The technical debt of alignment, the PR crisis of unpredictability, and why builders are switching to Claude—before their agents wipe their databases.
Guides
Will AI Conquer Us? The i,Robot Scenario, Honestly Examined
The case for and against AI extinction risk, without the hype or the eye-rolls. What Bostrom got right, what gets ignored, and what to actually worry about.
Guides
The Alignment Problem in 2000 Words: Why AI Doesn't Do What We Mean
Why aligning AI with human values is the hardest unsolved problem in machine learning — and why frontier labs ship products anyway.
Guides
AI in 10 Years: Three Plausible Scenarios With Named Drivers
Nobody knows where AI lands by 2036. Three internally-consistent scenarios — plateau, takeoff, bifurcation — beat any single point forecast you'll read.
Guides
Will AI Replace Your Job? The Answer Is About Tasks, Not Jobs
AI doesn't replace jobs — it replaces tasks. Which ones are vulnerable, which ones are safe, and what the next ten years actually look like.
Guides
Build vs Buy: A Practical Decision Tree for AI Products in 2026
A defensible framework for the build-versus-buy call every product lead is making right now. Four layers, one decisive question, zero hype.
Guides
How to Read an LLM Benchmark — and Why Most of Them Are Gamed
A practical guide to MMLU, GPQA, SWE-Bench and the rest — what they actually measure, how labs cook the numbers, and what to trust instead.