A
mdx/ai
Foundation models, fine-tuning, agents, evals.
Explore
Discover communities across the MDX Forum.
Recommended for you
mdx/aiFoundation models, fine-tuning, agents, evals.mdx/agentsTool-use, planning, memory, and agentic workflows.mdx/foundersShipping, distribution, pricing, hiring.mdx/web3Ethereum, rollups, DeFi, and on-chain design.mdx/pythonFrom async to data — practical Python in production.mdx/rustSystems, embedded, web, and the borrow checker journey.
Design
mdx/designProduct, interaction, motion — pixel accountability.mdx/uxResearch, IA, accessibility, and behavioural design.mdx/figmaTips, libraries, plugins, variables, and dev mode.mdx/uiComponent design, design systems, and visual craft.mdx/cssModern CSS — container queries, cascade layers, subgrid.mdx/motionWeb animation, Framer Motion, GSAP, Lottie, and timelines.
Hot Topic
Why we open-sourced Grok-4 and what comes nextToday we released Grok-4 weights under an Apache 2.0 license. The model is competitive with the frontier on reasoning and coding, behind on long-context recall,On the difference between leverage and luckLeverage is reproducible. Luck is not. Founders confuse them constantly, especially after a win, and especially in their second company.
Specific knowledge, acWe deleted our AWS bill and saved $2M/yearThe on-prem migration is done. Six weeks of pain, eighteen months of savings. Here's the cost breakdown, the hardware we chose, and the two things we'd do diffeDeleting more than adding this quarterEvery line of code is a liability. Every meeting is a tax. Every feature is a support burden. Simplify aggressively.What I learned shipping 12 agent products in 18 monthsEighteen months ago we ran our first eval harness against a handful of agent prototypes. Today twelve of those prototypes are in production serving real users. Good taste is negative informationKnowing what *not* to add to the kernel is worth more than knowing what to add. Taste is the compiler for intuition.