Latest
AIDiffusion LLMs explained: what Mercury 3's 1,000 tokens per second means for builders
Inception's Mercury 3 is a diffusion-based language model running at 1,068 tokens per second with a 1M-token context. Here's how diffusion LLMs differ from ChatGPT-style models, what Mercury 3 costs, and where speed matters.
Build GuidesDESIGN.md: give your coding agent a design system so every page looks the same
DESIGN.md is an open format from Google Labs for describing your visual identity to coding agents: tokens for exact values, prose for the why. Here's how to write one, lint it, and export it to Tailwind.
AIDeepSeek open-sources DeepInfer: what V4.1 Flash is, and can you run it yourself?
DeepSeek released DeepInfer, the inference engine behind its own service, plus a new V4.1 Flash checkpoint (1017). What the model card actually says, and the hardware DeepInfer needs.
TechDDR5 buying guide for late 2026: how much to buy, what speed, and whether to wait
DDR5 contract prices are still rising in Q4 2026. A plain guide to how much RAM you need, speed vs capacity, and how to decide whether to buy now.
FinanceWho is CXMT? China's DRAM maker, its giant IPO and the 4F² claim, explained
CXMT is China's biggest DRAM maker and now holds about 9.5% of the market, per TrendForce. What it makes, what's been reported, and what to watch on Oct 31.
TechWhat it costs to rent an H100, H200, B200 or B300 in October 2026
On-demand GPU prices from Lambda, Runpod and CoreWeave, checked Oct 22, 2026, with the math for what a real training or inference job costs.
Build GuidesContext windows explained: tokens, 1M-token models and what actually fits
A context window is a model's working memory: your system prompt, chat history, files, tool definitions and the reply all share it. Here's how it fills up, why 1M tokens isn't a free pass, and how to count before you send.
Build GuidesCodex CLI from zero: install, sign in, /init and your first task
A beginner's walkthrough of OpenAI's Codex CLI: install on Mac, Linux or Windows, sign in with your ChatGPT plan, generate an AGENTS.md with /init, and run your first task safely.
Build GuidesClaude Opus 5.5 for builders: price, 1M context, and the four changes that break Opus 5 code
A practical guide to Claude Opus 5.5: what it costs per job, how to pick an effort level, and the four breaking changes to fix before you swap the model ID.
AIClaude Duet, explained: Anthropic's proactive agent app, what it can do and who can get it
Claude Duet is a new Anthropic app where Claude checks your connected apps, sends a morning brief and runs tasks in the background, asking first before it sends, pays, signs or deletes. Here's what's in the research preview.
Build GuidesClaude Code subagents vs skills vs hooks: which one to use, with examples
Subagents, skills and hooks all extend Claude Code, but they do different jobs. A plain-English guide with a working example of each, plus the new \"Record a skill\" feature in Cowork for Mac.
FinanceChip earnings, Q3 2026: what TSMC, Lam Research and Texas Instruments reported
Three chip companies, three layers of the stack: the foundry, the equipment maker and the analog supplier. Their Q3 2026 numbers, straight from their releases.
AIChina's draft AI export controls: what they cover, what's exempt, and what it means for open models
China proposes adding 23 AI technologies, from large models to agents, to its export-control catalogue. Comments are open until Nov 15. Here's what the draft covers, why open-source work is excluded, and what it means for builders.
FinanceBig Tech earnings week (Oct 28 to Nov 2): the dates and the AI-capex lines to read
Microsoft and Alphabet report Oct 28, Amazon Oct 29, Apple Nov 2. Here's each company's last capex guidance and the lines worth checking first.
FinanceThe Anthropic IPO, explained: what a confidential S-1 is and what's been reported
Anthropic has confidentially filed for an IPO. What a draft S-1 is, what happens next, and what WSJ and Reuters say the draft shows (reported, not public).
FinanceWhat's an impairment? Gartner's $3.5 trillion AI warning, explained
Gartner says over $3.5T of AI infrastructure spending could face impairment by 2030. What impairment, depreciation and useful life mean, and how to spot them.
FinanceWho's paying for the $10 trillion AI build-out? A Columbia paper explains the plumbing
A Columbia paper for Brookings estimates $10.3T of AI infrastructure spending from 2025 to 2032. Here's how it's being financed, and why risk is harder to see.
TechAI data centres and the power grid: what the IEA numbers actually say
Data centres used about 485 TWh in 2025 and could hit 950 TWh by 2030, says the IEA. Here's what that means, what's driving it and where the grid pinches.
FinanceIs AI a bubble? Five numbers to check instead of guessing
A plain framework for judging the AI spending boom: capex versus revenue, versus cash flow, depreciation, debt and supplier orders. Worked through with Meta's Q3 2026 release.
AIAnthropic gives Agent Skills to the Agentic AI Foundation: what changes for your SKILL.md files
Anthropic donated the Agent Skills standard to the Linux Foundation’s Agentic AI Foundation on Oct 22, 2026. What moved, who maintains it, and what to do with your skills.
Build GuidesWhat is AGENTS.md? How to write one your coding agents follow
AGENTS.md is a Markdown file that tells AI coding agents how to work on your repo. What it is, which tools read it, and how to write one that works.
Build GuidesUTM tracking setup: a UTM naming guide for AI-built sites
A UTM tracking setup for GA4: what each UTM parameter does, a lowercase naming convention to copy, auto-tagging rules, and launch checks for AI-built sites.
Build GuidesHow to run coding agents safely: permissions, sandboxes and secrets
How to run coding agents safely with Claude Code permission modes, Codex sandbox and approval settings, deny rules, and habits that keep secrets out of reach.
Build GuidesMCP security best practices: how to use MCP servers without leaking data
MCP security best practices for vibe coders: prompt injection risks, what the MCP spec requires, and a checklist to use MCP servers without leaking data.