news.
LIVE
Anthropic eyeing $900B valuation · $50B round
Cerebras IPO targets $3.5B raise · Nasdaq · mid-May
Apple exploring Intel & Samsung for future chips
Five Eyes warn agentic AI poses resilience risks
NHS walls off GitHub repos over AI security concerns
Google commits $40B to Anthropic · 5GW compute
Ollama 0.23 · all models available in Claude Code
Cerebras Q4 revenue +76% YoY · $510M · net income $87.9M
Anthropic eyeing $900B valuation · $50B round
Cerebras IPO targets $3.5B raise · Nasdaq · mid-May
Apple exploring Intel & Samsung for future chips
Five Eyes warn agentic AI poses resilience risks
NHS walls off GitHub repos over AI security concerns
Google commits $40B to Anthropic · 5GW compute
Ollama 0.23 · all models available in Claude Code
Cerebras Q4 revenue +76% YoY · $510M · net income $87.9M
★ Top story
TechCrunch · 30 Apr 2026
Anthropic's $900B valuation round closing within two weeks
Investor demand has hit fever pitch. Anthropic is asking investors to submit allocations within 48 hours for a $50B round at a $900B valuation — potentially leapfrogging OpenAI as the world's most valuable AI startup. Revenue run rate is closer to $40B, sources say.
Round~$50B
Valuation$900B
Rev run rate$40B
SourceTechCrunch
All · Today
Tech
AI
Open Source
Web Design
UI / UX
Today · all
Cerebras on track for blockbuster IPO — $10B in orders for $3.5B raise
TECH · Banks already fielding $10B in orders. 28M shares at $115–$125, valuing the AI chip co at $26.6B. OpenAI deal worth $20B through 2028 is the anchor.
TECHCRUNCH2 days ago→
Apple exploring Intel & Samsung as chip suppliers amid AI demand surge
TECH · Bloomberg reports shortages driven by AI data center buildouts and Mac demand for on-device AI. First sign Apple may loosen its tightly controlled chip partnerships.
BLOOMBERG2 days ago→
Samsung Galaxy S26 confirmed: Snapdragon 8 Elite 2 across all variants globally
TECH · HARDWARE · Samsung breaks from its Exynos split, going all-in on Qualcomm. Benchmark leaks suggest 40% compute boost over S25 flagship. Reveal expected Q1 2027.
ARST.3 days ago→
Big Tech layoffs slow but don't stop — 18,000 roles cut across sector in Q1 2026
TECH · INDUSTRY · Google, Meta, and Microsoft trimmed mid-management while doubling AI headcount. Net employment in the sector rose slightly thanks to AI infrastructure hiring.
THE VERGE3 days ago→
Google commits $40B to Anthropic — $10B now, $30B on milestones
AI · Google also providing 5GW of compute capacity over five years via Google Cloud TPUs. Comes alongside Amazon's $5B top-up. Anthropic IPO eyed for October.
TECHCRUNCH12 days ago→
Five Eyes warn rapid agentic AI rollout creates serious resilience risks
AI · Security agencies from US, UK, Canada, Australia, NZ urge organisations to prioritise safety over speed. Guidance targets autonomous AI systems acting without human oversight.
THE REGISTER1 day ago→
Braintrust raises $80M Series B to monitor AI system reliability in prod
AI · Enterprise observability for production AI. Round reflects investor conviction in the "boring" reliability layer as AI moves from experiment to core infrastructure.
TECHSTARTUPS1 day ago→
MIT Technology Review: 10 AI things that matter right now in 2026
AI · Multi-agent cooperation, robot training data, deepfakes as weapons, AI in the war room. The ideas and trends shaping what's technically possible this year.
MIT TECH REVIEW15 days ago→
Ollama 0.23: all cloud models now pipe directly into Claude Code & Cowork
OPEN SOURCE · AI · Developers running local models through Ollama can now use them inside Claude's agentic tools without leaving Claude Desktop. Significant capability unlock.
OLLAMA2 days ago→
NHS orders wall-off of hundreds of open-source GitHub repos over AI risk
OPEN SOURCE · Maintainers given May deadline. Concern: frontier AI tools lower the cost of finding and exploiting vulnerabilities in public code at scale.
THE REGISTER1 day ago→
GitHub Copilot Workspace hits GA — full repo context, multi-file edits, PR drafts
OPEN SOURCE · DEV TOOLS · Generally available to all GitHub Copilot Enterprise customers. Workspace can now reason across entire repositories, draft PRs, and run CI checks autonomously.
GITHUB BLOG2 days ago→
CSS anchor positioning lands in all major browsers — the layout game changes
WEB DESIGN · CSS · After years in spec, CSS Anchor Positioning reaches full cross-browser support. Tooltips, popovers, and floating elements can now position themselves relative to any element without JS.
CSS-TRICKS3 days ago→
Smashing Mag: 2026's web design shifts — spatial grids, variable fonts, adaptive colour
WEB DESIGN · TRENDS · Designers are leaning into container queries and :has() to build truly context-aware layouts. Variable fonts move from novelty to baseline. Dark/light adaptive palettes become the new default.
SMASHING MAG4 days ago→
View Transitions API v2 — cross-document animations without a SPA framework
WEB DESIGN · BROWSER · Chrome 128 ships View Transitions v2 enabling film-like page transitions on plain HTML sites. Firefox and Safari implementations are now behind a flag.
WEB.DEV5 days ago→
Design systems in 2026: token-driven, AI-generated, and finally boring (in a good way)
WEB DESIGN · SYSTEMS · A List Apart looks at how design tokens have matured — W3C Design Tokens spec is in CR, and tools like Tokens Studio now export to Figma, CSS, and SwiftUI simultaneously.
A LIST APART6 days ago→
Nielsen Norman: AI is breaking UX conventions — and that's mostly a problem
UI/UX · NN/g's 2026 research finds chat interfaces and AI suggestions are eroding discoverability. Users struggle to form mental models when interfaces change output on every visit.
NN/G4 days ago→
State of UX 2026: adaptive interfaces, ethical dark patterns, and the return of simplicity
UI/UX · TRENDS · UX Collective's annual report surfaces a counter-movement: after years of complexity, top-rated products are stripping back — fewer options, clearer defaults, less choice paralysis.
UX COLLECTIVE5 days ago→
Figma AI Design Mode: generate full component specs from a plain-text brief
UI/UX · TOOLS · Figma's new AI Design Mode takes a natural-language prompt and produces annotated component specs, auto-layout variants, and interaction states. Currently in beta for Professional plans.
FIGMA BLOG3 days ago→
Laws of UX 2026 update: motion design principles added — timing, easing, intent
UI/UX · MOTION · Jon Yablonski expands the Laws of UX canon with six new motion principles. Covers cognitive load of animation, when easing adds clarity vs. when it just slows users down.
LAWS OF UX7 days ago→
EU AI Act enforcement begins — first compliance deadlines hit this month
TECH · POLICY · High-risk AI systems must now meet transparency and testing requirements or face fines up to €30M. Several US companies scrambling to file conformity assessments.
ARS TECHNICA1 day ago→
Nvidia GB300 Grace Blackwell Superchip ships to hyperscalers — 1.4TB/s memory bandwidth
TECH · HARDWARE · First GB300 units dispatched to AWS, Google, and Microsoft Azure. 288GB HBM3e per chip. Inference throughput doubles over H100 for large context workloads.
THE VERGE2 days ago→
Microsoft Recall relaunches with on-device-only processing after privacy overhaul
TECH · WINDOWS · After pulling the feature last year over data leak concerns, Microsoft ships Recall with mandatory local processing, encrypted snapshots, and an opt-in requirement.
WIRED4 days ago→
SpaceX Starship Flight 9 catches both boosters — first full stack reuse achieved
TECH · SPACE · Mechazilla caught the Super Heavy booster and Ship simultaneously for the first time. Orbital velocity confirmed. Reuse turnaround target is under 24 hours.
TECHCRUNCH4 days ago→
Qualcomm Snapdragon X Elite 2 benchmarks leak — 60% CPU uplift on Geekbench
TECH · CHIPS · The successor to the X Elite ships this autumn. Early pre-production units showing substantial gains in both single-core and NPU workloads. Targets MacBook Pro M4.
BLOOMBERG5 days ago→
Google Willow 2 quantum chip hits 1,000 logical qubits with below-threshold error rates
TECH · QUANTUM · Google's second-gen Willow chip demonstrates error correction at scale for the first time. Team calls it the first chip where adding qubits reduces net error rate.
THE REGISTER3 days ago→
OpenAI ships GPT-5 — multimodal, 1M context, real-time voice and vision
AI · GPT-5 launches to all tiers. Scores above PhD level on GPQA across all domains. Native code execution, image editing, and voice mode with sub-300ms latency ship day one.
OPENAI BLOG3 days ago→
Enterprise AI agents are here — but 70% fail within 90 days, new study finds
AI · ENTERPRISE · MIT Sloan study of 400 deployments finds most agentic failures trace to tool reliability and context window management, not model capability. Tooling layer is the bottleneck.
MIT TECH REVIEW3 days ago→
Meta releases Llama 4 Scout and Maverick — MoE architecture, 10M context
AI · OPEN SOURCE · Llama 4 Scout (17B active params) and Maverick (400B MoE) both released under a commercial licence. Scout runs on a single H100. Maverick matches GPT-4o on MMLU.
WIRED5 days ago→
AlphaFold 3 predicts novel antibiotic candidates — first real-world drug from AI structure
AI · SCIENCE · DeepMind and Isomorphic Labs announce a lead compound targeting drug-resistant TB, discovered entirely via AlphaFold 3 structure prediction. Phase I trial set for 2027.
DEEPMIND6 days ago→
30% of new code at Google now written by AI — CEO Sundar Pichai confirms
AI · DEV · Pichai says AI-generated code is reviewed by engineers before merge but the ratio has risen sharply since Q4 2025. Similar numbers reported at Meta and Microsoft.
HN / THE VERGE2 days ago→
Mistral launches Le Chat Pro — $20/mo, web search, canvas, file analysis
AI · PRODUCTS · France's Mistral enters the consumer subscription market. Le Chat Pro bundles their latest Mistral Large model with real-time web grounding and a collaborative canvas tool.
TECHCRUNCH2 days ago→
OpenTofu 1.9 ships statefiles encryption, provider functions, and improved CLI
OPEN SOURCE · INFRA · The Terraform fork continues to diverge. State encryption is a key enterprise ask that HashiCorp never shipped. 1.9 also lands provider-defined functions for custom validation logic.
GITHUB3 days ago→
Rust 2026 Edition announced — async closures stable, gen blocks, better error messages
OPEN SOURCE · LANGUAGES · The three-year edition cycle delivers async closures as a stable feature after years of RFC discussion. Gen blocks bring generator syntax for lazy iteration. Compiler error quality overhauled.
RUST BLOG4 days ago→
Bun 2.0 — Node.js compatibility at 99.4%, 3× faster installs, built-in SQLite ORM
OPEN SOURCE · JS · The JavaScript runtime hits a major milestone. 99.4% of the Node.js test suite now passes. The bundled SQLite ORM removes an entire dependency category for many projects.
BUN.SH5 days ago→
Astro 5 launches Content Layer API and Server Islands — static sites get dynamic slots
OPEN SOURCE · WEB · Server Islands allow individual components to be server-rendered within an otherwise static page, without a full SSR setup. Content Layer unifies local and remote content sources.
ASTRO BLOG6 days ago→
Vite 7.0 — Rolldown bundler replaces esbuild, cold start drops 40%
OPEN SOURCE · TOOLING · The move to Rolldown (Rust-based Rollup rewrite) cuts cold start and full rebuild times significantly. HMR latency unchanged. All major frameworks have committed to Vite 7 support.
VITEJS.DEV7 days ago→
Deno 2.2 ships OpenTelemetry built-in and Node.js `node:http` parity
OPEN SOURCE · RUNTIME · Built-in OTEL tracing with zero config is the headline. `node:http` now passes the full Node compatibility test suite, opening the door to Express and Fastify on Deno without shims.
DENO.COM8 days ago→
Hono v5 — JSX streaming, RPC client types, runs on 10 runtimes
OPEN SOURCE · WEB · The ultra-fast web framework adds JSX streaming support and typed RPC client generation from route definitions. Now officially supported on Bun, Deno, Node, Cloudflare Workers, and six more.
HONO.DEV9 days ago→
CSS inline if() lands in Safari 19 — conditional styles without a preprocessor
WEB DESIGN · CSS · The CSS Working Group's inline if() proposal ships in Safari 19 and Chrome 128. Developers can now write conditional values like `color: if(style(--theme: dark): white; else: black)` in plain CSS.
WEBKIT BLOG2 days ago→
OKLCH in production — why every design system should drop HEX this year
WEB DESIGN · COLOUR · Smashing's deep dive on OKLCH explains perceptual uniformity and why tints and shades look better than hex scales. Includes migration paths from Tailwind and CSS custom properties setups.
SMASHING MAG3 days ago→
Scroll-driven animations without JS — a Codrops practical guide for 2026
WEB DESIGN · ANIMATION · Pure CSS scroll-driven animations are now baseline. Codrops walks through animation-timeline, view(), scroll(), and how to replace most GSAP ScrollTrigger patterns with zero JS.
CODROPS4 days ago→
Container style queries hit baseline — components that respond to their own data
WEB DESIGN · CSS · Style queries let components change layout based on custom property values set by a parent, not just size. Enables truly context-aware design systems without class toggling.
WEB.DEV5 days ago→
AI-generated UIs are getting good — but they still can't do accessibility
WEB DESIGN · AI · NN/g tested 200 AI-generated interfaces for WCAG 2.2 compliance. Average pass rate: 34%. Focus management, ARIA labelling, and colour contrast remain the most common failures.
NN/G6 days ago→
Variable fonts in 2026: optical sizing, grade axis, and the end of system stacks
WEB DESIGN · TYPE · With Instrument Serif and Plus Jakarta Sans now in Google Fonts, variable font adoption is mainstream. CSS-Tricks benchmarks optical size axis performance gains across 10 popular typefaces.
CSS-TRICKS7 days ago→
Figma Variables 2.0 — number modes, boolean logic, and cross-library syncing
UI/UX · TOOLS · Figma's variables system gets a major expansion. Number modes allow responsive spacing scales to be token-driven. Boolean logic gates enable if-then component states without plugins.
FIGMA BLOG1 day ago→
The hamburger menu is finally dying — what's replacing it in 2026
UI/UX · PATTERNS · UX Collective analysis of 500 top apps shows hamburger usage down 41% YoY. Tab bars, gesture navigation, and contextual command palettes are the main replacements — each with different task-completion trade-offs.
UX COLLECTIVE2 days ago→
Inclusive design in the age of AI — when personalisation conflicts with accessibility
UI/UX · ACCESSIBILITY · A List Apart explores the tension between AI-adaptive UIs and WCAG compliance. Adaptive interfaces that rearrange content for efficiency can break screen reader expectations and cognitive accessibility patterns.
A LIST APART3 days ago→
Chat is the worst UI pattern for most tasks — NN/g's 2026 research is clear
UI/UX · RESEARCH · Two years of lab testing confirms chat interfaces perform poorly for task completion vs. structured UIs. Users struggle with blank prompts, get lost in long threads, and can't undo. Hybrid UI is the answer.
NN/G4 days ago→
Spatial computing UI patterns for Vision Pro and Android XR — a 2026 field guide
UI/UX · SPATIAL · With Vision Pro in its second year and Android XR devices shipping, UX Collective documents the emerging patterns: gaze-dwell selection, wrist anchors, and world-locked panels that don't induce fatigue.
UX COLLECTIVE5 days ago→
State of Design Systems 2026 — 78% now token-driven, AI generation still fringe
UI/UX · SYSTEMS · Figma's annual survey of 3,200 practitioners shows token adoption has hit mainstream. AI-generated components remain under 10% of actual production usage despite heavy press coverage.
FIGMA BLOG6 days ago→
Apple WWDC 2026 preview — on-device LLM, redesigned Siri, and visionOS 3
TECH · APPLE · Expect a major Siri overhaul built on Apple's own on-device model, a redesigned home screen for iOS 20, and visionOS 3 with hand-tracking improvements. Keynote 9 June.
THE VERGE1 day ago→
Pixel 10 Pro spec sheet leaks — Tensor G5 built on TSMC 3nm, 16GB RAM standard
TECH · HARDWARE · First Pixel chip not made on Samsung silicon. TSMC 3nm reportedly cuts thermal throttling by half. 50MP triple camera, 7 years of updates, launching October.
ARS TECHNICA2 days ago→
xAI releases Grok 3 — trained on 100k H100s, claims top LMSYS Chatbot Arena score
AI · Elon Musk's xAI ships Grok 3 to X Premium+ subscribers. Claims highest Elo on Chatbot Arena at launch. 128k context, real-time X data grounding, and a new reasoning mode called "Big Brain".
BLOOMBERG1 day ago→
Perplexity raises $500M Series E at $14B valuation — targets Google Search directly
AI · FUNDING · Round led by SoftBank with participation from Jeff Bezos. Perplexity now has 15M DAUs. New funding targets international expansion and an enterprise tier launching Q3 2026.
TECHCRUNCHToday→
Node.js 24 LTS — native TypeScript stripping, built-in test runner stable, npm 11
OPEN SOURCE · JS · The new LTS ships with native .ts file execution via type stripping (no transpilation). The built-in test runner graduates to stable. npm 11 included with workspaces improvements.
NODE.JS BLOG3 days ago→
PostgreSQL 17 — incremental backups, JSON_TABLE, 2× faster vacuum on large tables
OPEN SOURCE · DATABASE · pg17's headline is incremental backup support via pg_basebackup. JSON_TABLE lands as a SQL/JSON standard feature. Vacuum memory management overhauled for tables over 1TB.
POSTGRESQL.ORG5 days ago→
CSS subgrid in production — the definitive guide after 18 months of browser support
WEB DESIGN · CSS · Subgrid has been baseline for long enough that real production patterns have emerged. Smashing documents card alignment, form layout, and magazine-style grids that were impossible before.
SMASHING MAG1 day ago→
Designing dark mode properly in 2026 — contrast, colour science, and the grey trap
WEB DESIGN · COLOUR · Codrops digs into why most dark modes look wrong. The grey trap (using neutral greys instead of dark saturated hues), contrast ratios at low luminance, and why OKLCH makes dark palettes trivial to get right.
CODROPS2 days ago→
Generative UI patterns — NN/g documents six emerging interface archetypes for 2026
UI/UX · AI · Nielsen Norman Group's new report names six recurring patterns in AI-driven UIs: ambient oracles, progressive disclosure agents, command surfaces, narrated results, scaffolded forms, and hybrid pickers.
NN/G1 day ago→
Micro-interactions in 2026 — what still works and what's become noise
UI/UX · MOTION · After a decade of overuse, UX Collective's analysis of 1,000 apps finds micro-interactions improve completion rates only when tied to state changes. Decorative animations reduce trust scores by 12%.
UX COLLECTIVE2 days ago→
Trending topics
#anthropic-900b↑ 210%
#cerebras-ipo↑ 178%
#agentic-ai-risk↑ 94%
#google-anthropic↑ 82%
#ollama-023↑ 61%
#ai-chips↑ 44%
#multi-agent↑ 38%
#crypto↓ 14%
Sources · 21
── Tech & AI ──
TechCrunch · AILive
Bloomberg TechLive
CNBC TechLive
The Register · AILive
MIT Tech ReviewLive
Ars Technica · AILive
Wired · AILive
Tech StartupsLive
The Verge · TechLive
── Dev & Open Source ──
Hacker NewsLive
GitHub BlogLive
Simon Willison's BlogLive
LLM Stats · NewsLive
Tech Reader DigestLive
── Design & UI ──
Smashing MagazineLive
CSS-TricksLive
UX CollectiveLive
Nielsen Norman GroupLive
Figma BlogLive
A List ApartLive
web.dev · ChromeLive
CodropsLive