Search the archive

Moment

ANTHROPIC DEBUTS CLAUDE DESIGN FOR CREATING PROTOTYPES, PITCH DECKS, AND MOCKUPS

Claude Design can create designs, prototypes, slides, and more using Opus 4.7, an AI model that was introduced earlier this week. It can see images in greater resolution than Anthropic's previous models and is more 'tasteful and creative' when performing professional tasks. Claude Design allows founders, product managers, and marketers without a design background to create high-quality interfaces, slides, and docs. Teams in testing have been using it to create realistic prototypes, wireframes, mock-ups, design explorations, pitch decks, presentations, and social media assets. It is now available as a research preview for Claude Pro, Max, Team, and Enterprise customers.

APPLE'S REVAMPED SIRI INTERFACE IN IOS 27 IS HIDDEN IN WWDC TEASER

Apple's WWDC 2026 teaser provided a glimpse of the revamped Siri interface that will be part of iOS 27. The teaser hints at a complete overhaul of Siri that includes previously announced features along with several additional enhancements. The enhanced Siri will be able to handle multiple commands within a single query and support third-party AI agents. The design is still being tested within the company. The Siri announcement will be a central part of the announcements at the upcoming WWDC.

BLUE ORIGIN SUCCESSFULLY RE-USES A NEW GLENN ROCKET FOR THE FIRST TIME EVER

Blue Origin successfully reused one of its New Glenn rockets, but the overall success of the mission is in question as the communications satellite that the rocket carried to space wound up in an off-nominal orbit. The company missed its target, which suggests something may have gone wrong with the rocket's upper stage. It is unclear whether the upper stage performed its second burn, scheduled an hour after the rocket lifted off. The re-use feat was achieved on just the third-ever launch of New Glenn.

PANCREATIC CANCER MRNA VACCINE SHOWS LASTING RESULTS IN AN EARLY TRIAL

mRNA vaccines, personalized for each individual using genetic material taken from their unique tumor cells, are showing promise in the fight against pancreatic cancer. These vaccines work as a type of immunotherapy that stamps out lingering, undetected cancer cells after surgery to remove larger tumors. Traditional immunotherapies only work in about 20% of all cancers, and pancreatic cancer is the poster child for difficult-to-treat cancers. The drugmakers behind the vaccine have now started Phase 2 clinical trials.

instead.

THE AGENT STACK BET

Most 'production agents' shipping today feature custom plumbing, fragile session logic, shared service accounts, and weak security models. While models and tools have gotten dramatically better, there is a stack ceiling that is quietly creating a governance and reliability gap that agentic systems can't grow through. The current stack has excessive agency, with agents leaving a trail of corrupted state behind them. This article suggests four architectural bets teams should consider to move to the next generation of agentic systems.

OUTPUT ISN'T DESIGN

Many people mistake design as the act of producing something. However, the hard part of design is actually understanding the problem enough to know what and how something should exist at all. There is use and place for tools, but the tools aren't part of the design process. Design is the search for a good fit between a form and its context. Good design considers human needs, technical constraints, conflicting requirements, habits, edge cases, and easy-to-miss relationships that require time to identify.

INSIDE NOTION

Notion is a company built on ideology, brand, and community rather than technical edge. The company's software gives users blank pages they can customize with modular blocks into whatever they want. Notion became the canonical product-led growth company after it first took off in 2018, propelled by brand ambassadors, influencers, and fandoms across social media. This article explores how the company works and how it became one of the top 10 most-used AI products in the world, despite not being 'AI-native'.

META TARGETS 20 MAY FOR 8,000 LAYOFFS AS IT REDIRECTS BILLIONS TOWARD AI INFRASTRUCTURE

Meta plans to cut approximately 10% of its workforce starting on May 20. The layoffs will hit teams across Reality Labs, the Facebook social division, recruiting, sales, and global operations. The restructure is being driven by a reallocation toward AI infrastructure spending. Teams will be reorganized into AI-focused 'pods' under Superintelligence Labs.

FULLY REMOTE)

TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. Learn more [15].

ON DWARKESH PATEL'S PODCAST WITH NVIDIA CEO JENSEN HUANG

Most CEO interviews are generally self-serving, and they can be unreliable narrators, but Jensen Huang is usually not like that.

VERCEL CONFIRMS BREACH AS HACKERS CLAIM TO BE SELLING STOLEN DATA

Vercel, known for developing Next.js, says a limited subset of its customers have been affected by a security breach that involves unauthorized access to certain internal Vercel systems.

THOUGHTS AND FEELINGS AROUND CLAUDE DESIGN

Unlike Figma Make, Claude Design is HTML and JS all the way down.

PGQUE

PgQue is a zero-bloat Postgres queue that only requires one SQL file to install.

‘FITBIT AIR' IS A SCREEN-LESS WHOOP COMPETITOR DEBUTING WITH ‘GOOGLE HEALTH' SUBSCRIPTION

The 'Air' moniker likely refers to a thinner design that allows users to wear the device all day.

ROBOTS BEAT HUMAN RECORDS AT BEIJING HALF-MARATHON

The winning runner at a Beijing half-marathon for humanoid robots finished the race today in 50 minutes and 26 seconds.

REGISTER FOR YOUR CITY [5] | THE AC/DC VISION [6] | SCALE WITH SONARQUBE

INTRODUCING CLAUDE DESIGN BY ANTHROPIC LABS

Claude Design is a tool for creating visual work using the Claude Opus 4.7 vision model. It allows users to design prototypes, pitch decks, and marketing materials, automating brand consistency and enabling collaboration. The tool integrates with Claude Code for seamless transition from prototype to production.

CURSOR IN TALKS TO RAISE $2B+ AT $50B VALUATION AS ENTERPRISE GROWTH SURGES

Cursor, an AI coding startup, is close to raising $2 billion in funding, potentially doubling its valuation to $50 billion. Thrive and Andreessen Horowitz are expected to lead the round, with Battery Ventures and Nvidia possibly participating. Cursor aims to triple its annualized revenue to over $6 billion by the end of 2026, driven by its proprietary Composer model and strategic shifts to achieve profitability.

KEVIN WEIL AND BILL PEEBLES EXIT OPENAI AS COMPANY CONTINUES TO SHED ‘SIDE QUESTS'

Two of the architects of OpenAI's most ambitious moonshots, Kevin Weil and Bill Peebles, are leaving the company. Weil led the company's science research initiative while Peebles was the researcher behind Sora. Both announced their departures on Friday. OpenAI recently decided to cut back on 'side quests' as it consolidates around enterprise AI and its forthcoming superapp.

ARE THE COSTS OF AI AGENTS ALSO RISING EXPONENTIALLY?

The length of tasks AI agents can perform has been growing exponentially over the last seven years. The latest models can (sometimes) do tasks that would take a human a few hours. However, the cost to achieve these time horizons is growing exponentially, and the hourly costs for some models are now close to human costs. There will eventually be a divergence between what time horizon is possible and what is economically feasible.

BUILDING A FAST MULTILINGUAL OCR MODEL WITH SYNTHETIC DATA

NEMOTRON OCR V2, developed using synthetic data, is a fast, accurate multilingual OCR model achieving significant accuracy improvements, lowering NED scores to near-zero for non-English languages. Utilizing a synthetic data pipeline with mOSCAR text and diverse fonts, the model trains with pixel-perfect annotations across languages, enabling generalization to real-world documents. The unified architecture reuses feature maps, achieving speeds of 34.7 pages/second on a single A100 GPU, outperforming specialized models in diverse language OCR tasks.

CHANGES IN THE SYSTEM PROMPT BETWEEN CLAUDE OPUS 4.6 AND 4.7

Anthropic is the only major AI lab to publish the system prompts for its user-facing chat systems. The company just released Opus 4.7, which features an updated system prompt. This author used Claude Code to take the Markdown version of its system prompts, break them into separate documents, and then construct a Git history of those files with fake commit dates. The result is in the post.

THE TWO SIDES OF OPENCLAW

Peter Steinberger's talks highlighted OpenClaw's contrasting stories: an inspiring public narrative and a more serious engineering perspective on security and scaling challenges. Anthropic launched Claude Design to compete with tools like Figma, expanding beyond chat/coding into design and prototyping with its OPUS 4.7 engine, which received mixed initial user reviews despite strong benchmark performances. Open-source agent stacks like Hermes and advancements in computer-use UX are gaining traction, while applied AI advances in science, medicine, and infrastructure underscore the sector's dynamic growth.

See how it works

EXPERIMENTAL HYBRID INFERENCE AND NEW GEMINI MODELS FOR ANDROID

Hybrid inference for Android is a new API for Firebase AI Logic that leverages both on-device and Cloud inference. It supports Google's new Gemini models, including the latest image generation Nano Banana models. The new API allows apps to dynamically switch between Gemini Nano running locally on-device and cloud-hosted Gemini models. It is still experimental.

XAI LAUNCHES GROK STT AND TTS APIS

xAI has launched standalone Grok Speech to Text (STT) and Text to Speech (TTS) APIs, enhancing developer options for integrating advanced speech capabilities. These APIs support high accuracy, low latency, word-level timestamps, speaker diarization, and intelligent inverse text normalization across 25+ languages. Grok STT excels in transcription accuracy within phone calls and video/podcasts, providing robust solutions for medical, legal, and financial sectors.

PREFILL-AS-A-SERVICE: KVCACHE OF NEXT-GENERATION MODELS COULD GO CROSS-DATACENTER

Prefill-as-a-Service (PrfaaS) is a cross-datacenter serving architecture that selectively offloads long-context prefill to standalone, compute-dense prefill clusters and transfers the resulting KVCache over commodity Ethernet to local PD clusters for decode. It combines model-side KV efficiency with system-side selective offloading, bandwidth-aware scheduling, and cache-aware request placement. The design removes the requirement that heterogeneous accelerators share the same low-latency RDMA fabric and enables independent scaling of prefill and decode capacity across loosely coupled clusters. A PrfaaS-augmented heterogeneous deployment achieves higher serving throughput while consuming only modest cross-datacenter bandwidth.

DIVE INTO CLAUDE CODE: THE DESIGN SPACE OF TODAY'S AND FUTURE AI AGENT SYSTEMS

Claude Code is an agentic coding tool that can run shell commands, edit files, and call external services on behalf of the user. This study describes its architecture by analyzing the publicly available TypeScript source code and further comparing it with OpenClaw. The core of the system is a simple while-loop that calls the model, runs tools, and repeats. OpenClaw uses the same recurring design questions but produces different architectural answers when the deployment context changes. The study also identifies six open design directions for future agent systems.

GOOGLE IS IN TALKS WITH MARVELL TO BUILD CUSTOM AI INFERENCE CHIPS AS IT DIVERSIFIES BEYOND BROADCOM

Google is in talks with Marvell Technology to develop a memory processing unit and an inference-optimized TPU. The custom ASIC market is projected to grow 45% in 2026 and reach $118 billion by 2033. Broadcom, Google's primary custom chip partner, recently announced a long-term agreement to design and supply TPUs and networking components through to 2031. Talks with Marvell have yet to produce a signed contract.

BETTER AI MODELS ENABLE MORE AMBITIOUS WORK

Better AI models, like Opus 4.5 and GPT-5.2, increased developer AI usage by 44%, enabling more complex tasks after an initial adjustment period. Industries including media and advertising saw notable usage growth, driven by competitive pressures and new opportunities. Developers shifted toward managing AI output, with significant increases in documentation, architecture, and learning tasks.

COMPOSING A SEARCH ENGINE

Canon defines its search pipeline as a DAG to ensure automatic concurrency across classification, localization, retrieval, and ranking processes. This setup allows independent node control, enabling easy subsystem changes without affecting the overall pipeline. The DAG framework optimizes execution through parallelism, supports durable execution with node-specific retries, and provides clear introspection and execution decoupling.

FULLY REMOTE)

TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. Learn more [25].

CANVA STARTS PREVIEWING A MORE POWERFUL VERSION OF ITS AI ASSISTANT

Canva AI 2.0 is available as a research preview starting today to the first one million people who visit the Canva website.

GOOGLE TESTS GOOGLE AI SUBSCRIPTION SUPPORT FOR AI STUDIO

Google is testing a bridge between Gemini subscriptions and AI Studio, offering users subscription-based token access instead of separate API-key billing.

MULTIMODAL 3D WORLD MODEL

Tencent released a multimodal framework that generated and reconstructed 3D worlds from text, images, and video using a staged pipeline and a unified feed-forward model.

INTRODUCING CLAUDE DESIGN BY ANTHROPIC LABS

Claude Design is a new product by Anthropic that allows users to collaborate with Claude to create visual work like designs, prototypes, and presentations using the Claude Opus 4.7 vision model. The tool is available in research preview for Claude Pro, Max, Team, and Enterprise subscribers and can automatically apply team design systems, import from various sources, and export to multiple formats. Teams are using it for realistic prototypes, product wireframes, design explorations, pitch decks, and marketing materials with natural conversation-based refinement.

REDDIT'S IOS APP IS ADDING LIQUID GLASS SOON, HERE'S HOW IT LOOKS

The latest beta version of Reddit's iOS app introduces support for Apple's Liquid Glass design, a visual style first launched with iOS 26 that is gradually being adopted by third-party apps. The update, currently available via TestFlight, showcases a polished implementation that aligns Reddit with Apple's evolving design language. While the feature is still in beta, early impressions suggest it is stable and could roll out to all users in the near future.

IOS 26.4 ADDS BRAND NEW WIDGETS FOR YOUR IPHONE'S HOME SCREEN

iOS 26.4 introduces new Ambient Music widgets that allow users to instantly start mood-based playlists—such as Chill, Productivity, Sleep, and Wellbeing—directly from the Home Screen, with options for both a single-playlist widget and a larger version displaying all four moods. Users can customize each mood with Apple's built-in playlists or their own, but the feature is limited to Apple Music and does not support third-party services like Spotify.

OH, BUT THERE'S ONE MORE THING

AI can produce convincing outputs, but it lacks the “taste” needed to choose meaningful problems, apply cultural and domain judgment, and guide ideas toward truly innovative outcomes—something exemplified by Steve Jobs' approach at Apple. As a result, designers' key role is defining problem spaces and making strategic decisions, positioning them above AI as judgment-driven leaders rather than tool operators.

YOUR AI AGENT CAN READ YOUR CODEBASE. IT DOESN'T KNOW YOUR PRODUCT

AI coding agents can read codebases well but still produce generic results because they lack a product's design context—its behavior, principles, voice, and identity—which mostly exists outside the code. By structuring this missing context into a reusable “skill” (covering areas like UX principles, brand voice, and positioning), outputs become more aligned with the product, reducing generic solutions. However, building this system is difficult because the most important part—design judgment and taste—can't be auto-generated and must be explicitly defined and continuously maintained by humans.

AI AGENT ORCHESTRATION: THE NEW DESIGN SKILL

AI agent orchestration has become a core design skill as the barrier between design and engineering has collapsed, with designers now able to build apps in plain language rather than complex coding syntax. In this article, a designer reveals how he shipped three apps in 2025 by treating AI as a collaborative partner, providing detailed context and intent rather than simple commands. The process involves four phases: accepting basic outputs, learning to debug through conversation, applying systems thinking with structured documentation, and knowing when to ship and iterate.

AI VIDEO CLIPPING WITH EDITORIAL CONTROL

Turn long-form videos into publish-ready clips with sentence-level editing.

AI ROOM DESIGNER

Transform any room instantly with AI. Upload a photo, choose a style, and visualize your future home with photorealistic precision.

LESS ADMIN. MORE CREATIVE FLOW.

One workspace built specifically for freelance creative professionals who run the whole project.

FOLDER INSTRUCTIONS — INSTRUCTIONS FOR SYSTEM-LEVEL AI

AI assistants are becoming more powerful by organizing work within file systems, turning folders into persistent, context-rich environments rather than just storage. Building on this, the concept of “Folder Instructions” lets users define a folder's purpose and behavior—automating tasks like organizing, sourcing, or acting on files—so the system continuously manages itself without constant prompts. This shifts interaction from chat-based commands to structured, always-on workflows, though it introduces challenges like higher computational demands and the need for clear, well-defined rules.