Search the archive

WE'VE BEEN BUILDING AN INTERNAL CLAUDE CODE PLUGIN SYSTEM

Intercom's internal Claude Code plugin system has 13 plugins, over 100 skills, and hooks that turn Claude into a full-stack engineering platform.

SOFTWARE IS EATING ITSELF

Software has always wanted to be free, and it eventually will be, so the business of software needs a different approach.

A SUFFICIENTLY DETAILED SPEC IS CODE

Coding agents can't reliably fill in missing clarity and detail as they aren't mind readers, and even if they were, there's not much they can do if your own thoughts are confused.

FULLY REMOTE)

As the first engineer on TLDR's new Applied AI team, you'll build AI agents to automate end-to-end workflows and lead our context engineering project to allow non-technical teammates create their own AI workflows. Learn more [19].

GOOGLE ENGINEERS LAUNCH "SASHIKO" FOR AGENTIC AI CODE REVIEW OF THE LINUX KERNEL

Sashiko ingests patches from mailing lists, analyzes them using AI-powered prompts, and provides feedback to help maintainers and developers ensure code quality and adherence to kernel standards.

META WILL PAY INSTAGRAM, TIKTOK, AND YOUTUBE CREATORS WITH BIG FOLLOWINGS TO POST ON FACEBOOK

Meta's new Creator Fast Track program offers creators with an established following guaranteed monthly payments and increased reach on Facebook.

WE HAVE LEARNED NOTHING

Any time someone insists on a method for how to build a successful startup, you should do something different.

Solutions

HOW CHINA IS GETTING EVERYONE ON OPENCLAW, FROM GEARHEADS TO GRANDMAS

China aggressively promotes OpenClaw, a popular AI assistant, through public events organized by major tech firms like Baidu and Tencent. This AI tool enables personal automation, supporting China's vision for widespread AI integration by 2030.

MINIMAX LAUNCHES M2.7 MODEL ON MINIMAX AGENT AND APIS

MiniMax's M2.7 model is now publicly available via the MiniMax Agent and the MiniMax API Platform. It supports complex workflows in software engineering, office productivity, and research environments. The model features capabilities like autonomous debugging and research agent harnesses. MiniMax's release marks a shift towards models that participate in their own evolution.

XIAOMI STUNS WITH NEW MIMO-V2-PRO LLM NEARING GPT-5.2, OPUS 4.6 PERFORMANCE AT A FRACTION OF THE COST

Xiaomi's MiMo-V2-Pro is a 1-trillion-parameter foundation model with performance approaching that of models from OpenAI and Anthropic but at a fraction of the cost. The model uses a sparse architecture that only activates 42 billion parameters during any single forward pass. It has a Multi-Token Prediction layer that allows it to anticipate and generate multiple tokens simultaneously, drastically reducing the latency required for 'thinking'. The model is currently only available via Xiaomi's first-party API. Xiaomi plans to release an open source variant of the model.

WHAT 81,000 PEOPLE WANT FROM AI

Anthropic conducted a large-scale study with 80,508 global participants to understand their hopes and fears about AI usage. Key desires include professional excellence and life improvements, while main concerns revolve around AI unreliability and job displacement. People experience AI as both a productivity tool and a potential dependency, highlighting its dual role in enhancing and complicating life.

GPT 5.4 IS A BIG STEP FOR CODEX

GPT 5.4 brings a ton more simple usability and 'agentness'. It is the first OpenAI agent that feels like it can do a lot of random tasks. The instruction following of the model is precise, and the model just does what you tell it to do. GPT 5.4 will likely appeal to the master coordinator who wants to unleash an AI army on distributed tasks.

HOW DID ANTHROPIC DO IT?

Anthropic was popular early on with the earliest adopters. The company is now leveraging that early-adopter base to go mainstream. Data suggests that the company's tools are becoming the consumer default. Anthropic still struggles to meet its own demand and is actively turning away revenue because it doesn't have the compute to serve it. Demand for its products is growing despite the company charging more for roughly equivalent performance.

legacy stack look like a filing cabinet. Read the report.

INTRODUCING THE MACHINE PAYMENTS PROTOCOL

Stripe announced support for programmable LLMs, enabling businesses to create customized payment experiences. This feature allows companies to integrate AI models into their payment processes for enhanced user interactions. Stripe's new LLM support aims to streamline and personalize the payment journey for users.

AGENT PACKAGE MANAGER

Microsoft's Agent Package Manager is an open source, community-driven dependency manager for AI agents. It allows every developer who clones a repository to get a fully configured agent setup in seconds. Developers just have to declare their project's agentic dependencies once in a YML file. The package manager works with GitHub Copilot, Claude Code, Cursor, and OpenCode.

ENTERPRISE VISION-LANGUAGE MODELS

Baidu's Qianfan‑VL is a series of enterprise-focused vision‑language models optimized for industrial use cases such as document parsing, OCR, and complex visual reasoning while maintaining general multimodal capabilities.

THE INTERNET RUINED CUSTOMER SERVICE. AI COULD SAVE IT.

AI can revolutionize customer service by offering personalized, scalable, and concierge-like experiences for all consumers, bridging the gap between luxury services and mass-market customer support. Decagon, a leading AI platform, enhances customer service by resolving over 80% of inquiries autonomously, boosting satisfaction, and drastically reducing costs. If successful, this model transforms customer interaction into a continuous, proactive relationship, fundamentally changing the way businesses engage with consumers worldwide.

HOW TO MAKE SENSE OF AI

Effective AI sensemaking involves focusing solely on detailed field reports while ignoring opinions, predictions, and speculative essays. The method emphasizes four key questions: new outcomes, potential actions, relative outcome values, and causal relationships.

EXPLORE OPENSEARCH

GREETINGS, EARTHLINGS: PHILIP JOHNSTON OF STARCLOUD ON DATA CENTERS IN SPACE

Starcloud's approach utilizes space-based solar power, avoiding terrestrial transmission losses and permitting delays.

COMET BROWSER ON IOS

Perplexity Comet for iOS brings Perplexity's AI‑native browser to iOS with voice interaction, hybrid search results, and an in‑browser assistant that can answer questions about open pages or help complete tasks.

HOW A SMALL TEAM SCALED AI INFRASTRUCTURE

A six‑engineer team scaled an AI platform serving millions of businesses by consolidating infrastructure into a single codebase and platform to simplify operations and support rapid growth.

Read the blog

SEC ISSUES TOKEN TAXONOMY FOR CRYPTO ASSETS

The SEC released interpretive guidance on March 17 that establishes a four-category token taxonomy covering digital securities, digital commodities, digital collectibles, and digital tools/stablecoins, formally defining how federal securities laws apply to each class of crypto asset. Chairman Atkins stated that "most crypto assets are not themselves securities," and the guidance places airdrops, protocol staking, and protocol mining outside SEC jurisdiction, removing a major source of enforcement uncertainty for protocol developers and DAOs. Investment contracts lose their securities classification once an issuer fulfills or fails to satisfy its original representations, a definitional shift with direct consequences for token launch structuring and secondary market compliance. The CFTC co-signed the taxonomy in a coordinated action, and Atkins indicated a formal rulemaking exceeding 400 pages, including an innovation exemption for crypto firms, is expected within weeks.

BITCOIN ETFS POST LONGEST INFLOW STREAK SINCE OCTOBER 2025

US spot Bitcoin ETFs recorded $199.4 million in net inflows on Tuesday, extending the streak to seven consecutive days and attracting roughly $1.17 billion over that span. BlackRock's IBIT led with $169 million, followed by Fidelity's FBTC at $24.4 million, with Ark/21Shares and VanEck also posting gains. The seven-day run is the longest since October 2025, and the funds are on track for a fourth consecutive week of net inflows, the longest weekly streak since September.

MOODY'S BECOMES FIRST CREDIT RATING AGENCY TO BRING ANALYSIS ONCHAIN

Moody's Ratings launched its network-agnostic Token Integration Engine (TIE), becoming the first credit rating agency to deliver independent credit analysis directly within onchain workflows. TIE connects Moody's existing analytical outputs with blockchain networks, allowing authorized institutional users to access credit insights in real time. The initial deployment operates on the Canton Network, built for the privacy and regulatory requirements of institutional finance. Moody's plans to expand TIE to additional blockchain networks, asset classes, and financial instruments over time.

X402 SHIPS SIWX CREDENTIAL-BASED AUTHORIZATION

x402 added credential-based authorization via SIWX (Sign-In-With-X), a CAIP-122 wallet authentication layer that removes the per-request on-chain settlement requirement from V1. V1's settlement latency (1-2 seconds on Base, 400ms on Solana) made high-frequency use cases like LLM inference, data feeds, and multi-tool agent sessions impractical, and the model could not support stateful APIs or websites that fire 20-50 sub-requests per page load. With SIWX, users pay once and authenticate subsequent requests through locally verified wallet signatures in microseconds, with no blockchain round-trip or facilitator, enabling paywalled websites, streaming media, stateful APIs such as Crossmint's embedded wallets, and persistent-identity AI agent workflows. The tradeoff shifts business logic complexity to servers, as each service must define its own access semantics independently.

THE FLATTENING

AI is collapsing information moats by reducing the cost of bridging closed platforms to near zero, converting proprietary APIs into open ones and threatening business models built on data scarcity at companies like Facebook, Yelp, and Google. Crypto offers a tested template for what survives this shift, having always operated with open APIs and been forced to build power through alternative forms of scarcity. The canonical example is onchain exchanges, which accrue value through concentrated liquidity network effects rather than data exclusivity, even as order books and trade history remain open. For traditional internet businesses, the prescription is to anchor value in scarce complements that AI cannot commoditize: physical assets, regulatory position, brand, and trust.

WHY ETHEREUM NEEDS A DYNAMICALLY AVAILABLE PROTOCOL

Ethereum's heartbeat layer should guarantee liveness whenever a majority of online stake is honest, treating offline validators as absent rather than as liveness threats, a property termed dynamic availability that allows the chain to self-recover from client bugs, cloud outages, or adversarial censorship without coordinated social intervention. An availability-finality dilemma, analogous to CAP, makes a single-protocol solution impossible: simultaneously guaranteeing liveness under dynamic participation and safety under network partitions cannot be achieved in one layer, mandating a two-layer design of a dynamically available heartbeat plus a trailing finality gadget. Standard BFT protocols (PBFT, Tendermint, and HotStuff) halt when participation drops below fixed-quorum thresholds, while candidate heartbeat protocols Goldfish and RLMD-GHOST use small per-slot committees (~256 validators) to provide provable dynamic availability. Small committees enable faster slots and eliminate the need for large-scale signature aggregation, opening a near-term path to post-quantum security for the heartbeat layer.

ETHEREUM'S FAST CONFIRMATION RULE CUTS L1 CONFIRMATION TIMES

The Fast Confirmation Rule (FCR) reduces Ethereum L1 confirmation times from roughly 13 minutes to approximately 13 seconds, an 80-98% improvement for most L2s and exchanges. FCR works by counting attestations rather than blocks and requires no hard fork. Consensus layer client teams are already implementing it, and nodes will run the rule automatically once deployed. The system assumes network synchrony and less than 25% adversarial stake, with a built-in fallback to finality-grade confirmation if conditions deteriorate. Exchanges, L2s, bridges, and solvers all stand to benefit from faster deposit times, reduced capital lockup, and a more efficient blob fee market.

WHY RICH PEOPLE DON'T CARRY CASH

The relationship between people and money inverts as wealth scales. It goes from cash-based personal trust at street level to systemic, infrastructure-based trust at the institutional tier. SWIFT processes roughly 45 million messages per day across 11,000 institutions in 200+ countries, while the FICC clears $7 trillion in daily Treasury volume, all largely invisible. Institutions select financial infrastructure based on track record under stress rather than innovation, and positions real-time blockchain settlement as the next evolution of that plumbing.

SEC FORMALLY CLASSIFIES SOL AS A DIGITAL COMMODITY

The SEC has formally classified SOL as a digital commodity in its new crypto asset taxonomy, placing it alongside BTC, ETH, and 14 other assets.

TANGEM DOUBLES ANNUAL REVENUE TO MORE THAN $60 MILLION

Hardware wallet maker Tangem doubled its annual revenue to over $60 million as self-custody adoption shifts from passive storage toward active on-chain use.

POLYGON OUTPACES MAJOR L2S IN TPS

Polygon is currently processing more transactions per second than Base, Arbitrum, MegaETH, and Optimism.

TALLY IS SHUTTING DOWN

Tally, the DAO governance platform, is shutting down after five years and canceling its planned ICO.

Get your free year

FIREFOX TEASES NEW LOGO WITHOUT ICONIC FOX

Firefox has teased a new logo by posting a fox-less version on Instagram, showing only a purple orb. The company promises the fox will return "next week," likely featuring their new mascot character, "Kit," with a visible face design. This logo change is part of a broader brand refresh that includes pastel color palettes and softer, more rounded aesthetics.

GAMMA ADDS AI IMAGE-GENERATION TOOLS IN BID TO TAKE ON CANVA AND ADOBE

Gamma Imagine is an AI image-generation tool that creates brand-specific marketing assets via text prompts, competing with Canva and Adobe. The platform, which raised $68 million at a $2.1 billion valuation, aims to serve knowledge workers who need visual communication tools but lack design resources. Gamma is approaching 100 million users and integrates with tools such as ChatGPT, Claude, and Zapier to power its AI-driven asset-creation features.

IPHONE 18 PRO: THREE NEW DESIGN UPDATES ARE COMING THIS YEAR

Rumors suggest Apple will give the iPhone 18 Pro and Pro Max a more unified rear design (removing the two-tone look), shrink the Dynamic Island by about 35% with partial under-display Face ID, and continue experimenting with bold new colors like brown, purple, or burgundy, possibly without offering a standard black option.

DURABLE PATTERNS IN AI PRODUCT DESIGN

Recurring design patterns across AI product evolution consistently prove their value as interfaces shift from background AI to chat to agentic systems. As AI gains the ability to use tools and orchestrate sub-agents, a core UX challenge becomes balancing transparency with cognitive load, with collapsed, expandable action logs emerging as a reliable solution. Capability awareness, context awareness, and output readability remain the central unsolved problems shaping how AI products are built today.

DETAILS THAT MAKE INTERFACES FEEL BETTER

Great interfaces result from a collection of small details that compound into better user experiences. Key techniques include balanced text wrapping, concentric border radii for nested elements, contextual icon animations, crisp text rendering, and tabular numbers for consistent digit spacing. Making animations interruptible and using CSS transitions instead of keyframe animations creates more responsive interfaces that adapt to user behavior changes.

A DESIGNER'S FIELD REPORT ON THE ICONIC BLIND SPOT IN AI WORLD MODELS

Experiments with Gemini and other leading AI systems highlight a core limitation of current LLMs: they lack true spatial and physical understanding because they are not grounded in real-world, embodied experience. Failures in generating and reasoning about simple physical scenarios reveal three key gaps—Continuity (no coherent 3D spatial model), Gravity and Physics (no sense of physical constraints), and Reversibility of Thought (inability to reset or backtrack reasoning), showing that modern AI operates as a top-heavy symbolic system that can produce fluent outputs but often lacks structural correctness.

AI VIDEO AND IMAGE GENERATOR

Create professional-quality videos and images in seconds with Crano.