Search the archive

Read the ThreatSpike case study.

HOW I USE CLAUDE CODE FOR WORK, FUN, AND AS A TEXT EDITOR

This article looks at several use cases for Claude Code. These include vibe coding CRUD, using agents to control a VPS, migrating real production projects, and other projects. Claude Code can even be used as a text editor - this opens up a lot of new design possibilities.

REFLECTIONS ON THE REACT COMMUNITY

React is boring tech at this point. It has one of the largest ecosystems of developers, packages, and educational content for building on the web. The composition model of React was the right architecture choice, and it continues to have benefits. However, the React of today doesn't feel like the React of 2015-2020. This article looks at what has changed and makes suggestions on how to grow the community today.

THE ENGINEERING MARVEL THAT CHINA HOPES WILL HELP WEAN IT OFF FOREIGN ENERGY

China has started building a giant hydropower project at the edge of the Tibetan plateau. The $167 billion facility could generate triple the output of the world's largest hydroelectric facility, China's Three Gorges Dam, which can power 40 million Chinese homes. China is determined to become self-reliant in areas of national importance. It imported nearly a quarter of its energy supply in 2023, a level of dependency that it is working hard to undo. The project has stirred accusations that it will give China power to strangle resources beyond its borders and raised fears of environmental damage.

IT'S NOT 10X. IT'S 36X - THIS IS WHAT IT LOOKS LIKE TO KILL A $30K MEETING WITH AI

Improvements in AI come from embracing constraints. Context engineering is about understanding the shape of a problem and working with constraints rather than against them. The constraint is the feature, not the enemy. You can become significantly more efficient not by making AI smarter, but by understanding why your current approach fights human nature, and stopping that fight.

TESLA FSD V14 GETS TENTATIVE RELEASE DATE

Elon Musk stated during Tesla's second-quarter earnings call that he believes Unsupervised FSD will be available for consumers in certain geographies by the end of the year.

RUMORED $599-$699 MACBOOK COULD BE IN THE WORKS — 12.9-INCH MODEL WITH IPHONE 16 PRO SOC TIPPED FOR Q3 2025 PRODUCTION

The cheaper MacBook will apparently use a 12.9-inch diagonal screen and have limited RAM and storage options.

GM PLANS RENEWED DRIVERLESS-CAR PUSH AFTER CRUISE DEBACLE

GM's new project will be focused on autonomous cars for personal use rather than a robotaxi service.

STRIPE BUILDS PAYMENTS BLOCKCHAIN ‘TEMPO' WITH PARADIGM IN CRYPTO EXPANSION

Stripe's Tempo will be a high-performance, payments-based Layer 1 blockchain that will be compatible with Ethereum's programming language.

WRITING A BRAND-NEW OS IS ALMOST IMPOSSIBLE BY NOW

Creating a consumer-friendly, fully working OS from scratch also involves getting the OS adopted, which means convincing both users and developers of its unique features.

📥 Get the whitepaper

OPENAI'S REASONING SYSTEM WINS GOLD AT INTERNATIONAL OLYMPIAD IN INFORMATICS

OpenAI's reasoning models placed first among AI participants and outperformed 325 of 330 human competitors at the prestigious programming olympiad. The system operated under identical constraints as the humans and used general-purpose models without competition-specific training. It jumped from last year's near-bronze to 98th percentile performance.

NVIDIA'S PUSH ON PHYSICAL AI

NVIDIA unveiled innovations in neural rendering, 3D generation, simulation, and reasoning models for physical AI at SIGGRAPH 2025. Highlights included new Omniverse NuRec libraries for large-scale reconstruction, Cosmos Reason VLM for physics-aware reasoning, and updates to the Metropolis vision AI platform.

ANTHROPIC'S CLAUDE CHATBOT CAN NOW REMEMBER YOUR PAST CONVERSATIONS

Anthropic released a memory function for its Claude chatbot that allows users to recall past conversations on demand. The feature supports web, desktop, and mobile, and is currently available to Max, Team, and Enterprise subscribers. Unlike OpenAI's ChatGPT, Claude doesn't build a user profile and retrieves past chats only when requested.

GPT-5S ARE ALIVE: BASIC FACTS, BENCHMARKS, AND THE MODEL CARD

OpenAI released several good models at once: GPT-5, GPT-5-Thinking, GPT-5-With-The-Router, GPT-5-Pro, and GPT-5-API. These models cut down on errors and hallucinations and are easier to use. This post provides an introduction to these models, including basic facts, benchmarks, and the model card. GPT-5-Thinking and GPT-5-Pro look like substantial upgrades to o3 and o3-Pro.

DEEPMIND CEO ON WORLD MODELS AND AI BENCHMARKS (31 MINUTE VIDEO)

Demis Hassabis discusses advances like Genie 3's world model capabilities, the Deep Think features in Gemini 2.5, and Kaggle's Game Arena as tools for pushing AI systems closer to AGI-level reasoning and understanding.

HOW MUCH POWER WILL FRONTIER AI TRAINING DEMAND IN 2030?

Epoch AI estimates the largest AI training runs will require 4-16 gigawatts by 2030, enough to power millions of homes, if current trends hold. US AI data center capacity would need to increase 10x over the next 5 years to 50 gigawatts to support this scale, roughly 5% of the nation's total power generation.

AN ESSENTIAL PRIMER FOR BUILDING ENTERPRISE-READY MCP SERVERS

MCP doesn't require compromising security - it enables security to be applied intelligently so that teams can embrace AI safely, confidentially, and without unnecessary barriers. Overly restrictive access slows innovation, frustrates developers, and prevents teams from realizing the full potential of MCP-based integrations. Organizations should already have guardrails proven to protect critical systems - they should have faith in those foundations.

FORWARD DEPLOYED AI RESEARCH

Forward Deployed Research is the discipline of living with a problem, building just enough to learn, throwing it into the wild, and pulling the feedback directly into the next version. It is where the real progress in AI is happening now: the people pushing the boundaries refuse to accept the traditional divide between research and deployment. AI may cause the distinction between scientist and engineer to blur.

LANGDIFF

Modern AI applications increasingly rely on LLMs to generate structured data, but streaming these outputs poses unique challenges. LangDiff is a Python library that helps stream structured LLM outputs to frontends. It can be used to build responsive AI applications where the backend structures and frontend experiences can evolve independently.

PHYSICALLY CONTROLLABLE RELIGHTING OF PHOTOGRAPHS

This study presents a self-supervised approach to in-the-wild image relighting that enables fully controllable, physically based illumination editing. The approach combines the physical accuracy of traditional rendering with the photorealistic appearance made possible by neural rendering. It represents a significant step in bringing the explicit physical control over lights available in typical 3D computer graphics tools to in-the-wild relighting. A 5-minute-long video explaining the study with examples is available.

HOW BENCHMAXXED IS GPT-OSS-120B?

An analysis of currently available cutting-edge open source models found that gpt-oss-120b scored the worst on LiveBench, a private benchmark that uses secret questions. While the model looked good on release day, a bunch of critical reports have since been released. This isn't conclusive evidence that OpenAI is gaming benchmarks, but it brings up the question of whether it is worth creating meta-benchmarks that try to measure overfitting in more robust ways.

ELON MUSK'S XAI RELEASES GROK 4 FOR FREE GLOBALLY, CHALLENGES OPENAI'S GPT-5 LAUNCH

xAI's Grok 4 is now freely accessible to all users worldwide. Free users have generous usage limits for a limited time. Grok 4 features an Auto Mode, where the AI decides if a user prompt requires deeper reasoning or a simple response, and an Expert Mode, which allows users to manually trigger more in-depth answers. Access to Grok 4 Heavy remains exclusive to SuperGrok Heavy subscribers.

HERE IS HOW OPENAI WILL PRIORITIZE COMPUTE OVER THE NEXT COUPLE OF MONTHS

OpenAI will first make sure that current paying ChatGPT users get more total usage than they did before GPT-5, then it will prioritize API demand up to the allocated capacity and commitments it has made to customers, increase the quality of the free tier of ChatGPT, and prioritize new API demand.

BILLIONS FLOW TO NEW HEDGE FUNDS FOCUSED ON AI-RELATED BETS

AI-focused investors argue the long-term trend of development and adoption are inevitable.

ON PRICING

OpenAI may have erased a bunch of market cap by pricing ChatGPT at $20.

PAXOS AIMS FOR NATIONAL BANK CHARTER

Paxos is applying for a national trust bank charter with the OCC, joining Ripple and Circle in seeking federal oversight as the GENIUS Act reshapes stablecoin regulation. The issuer of PayPal USD and Pax Dollar says the move will reinforce its commitment to safety and transparency, but comes just a week after settling with NYDFS over past AML and due diligence lapses tied to its Binance partnership.

ETH TOPS BTC IN YTD GAINS AMID ETF AND TREASURY DEMAND

ETH has overtaken BTC in year-to-date gains, rising 29% compared to Bitcoin's 28%, as institutional demand pushes Ethereum above $4,000 for the first time since December. The increase has been sparked by $5B in recent inflows to spot ETH ETFs and growing allocations from digital asset treasury firms mimicking Strategy's Bitcoin playbook.

STRIPE BUILDING ‘TEMPO' BLOCKCHAIN WITH CRYPTO VC PARADIGM

Stripe is developing a payments-focused layer-1 blockchain called “Tempo,” per a now-removed product marketing job posting that described a stealth five-person team working with Paradigm. The chain is rumored to be EVM-compatible, and follows Stripe's recent stablecoin push, including the $1.1B acquisition of Bridge and the purchase of wallet developer Privy, positioning Stripe to own more of the stablecoin stack from issuance to transaction processing.

SHARPLINK RAISES $900M TO BACK ETH STRATEGY

SharpLink secured nearly $900 million in new capital last week, signaling strong market conviction in its Ethereum-focused treasury strategy. The raise highlights growing institutional alignment around ETH's long-term role in digital finance.

STRING IT TOGETHER: HOW STABLECOINS GO MAINSTREAM

Stablecoins can achieve mainstream US consumer payments adoption by repeating credit cards' two-step path: first win niche segments on intrinsic advantages like comparative convenience and powerful incentives, then connect those “fiefdoms” via an open, standardized, participant-owned network. Near-term wedges include cross-border shoppers underserved by dollar rails and incentive-first programs where merchants launch transferable, whitelabeled stablecoins funded by float yield, banks' ability to custody stables, and UX improvements like account abstraction and gas sponsorship. Programmability enables novel rewards and financing models that can pull balances into stables and drive spend if checkout is seamless. To scale, stablecoins need a neutral, user-owned network to enforce interoperability, compliance, and consumer protections across chains and issuers.

REFLECTIONS, RANT, AND ROAD AHEAD FOR CRYPTO X AI AGENTS

Phase 1 of crypto x AI saw speculative euphoria with a market cap topping $10B, followed by a collapse where liquidity rotated out, tokens fell 50-90%, teams went silent, and “L1 for agents” frameworks/launchpads plus chat UIs failed to deliver production utility. Today, chat front ends remain too slow/error-prone for transactions, and many Solana AI tokens trade as memecoins, but improving agentic models, MCP standardization, and walleted agents are emerging as workable primitives. Over the next 6-12 months, viable use cases are expected in complex agentic workflows rather than simple swaps. In 1+ years, agents could drive stablecoin payments with x402, enable crypto-native revenue sharing, spur AI-embedded protocols, verifiability/coordination networks, and an on-chain personal context layer.

DESIGN TESTNETS FOR FEEDBACK, NOT FARMING

Fluent is launching a testnet experience designed around collecting app feedback rather than traditional "chain farming," arguing that current testnets have devolved into spam-filled environments where users perform mindless tasks hoping for token rewards. The company criticizes how chains typically optimize for vanity metrics like transaction volume and user counts instead of fostering the development of useful applications that drive long-term success. Fluent's approach focuses on helping builders iterate toward product-market fit by systematically collecting and analyzing user feedback across areas like usability, engagement, and performance. This "apps first" philosophy positions chains as infrastructure that succeeds when the applications built on them succeed, similar to how AWS benefits from Netflix's success.

STRIPE'S STABLECOIN STRATEGY CULMINATES IN NEW L1

Stripe's stablecoin journey has rapidly evolved from basic onramps in 2022 to launching its own L1 blockchain, “Tempo,” in 2025. After acquiring Bridge Stablecoin and Privy for over $1B each, Stripe is now positioned to vertically integrate payments, wallets, and chain infrastructure.

COINBASE VENTURES HOLDING TON TOKEN

Coinbase Ventures has taken a liquid token position in the TON token, which is the canonical blockchain for Telegram.

HIGHEST-PAYING CRYPTO JOBS

Crypto x AI jobs and infrastructure projects pay the most of any sub-industry in crypto, with median salaries of $400k and $288k, respectively.

ETH FUTURES VOLUME HITS ALL-TIME HIGH ON CME

CME's Ethereum futures hit a record $118B in trading volume for July, up 82% month-over-month, with open interest jumping 75% to $5.21B.

DO KWON MAY PLEAD GUILTY IN U.S. FRAUD CASE

Kwon faces multiple charges, including wire fraud, securities fraud, and market manipulation.

HERE'S THE REDESIGNED AIRDROP ICON IN IOS 26

Apple has redesigned the AirDrop icon to match the new glass-like style in iOS 26 beta 5, replacing the white background with a fully blue icon featuring layered opaque glass to create a more physical radar effect. The updated icon appears in the share sheet on iPhone, iPad, and in macOS Tahoe 26 beta. The old version still shows in Finder search.

FRAMER ON-PAGE EDITING

Framer's On-Page Editing enables team members to update websites directly from the browser. The feature allows editing of text, images, and components with instant syncing to the Framer project. Anyone can contribute to site updates without design experience by simply clicking Edit on the live site.

APPLE APPLIES FOR PATENT ON THAT ‘SINGLE SLAB OF GLASS' IPHONE

Apple is pursuing Jony Ive's vision of an iPhone that looks like a single slab of glass. The company has filed a new patent describing a six-sided glass enclosure that displays content and can detect touch on all sides. While still far from production, the design aims for a seamless, bezel-free look. The iPhone 17 Air is expected to be the thinnest step toward that goal.

DESIGN SYSTEM TEAMS ARE DYING

Companies fall into different categories based on whether their design systems serve products or vice versa, affecting shipping speed. AI will eliminate traditional design system teams within two years by automating consistency and maintenance tasks. Successful companies embed system expertise directly into product teams while using AI to maintain brand and quality standards.

THE DUOLINGO METHOD: COLLABORATION AS A CORE PRACTICE

Duolingo's Math team abandoned traditional design-to-engineering handoffs in favor of collaborative co-creation throughout the product development process. The team ideates together in shared files, builds rapid prototypes, and iterates constantly using a "show don't tell" philosophy. This collaborative approach enables faster decision-making, maintains quality standards, and builds trust while developing entirely new educational games and modules.

I'VE DESIGNED AI ASSISTANTS — HERE'S WHAT ACTUALLY WORKS

Effective AI assistant design requires understanding users' specific jobs-to-be-done rather than creating generic ChatGPT clones. Success depends on building reusable UI components like context-aware prompts, visual activity indicators, and structured response blocks. Trust emerges through transparency about AI capabilities, designing for unpredictable responses, and maintaining user control over automated actions.

THE HUMAN-AI CREATIVE PLATFORM

Instantly turn ideas into images, videos, 3D, and immersive worlds. Haimeta provides pre-designed templates and intuitive scenarios that guide you through the creative process.

UI COMPONENTS BUILT WITH ARK UI AND TAILWIND

Tark UI is an open-source collection of 300+ copy-and-paste components for quickly building application UIs.

BUILD INTERACTIVE TOOLS THAT CONVERT

Instantly generate, enhance, and refine embeddable experiences with advanced AI by your side.