Search the archive

judgment works when it's designed, not assumed. Talk to an expert

INTELLIGENT AI DELEGATION

AI agents can tackle increasingly complex tasks, but to achieve more ambitious goals, they need to be able to meaningfully decompose problems into manageable sub-components and safely delegate their completion across to other AI agents and humans. This study proposes an adaptive framework for intelligent AI delegation. The proposed framework is applicable to both human and AI delegators and delegatees in complex delegation networks.

PERSONAPLEX: VOICE AND ROLE CONTROL FOR FULL DUPLEX CONVERSATIONAL SPEECH MODELS

Personaplex is a real-time speech-to-speech conversational model that jointly performs streaming speech understanding and speech generation. It operates on continuous audio and can predict both text tokens and audio tokens auto-regressively to produce its spoken responses. Personaplex's design allows the model to update its internal state based on users' ongoing speech while still producing fluent output audio, so it can support highly interactive conversations. The model is ready for commercial use.

THE RL ARCHITECTURE BEHIND MINIMAX M2.5, EXPLAINED CLEARLY

Minimax M2.5 is fast, cheap, and great at coding. While the team behind the model has released a technical article on how the model works, it is quite dense. This post aims to make the content of that article more accessible for readers. The core problem the team is solving is how to make RL work at massive scale for being good at Agentic AI. Training LLMs to be good at agentic tasks imposes many challenges: huge volumes of training data need to be processed quickly, training needs to be stable, and agents need to be good at a diverse range of tasks.

COMPOSED PROMPTS WITH RLVR

Composition-RL reused “too-easy” pass-rate-1 prompts by automatically combining multiple problems into new verifiable questions for RLVR training. Across 4B–30B models, the approach improved reasoning over training on the original dataset.

NOBODY IS TALKING ABOUT GENERALIZED HILL-CLIMBING (AT RUNTIME)

You have to know where you're going before you take that first step. The first thing to do is always to define your goal. Reverse engineer requests and combine that with context to create discrete, boolean, testable Ideal State Criteria. This criterion can carry through to become the verification criteria as well.

PENTAGON USED ANTHROPIC'S CLAUDE IN MADURO VENEZUELA RAID

Anthropic's partnership with Palantir means that the Defense Department and federal law enforcement have access to Claude. This means it was possible that the technology was used in matters related to the US military's operation to capture former Venezuelan President Nicolás Maduro. Anthropic's usage guidelines prohibit Claude from being used to facilitate violence, develop weapons, or conduct surveillance. A company spokesperson said that Anthropic has not discussed the use of Claude for specific operations, but the company is committed to using frontier AI to support the US in national security.

MANUS AI LAUNCHED 24/7 AGENT VIA TELEGRAM AND GOT SUSPENDED

Manus AI launched a 24/7 Agent via Telegram but faced immediate suspension without explanation.

MICROSOFT'S AI CHIEF TARGETS AI SELF-SUFFICIENCY AND OPENAI INDEPENDENCE

Microsoft plans to reduce its reliance on OpenAI by developing its own AI models, spearheaded by AI chief Mustafa Suleyman.

SOFTWARE AS WIKI, MUTABLE SOFTWARE

An example of how an agent can be used to edit parts of software like a wiki.

NVIDIA, GROQ, AND THE LIMESTONE RACE TO REAL-TIME AI: WHY ENTERPRISES WIN OR LOSE HERE

The exponential growth of AI is due to a series of bottlenecks being smashed, and Groq's LPU is the next step.

FREE MODELS ROUTER

openrouter/free is a router that selects free models at random from the models available on OpenRouter.

TOMASZ STANCZAK STEPS DOWN FROM ETHEREUM FOUNDATION

Tomasz Stanczak is stepping down from his co-Executive Director role at the Ethereum Foundation after less than a year. Bastian Aue has been named interim co-ED alongside Hsiao-Wei Wang. Stanczak, co-founder of Nethermind, had been pushing internal reform but faced friction over disclosed conflicts of interest with Nethermind, Flashbots, and VC ties. Industry commentary suggests the departure reflects power struggles within the foundation.

COINBASE Q4 REPORT

Coinbase has reported record trading volumes alongside 9 consecutive quarters of native unit growth, all-time high USDC market cap and on-platform balances, and 4x increase in US derivatives market share following its 24/7 perps launch and Deribit integration. The exchange has secured partnerships with five major global banks, including JPMorgan, Citi, PNC, and Standard Chartered. Six months after announcing its "Everything Exchange" vision, Coinbase has launched prediction markets, begun offering 24/5 stock trading, and integrated millions of Solana assets into its DEX.

X TO LAUNCH ‘SMART CASHTAGS' FOR REAL-TIME CRYPTO AND STOCK DATA

X will roll out “Smart Cashtags” in the coming weeks, letting users tap tickers in their timeline to view live price charts, related posts, and direct trading links for stocks and onchain assets, including smaller-cap tokens. While X will not execute trades itself, the feature deepens its push into financial services alongside X Money. It comes with strict API rules aimed at curbing crypto spam and bot-driven manipulation.

PREDICTION MARKETS FOR HEDGING

Vitalik Buterin argues prediction markets are converging toward unsustainable "corposlop" (short-term crypto price bets, sports betting) and should prioritize hedging as the core use case. Prediction markets on price indices for all major goods and services could be used as a type of hedged currency, with local LLMs generating personalized baskets representing each user's expected future expenses. The system requires interest-bearing denominations like wrapped stocks or ETH rather than fiat, and could incorporate autonomous agents for more accurate and full-time information gathering.

CRYPTO CAN KILL INTERNET GODS

Tech platforms have built "psychological extraction machines" that monetize user behavior while harvesting data value, creating a digital caste system where users generate content unpaid. Crypto is "Dharma through Code" – cryptography as an immutable mathematical truth that enables decentralization to redistribute power from centralized platforms back to individuals through private key sovereignty. Rather than viewing crypto as a speculative casino, treat it as an ark for building parallel societal infrastructure governed by mathematical laws instead of corporate algorithms.

THEIA - 2025 ANNUAL LETTER

Theia's 2025 annual letter highlights returns generated purely through asset selection in a challenging environment where BTC returned 35%, but median tokens dropped 60-80%. The fund's thesis centers on crypto evolving from speculative "digital gold" to functional financial infrastructure, evidenced by stablecoins facilitating trillions in annual transfers and onchain lending protocols like Maple and Morpho surpassing $100B. Investment focus areas include governance innovation through prediction markets and futarchy (MetaDAO), alongside emerging AI-crypto infrastructure for agent identity, payments, and coordination.

PREDICTION MARKETS MUST STOP CATERING TO “DUMB OPINIONS”

Vitalik Buterin warns prediction markets are “over-converging” on short-term, dopamine-driven bets dominated by uninformed retail gamblers and risks collapse in bear markets, urging platforms to pivot away from speculative gambling. He proposes turning markets into AI-powered hedging tools. These would enable personalized, regionally indexed baskets that offset real-world costs and could be denominated in productive assets, potentially replacing fiat-pegged stablecoins.

EVERYTHING IS MARKET

Markets are evolving from outputs into infrastructure, resembling "markets as APIs" that produce real-time data through adversarial competition that's costly to fake, unlike conventional APIs returning stored information. The pattern is already emerging in production: Coinbase's Borrow/Earn products query Morpho's onchain lending markets for rates, while Polymarket odds appeared at the Golden Globes and correctly predicted 26 of 27 winners. There is a "DeFi mullet" implementation model (fintech front-end, DeFi back-end) using permissionless markets as regulatory forcing functions, with the vision extending to one billion traders plus AI agents consuming markets as programmatic infrastructure.

SOLANA'S 2026 PLAN IS ALL ABOUT THE APPLICATION LAYER

Solana is shifting toward institutional finance.

ETHEREUM AND SOLANA DOMINATE RWA DEFI TVL

Ethereum holds ~57% of RWA DeFi active TVL and Solana about 32%, meaning nearly 90% of tokenized assets in DeFi sit on just two chains.

ETHEREUM HITS RECORD TRANSACTIONS AS FEES COLLAPSE

Ethereum's weekly transaction count has reached an all-time high of 17.3 million while median transaction fees have fallen to an all-time low of $0.008, marking a sharp divergence between usage and cost.

VAULTS: CONNECTING INSTITUTIONAL CAPITAL TO DEFI

BYTEDANCE'S NEXT-GEN AI MODEL CAN GENERATE CLIPS BASED ON TEXT, IMAGES, AUDIO, AND VIDEO

ByteDance has launched Seedance 2.0, an advanced AI video model that can generate 15-second clips with audio from text, images, and videos while accounting for camera movement and visual effects. The model represents ByteDance's entry into the competitive AI video generation market, joining Google's Veo 3, OpenAI's Sora 2, and Runway's latest offerings. Currently available on ByteDance's Dreamina platform and the Doubao AI assistant, the tool has already sparked concerns about copyright infringement and the future of the creative industries.

STATE OF THE DESIGNER 2026: DESIGNERS ARE LEANING INTO THE MESSY MIDDLE

Companies are increasing their demand for designers in the AI era, with artificial intelligence actually driving renewed momentum in design hiring. A recent study examined why this trend is occurring and identified what hiring managers prioritize when recruiting designers. The research also highlights the key skills designers need to develop to advance their careers in this evolving landscape.

AIRBNB TESTS AI SEARCH TO STREAMLINE BOOKINGS

Airbnb is piloting an AI-powered conversational search that allows guests to describe their ideal stay in plain language and ask follow-up questions, aiming to improve discovery, reduce booking friction, and increase conversions as part of its broader strategy to build a more personalized, AI-native platform that supports travelers throughout the entire trip.

HOW DESIGN SYSTEMS OFFER CREATIVE SAFETY FOR PRODUCT TEAMS

Design systems offer "creative safety" by providing clear definitions that distinguish between healthy cohesion (when similar things look the same) and unhealthy sameness (when different things look the same). Clear terminology helps product teams make better decisions about when to standardize versus when to innovate, leading to more productive critiques and accountability. This clarity reduces reliance on institutional knowledge and enables design system teams to act as enablers rather than gatekeepers.

AI'S TEXT-TRAP: MOVING TOWARDS A MORE INTERACTIVE FUTURE

As AI assistants become standard in SaaS—either embedded in products or connected externally via protocols like MCP—there's a risk that text-only interfaces commoditize user experience and limit complex workflows. Integrate AI with a product's design system to enable richer visual outputs, structured inputs, and collaborative workspaces. True differentiation will come from strong design fundamentals and systems thinking rather than generic chat experiences.

HOW ADOBE LOST THE WEB AND UX DESIGN MARKET

Adobe lost its dominance in web and UX design by discontinuing popular tools like Fireworks and forcing designers to use inadequate alternatives like Photoshop. While competitors like Sketch and Figma gained market share through better-suited tools and collaboration features, Adobe's XD failed to compete and is now in maintenance mode. Adobe now faces new threats from Canva's acquisition of Serif and the free release of professional Affinity tools, challenging Adobe's expensive subscription model.

ENDLESS DESIGNS FOR YOUR IDEAS, JUST SCROLL

Most tools make you explain what you want. Variant just shows you.

FREE AI IMAGE AND VIDEO GENERATOR

Fizzly AI is a free platform that offers professional AI tools for creators to generate stunning images and videos using advanced models like Flux, SDXL, and GPT-Image 1.5.

THE PRD ISN'T DEAD, IT'S EVOLVING FOR THE AI ERA

IOS 26 RECENTLY ADDED FIVE NEW WAYS TO CUSTOMIZE YOUR IPHONE

iOS 26 and its updates introduce new customization options, including a toggle to adjust the Liquid Glass design's clarity, the ability to disable the Lock Screen camera swipe, an option to replace slide gestures with single-tap actions for alarms and timers, a new screen-flash notification alert, and an expanded opacity slider for the Lock Screen's Glass-style clock.

CHIARA XIE CREATES ILLUSTRATIONS THAT GO WITH THE FLOW

Montreal-based illustrator Chiara Xie has developed a distinctive, motion-driven style shaped by her multicultural background, artistic influences, and interest in vitality, and now plans to expand into botanical themes and larger commercial projects.

STUNNING DAILY RENDERS BLENDING CINEMA 4D AND AI INTO OTHERWORLDLY SCENES BY WILL TOULAN

Will Toulan is a Massachusetts-based digital artist who creates daily renders combining Cinema 4D and AI to produce surreal, cinematic landscapes and structures.

DESIGN SHOWS FOR THE SEASON

Five international design exhibitions opening this season that explore how creativity, science, innovation, and technology intersect to shape contemporary culture.

TRY WISPR FLOW FREE (MAC / WINDOWS / IOS) →

COMPOUND ENGINEERING: MAKE EVERY UNIT OF WORK COMPOUND INTO THE NEXT

Compound Engineering is an AI-native philosophy designed to make each unit of engineering work easier for subsequent tasks. Aiming to fundamentally change how software is built, its core process involves a four-step loop: Plan, Work, Review, and Compound. The "Compound" stage helps with continuous improvement and knowledge codification within the system. This method prevents complexity accumulation by teaching the system new capabilities and capturing insights, making codebases progressively easier to understand, modify, and trust.

TWO DIFFERENT TRICKS FOR FAST LLM INFERENCE

Anthropic and OpenAI have both introduced "fast mode" for their coding LLMs, using distinct strategies to accelerate inference. Anthropic's method offers moderate speed increases (up to 2.5x) by optimizing low-batch-size inference, serving its full Opus 4.6 model at a higher cost. In contrast, OpenAI has faster speeds (15x) by using specialized Cerebras chips and deploying a new, smaller, and less capable model called GPT-5.3-Codex-Spark.

SCALING LLM POST-TRAINING AT NETFLIX

Netflix has developed an internal Post-Training Framework to effectively scale the adaptation of LLMs for improving member experiences in areas like recommendations and search. This framework addresses critical engineering challenges encountered when fine-tuning production-grade LLMs, including complex data pipelines, distributed GPU coordination, and intricate multi-stage training workflows. It offers modular components for managing data, models, compute, and diverse workflows, supporting methods from supervised fine-tuning (SFT) to advanced reinforcement learning (RL).

THE AI VAMPIRE

The AI Vampire phenomenon is where AI tools like Claude Code increase productivity but also drain employees, leading to widespread fatigue and burnout. This arises because companies capture most of the AI-generated value, while the addictive nature of agentic software and unrealistic productivity standards set by early adopters intensify the pressure to overwork. Individuals face a dilemma: overwork to keep pace and suffer exhaustion, or risk being left behind by an accelerating, extraction-focused work culture driven by both enterprises and frenzied AI-native startups.

TASTE IS NOT A MOAT

AI is progressing fast, quickly absorbing tasks once thought to require human taste and judgment, turning taste into a decaying "alpha" rather than a durable "moat." The baseline of AI capabilities is rising so fast that human taste is valuable only relative to what AI can already produce, an advantage that constantly diminishes. Consequently, the human role is shifting towards "taste extraction," where individuals articulate and encode their judgment into AI systems to steer outputs and maintain leverage.

DAYS OF MIRACLE AND WONDER

Many people, particularly critics, have lost their capacity for genuine wonder at recent technological advancements, instead focusing on skepticism and preventing acknowledgment of progress. It's better to maintain both skepticism and a sense of awe for what has been accomplished when it comes to AI.

OAT

Oat UI is an ultra-lightweight, zero-dependency UI library built with vanilla HTML, CSS, and minimal JavaScript. It uses semantic HTML and accessibility by styling native elements directly without classes, enforcing best practices and reducing markup pollution.

ROWBOAT

Rowboat is an open-source AI coworker that builds a long-lived knowledge graph from your email and meeting notes. It uses this accumulating context to help you with tasks like drafting documents, preparing for meetings, and generating reports. A key differentiator is its local-first design, storing all data as editable Markdown notes directly on your machine.

OPENCLAW, OPENAI, AND THE FUTURE

Peter Steinberger announced he is joining OpenAI to work on making AI agents accessible to everyone. As a "builder at heart," he believes teaming with OpenAI is the fastest way to achieve his vision for widespread, safe AI adoption, rather than focusing on building another large company. His personal project, OpenClaw, which has gotten significant attention, will transition into a foundation.

WHY STRIPE PAID $1B FOR METRONOME INSTEAD OF FIXING BILLING

Stripe acquired Metronome for $1 billion because its existing billing system, designed for traditional SaaS subscriptions, couldn't handle the real-time, high-volume usage metering and complex pricing models required by modern AI and other event-driven businesses. Stripe Billing's architecture fundamentally relies on pre-aggregated data pushed via HTTP, making it unsuitable for the event streaming and progressive billing necessary for these new models. Rebuilding their core system to support an event-first architecture would have involved a multi-year, breaking change with significant migration risks, making acquisition a more economically rational choice.