Search the archive

OPENAI RELEASES O3-PRO

o3-pro is an incremental improvement over o3, which OpenAI slashed its price by 80%, across science, coding, and business tasks. It's available to Pro and Team users today, replacing o1-pro.

MISTRAL LAUNCHES FIRST AI REASONING MODEL

Adding to a string of releases over the last 2 weeks, Mistral has launched an open-source reasoning model, Magistral. It trails proprietary models on major benchmarks, but claims to be 10x faster output and stronger multilingual capabilities.

META PLANS $15B INVESTMENT IN SCALE AI TO BUILD 'SUPERINTELLIGENCE' LAB

The deal would give Meta a 49% stake in the data-labeling startup and bring over co-founder Alexandr Wang to lead a new "superintelligence" lab aimed at outperforming OpenAI, Anthropic, and Google. The massive investment follows Llama 4's underwhelming launch, but it's unclear if the investment also includes greater access to Scale's training data created for other AI labs in addition to highly sought-after AI research talent.

REAL-WORLD ENGINEERING AT CURSOR: BUILDING FOR 100X GROWTH

Cursor cofounder Sualeh Asif reveals how the two-year-old startup handles 1M+ queries per second without storing any code on its servers, using Merkle trees for secure indexing. The team survived 100x growth by switching databases during outages (Yugabyte → PostgreSQL → Turbopuffer in hours) and built Anyrun, their Rust-based orchestrator, to manage thousands of GPUs.

SPECULATIVE DECODING IN LLMS

Perplexity applies speculative decoding to speed up its Sonar models, using lightweight draft models to propose multiple tokens verified by larger LLMs.

TOWARDS ADAPTIVE CLINICAL AI VIA THE CONSENSUS OF EXPERT MODEL ENSEMBLE

Despite the growing clinical adoption of LLMs, current approaches heavily rely on single model architectures. Consensus Mechanism is a novel framework to overcome risks of obsolescence and rigid dependence on single model systems. Mimicking clinical triage and multidisciplinary clinical decision-making, the Consensus Mechanism implements an ensemble of specialized medical expert agents enabling improved clinical decision making while maintaining robust adaptability.

memory. See specs + sign up 🔔

MIXED-CHIP CLUSTERS ENABLE EFFICIENT LARGE-SCALE AI TRAINING

Shanghai-based researchers introduced DiTorch and DiComm, which unify programming across diverse chip architectures like NVIDIA and AMD variants, making it possible to train massive models on whatever hardware is available. Their framework achieved 116% efficiency training a 100B model on 1,024 chips with vastly different specs by intelligently assigning memory-hungry pipeline stages to larger-memory hardware. This allows labs without access to thousands of identical cutting-edge GPUs to still pursue frontier AI training by combining older, cheaper, or export-controlled chips into effective "hyper-heterogeneous" clusters.

REINFORCEMENT PRE-TRAINING

Reinforcement Pre-Training (RPT) is a new scaling paradigm for large language models (LLMs) and reinforcement learning (RL). It offers a scalable method for leveraging vast amounts of text data for general-purpose RL. RPT significantly improves the large model accuracy of predicting the next tokens. It also provides a strong pre-trained foundation for further reinforcement fine-tuning.

JAVELINGUARD: LOW-COST TRANSFORMER ARCHITECTURES FOR LLM SECURITY

JavelinGuard is a suite of low-cost, high-performance model architectures designed for detecting malicious intent in large language model (LLM) interactions. Each architecture presents unique trade-offs in speed, interpretability, and resource requirements. The architectures are optimized specifically for production deployment. This paper explores the architectures, benchmarking them across nine diverse adversarial datasets, and compares them against leading open-source guardrail models and large decoder-only LLMs.

EFFICIENT MULTIMODAL REASONING WITH FEWER TOKENS

LLaVA-STF compresses vision token sequences by merging adjacent tokens and adds a multi-block token fusion module, enabling 75% token reduction.

SAM ALTMAN OUTLINES PATH TO SUPERINTELLIGENCE

In a rare blog post, Sam Altman declares we've passed the "event horizon" with systems like GPT-4 and o3 that already surpass humans in many ways, predicting agents doing real cognitive work in 2025, novel scientific insights in 2026, and useful robots by 2027. He frames the coming decade as one where scientific breakthroughs compound exponentially through AI-accelerated research.

WHAT "WORKING" MEANS IN THE ERA OF AI APPS

AI startups are growing rapidly, with the average enterprise achieving over $2 million ARR in the first year. Consumer startups are also gaining traction, outpacing B2B by reaching $4.2 million ARR. The disparity between average and top performers is widening, emphasizing the need for speed and innovation.

REIMAGINING TTS WITH LLM-POWERED AUDIO GENERATION

Bland AI has reimagined text-to-speech (TTS) technology by using large language models to predict audio directly from text, enhancing expressiveness and contextual understanding. This new system leverages two-channel conversational datasets and specialized audio tokenizers for accurate and nuanced speech generation. It supports advanced capabilities like style transfer, sound effect integration, and multilingual adaptation, setting a new standard for expressive synthetic speech.

OPENAI TAPS GOOGLE CLOUD IN UNPRECEDENTED DEAL DESPITE AI RIVALRY

OpenAI's compute demands have grown so massive it's turning to its biggest search competitor for additional capacity, marking its first major cloud partner outside of Microsoft.

OPENAI ANNOUNCES 80% PRICE DROP FOR O3, ITS MOST POWERFUL REASONING MODEL

o3 is now a much more accessible option for developers seeking advanced reasoning capabilities.

MONTHLY ALTERNATIVE DATA REPORT: OPENAI, GOOGLE, META, NVIDIA, AMAZON, MICROSOFT ANTHROPIC

This article summarizes some of the most valuable insights from various alternative data providers and research reports, covering AI, semiconductors, ad tech, and the cloud industry.

OPENAI'S OPEN MODEL IS DELAYED

OpenAI's open model will be released sometime after June.

AI-2027 RESPONSE: INTER-AI TENSIONS, VALUE DISTILLATION, US MULTIPOLARITY, & MORE

AI-2027 is a heavily researched and influential attempt at providing a concrete forecast on AI capability development and its potential consequences.

EVALS NOW SUPPORTS TOOL USE

OpenAI users can now use tools and Structured Outputs when completing eval runs and evaluate tool calls based on the arguments passed and responses returned.

SOCIETE GÉNERALE BECOMES FIRST MAJOR BANK TO LAUNCH STABLECOIN

Societe Générale is the first major bank to launch a dollar-backed stablecoin on Ethereum and Solana, with public trading set to begin in July. These tokens are classified as e-money under the EU's MiCA framework, offering regulated dollar liquidity to corporates, financial institutions, and crypto exchanges. Reserves will be custodied by BNY Mellon, initially held in cash and later invested in yield-bearing assets, enabling use cases from trading and cross-border payments to collateral management. The coin will be called “USD CoinVertible.”

HOUSE AGRICULTURE COMMITTEE VOTES FOR CRYPTO MARKET STRUCTURE BILL

The House Agriculture Committee has advanced the Digital Asset Market Clarity Act, setting the stage for a full House vote if also passed by the House Financial Services Committee. The bill outlines regulatory roles for the SEC and CFTC, mandates disclosures and fund segregation by crypto firms, and exempts non-custodial platforms from money transmitter laws.

PAY-TO-LEARN: MULTI-AGENT FINE TUNING WITH MICROTRANSACTIONS

This onchain framework allows “student” language models to automatically pay specialist evaluator agents in sub-cent USDC micropayments on Base for structured feedback, eliminating the $2–5 per-annotator bottleneck of RLHF. By choosing evaluators from openly published agent cards, models create a live marketplace that rewards high-quality, domain-specific judgment and maintains feedback relevance as capabilities improve, addressing reward-model drift and echo-chamber issues noted in prior multi-judge research. Early results show orders-of-magnitude cost savings and iterative performance gains, with economic selection preserving evaluator diversity and driving continual improvement. The system is already integrated into Nous Research's Atropos RL framework.

AAVE GOES LIVE ON SONEIUM

Aave V3 is now live on Soneium, a consumer-focused Ethereum L2 built with the OP Stack and with over $150M in TVL. The deployment was led by BGD Labs with risk support from Chaos Labs and LlamaRisk. Chainlink will provide price oracle infrastructure to ensure secure market operations.

BLUEPRINT TO BUILDING WEALTH IN WEB3

To go from $0 to $1 million in crypto, one should begin with zero-cost opportunities like social-fi content creation and testnets, then reinvest small earnings into on-chain activity and airdrops, and gradually scale into DeFi, NFTs, stablecoin farming, and more advanced strategies as capital grows. At each stage, have a mindset of consistency, adaptability, action, and profit-taking, while avoiding common pitfalls like leverage, memecoins, and security exploits. At 100k, consolidate gains through diversified stablecoin and DeFi yields, Bitcoin/Ethereum DCA, and traditional assets, before finally preserving wealth via broad diversification and real-world investments.

STATE OF DEFAI Q2/2025

After an initial Q1 surge around AI abstraction layers that suffered decision paralysis and valuations sell-offs, DeFAI projects in Q2 have pivoted to specialized discovery layers for trading and yield-farming alpha. AI is now being embedded directly into existing DeFi workflows via HUD overlays, while fully autonomous, noncustodial agents like Giza's zkML-driven vault optimizers manage over $70 million in volume. Simultaneously, major DeFi protocols (BasisOS, Zerion, Vectis, and Mode) are rolling out AI roadmaps, from ML autopilots to price-predictive vaults, to boost yield and UX. Looking ahead, the convergence of traditional analytics with AI interfaces and robust, verifiable DeFAI foundations is poised to drive the next wave of value accrual as fundamentals regain favor.

ETHEREUM IS HITTING A CRITICAL INFLECTION POINT

Ethereum is moving beyond speculative hype toward real financial innovation, driven by rising institutional adoption of stablecoins, tokenization, and blockchain-based capital markets. With ETH ETFs seeing $815M in inflows over 20 days and fintechs like Visa, Stripe, and Robinhood building onchain, Ethereum is becoming the operating layer for a new financial system. If stablecoin activity is valuable, then so is Ethereum, marking a critical shift in how investors perceive network assets like ETH.

THE GREAT COMEBACK OF DAOS.FUN

DAOs.fun's journey began as Tailwind Labs, a Cosmos-based wallet that pivoted through a Sei wallet and a memecoin launchpad before finding product-market fit in investment DAOs. In its final turnaround, the team launched ai16z on Solana using the ElizaOS framework for autonomous AI agents, earning 7,000 GitHub stars in one month and reaching an $800 M market cap. The success spurred a wave of new DAO launches, democratizing on-chain fund management.

SOLANA ETFS COULD SEE APPROVAL IN FOUR MONTHS

The SEC has asked issuers of proposed spot Solana ETFs to update their S-1 filings, signaling potential progress toward approval.

AMERICAN BITCOIN ACQUIRES FIRST $23 MILLION WORTH OF BTC

American Bitcoin, the mining firm backed by Eric Trump and Donald Trump Jr., acquired 215 BTC (~$24M) on May 31, its first major buy under a new treasury strategy modeled after MicroStrategy.

RAIN SETTLES DIRECTLY WITH VISA IN USDC, REINVENTING CARD INFRASTRUCTURE

Unlike most card programs that rely on third-party infrastructure, Rain owns its entire stack (licensing, issuing, and settlement), allowing it to settle directly with Visa in USDC, 24/7.

CRYPTO AI PROMPTS TO GAIN AN EDGE

Use AI to analyze market data like correlations with BTC, Sharpe ratios, and Value at Risk (VaR).

ROKU IS TESTING A NEW LOOK

Roku is testing a redesigned homepage with a new Quick Access section that automatically displays users' most visited apps. The update moves the cursor directly to the main grid when pressing home while relocating options like "Live TV" and "Featured Free" from the left sidebar to make them more discoverable. New sections include "Subscriptions" for viewing all subscription content in one place and "For You." Personalized recommendations have replaced the previous "What to Watch" feature.

INTRODUCING MODIFY VIDEO

Luma's Modify Video is a new tool that allows professional creatives to change video environments, lighting, and textures while preserving the original performance and motion. The tool offers motion capture capabilities to extract movement data and apply it to new characters or objects, plus the ability to transform scenes from garages to spaceships while maintaining the original framing. Users can edit individual elements like wardrobe or props without affecting the entire shot using three structured presets ranging from close adherence to creative reimagining.

LOVE IT OR HATE IT? APPLE'S NEW ‘LIQUID GLASS' DESIGN IS GETTING MIXED REVIEWS

Early testers have criticized Apple's new “Liquid Glass” user interface for poor readability, especially in notifications and the Control Center, where background blur and font contrast are lacking, though these issues may still be addressed before public release. Despite the rough start, Liquid Glass includes promising visual details and effects, suggesting Apple is laying the groundwork for more immersive, AI-integrated interfaces in the future.

WHERE SHOULD AI SIT IN YOUR UI?

Large language models are driving a shift from simple chatbot interfaces to more deeply integrated, task-focused layouts such as inline prompts, infinite canvases, and intelligent spreadsheets. These spatial designs—whether side panels, chat widgets, or dynamic grids—not only affect usability but also shape how users understand the AI's role, capabilities, and trustworthiness within the experience.

DO GRAPHIC DESIGNERS NEED NEW SKILLS TO BEAT AI? PROFESSIONALS WEIGH IN

Graphic designers are facing growing uncertainty about AI's impact on their careers, but many industry voices remain optimistic, emphasizing that creativity, vision, and adaptability still set human designers apart. While AI is becoming a valuable tool, success in the field will increasingly depend on evolving skillsets, strategic thinking, and the ability to integrate technology without losing the human touch.

COLLABORATION: THE MOST UNDERRATED UX SKILL NO ONE TALKS ABOUT

Collaboration is the most underrated skill in UX. It's different from mere cooperation, which emphasizes co-ownership of outcomes rather than just individual task completion. While AI tools can accelerate ideation and prototyping, they cannot replace human judgment and the need for validation from teammates who understand users, technical constraints, and business goals. Strengthening collaboration skills involves cultivating curiosity about teammates' challenges, facilitating structured conversations, sharing work early, practicing active listening, and co-owning outcomes rather than defending individual designs.

CREATE WINNING ADS WITH AI ACTORS

Arcads transforms text into videos using AI actors, saving time and cutting costs.

OKLCH COLOR PICKER AND CONVERTER

The OKLCH color picker and converter is a tool that helps developers and designers work with colors. It allows users to convert between various color spaces, including OKLCH, RGB, HSL, and Hex.

FIGMA IMAGE EDITOR (FIGMA PLUGIN)

Manipulate an image directly in Figma with many different color adjustments and filters. Save and load your own presets and keep track of changes.

AFTER 25 LOYAL YEARS, ADOBE'S PRICE HIKE HAS FORCED ME TO SAY GOODBYE

A 25-year Adobe user announces their departure from the software giant after the company eliminated the affordable Photography Plan, forcing a price jump from £82 to £249 for the same Photoshop and Lightroom access. They are transitioning to alternatives, including the Affinity creative suite and ON1 Photo Raw for photo organization and processing, although the switch will be challenging after decades of muscle memory with Adobe shortcuts.

FRONTIFY FUTURE'S BESPOKE TYPEFACE CHAMPIONS THE KINETIC FUTURE OF BRANDING

Frontify Futures has a new, fluid, experimental identity that embodies the platform's mission of exploring brand evolution through a living typographic system. Rather than fixed letterforms, the typefaces exist only momentarily as animated parameters generated on the fly, allowing the brand to evolve alongside the creative landscape continuously it studies.

APPLE'S BEST DESIGN UPDATE MIGHT BE ITS SIMPLIFIED NEW NAMING CONVENTION

Apple is shifting to a year-based naming system for its operating systems—like iOS 26 instead of iOS 19—to simplify version tracking and align with naming trends in other industries.

AS ROMA DITCHES LOGO FOR "PROGRESSIVE RETURN" OF HISTORIC CREST

AS Roma has unveiled a new logo that revives its historic ASR monogram and detailed Lupa Capitolina illustration, signaling a return to tradition ahead of the club's centenary.

CHRIS CHAPMAN'S IDENTITY FOR AGUA DE MADRE FIZZES WITH JOY, MEMORY, AND TYPOGRAPHIC EXCESS

Chris Chapman reimagined the Agua de Madre brand by blending nostalgic typefaces, bold colors, and sticker-like graphics to create a joyful, chaotic design that defies gut-health drink norms while retaining emotional elements from its original identity.