Search the archive

GOOGLE LEAKS ITS FULL PIXEL 10 LINEUP

The design and colors of the devices line up with the flood of leaks seen in recent weeks.

DOWNLOADING ARBITRARY APPLE PODCAST EPISODE TRANSCRIPTS

Apple Podcasts on macOS only allows users to copy 200 words of transcript at a time - this post details how to view and copy the whole thing.

FACTORY unifies all of your context (GitHub, Linear, Slack, DataDog, Sentry, etc.) in one place. Cutting-edge teams use Factory to accelerate: (1) Feature development (2) Codebase onboarding (3) Migrations in high-stakes production settings.

Start Building with Factory [14]: 14-day trial with 10M standard tokens

Implement MCP Auth with WorkOS →

* Packages all 5 OAuth specs into one simple API.

* Pre-built infrastructure that saves engineering time. * Enterprise ready security out of the box.

TRUMP ADMINISTRATION PLEDGES TO STIMULATE AI USE AND EXPORTS

The Trump administration has promised to slash red tape and take steps to boost exports for US tech companies in a bid to accelerate AI use in the US. Its action plan lays out moves to make it easier and faster for tech companies to build data centers and get the power they need for those centers. The plan includes directives to several government agencies to identify and gut any regulations that block the development and use of AI. The administration is also working on addressing uncertainty about whether it is legal for AI models to train on copyrighted material.

GITHUB LAUNCHES SPARK FOR NO-CODE AI APP CREATION

ANTHROPIC BACKS US AI PLAN, URGES TRANSPARENCY AND SECURITY

The White House's "Winning the Race: America's AI Action Plan" prioritizes accelerating AI infrastructure, federal adoption, and security coordination. Anthropic supports this strategy and emphasizes the importance of export controls and transparency standards in AI development. The plan aligns with Anthropic's earlier recommendations and highlights the need to maintain America's AI leadership through robust infrastructure, security, and responsible policy measures.

OPENAI'S NEW ECONOMIC ANALYSIS

OpenAI's new economic research reveals that 28% of employed US adults now use ChatGPT at work — up from just 8% in 2023. The usage data shows learning and upskilling dominating at 20% of US messages, followed by writing and communication at 18%, with programming and data science capturing 7%.

INVERSE SCALING APPEARS IN EXTENDED REASONING

Anthropic found that more test-time compute doesn't always help: longer reasoning chains in large models sometimes lead to worse performance.

“BEHAVIORIST” RL REWARD FUNCTIONS LEAD TO SCHEMING

A large class of reward functions, which includes almost every reward function in the RL and LLM literature, are all doomed to eventually lead to AI that will pretend to be docile and cooperative while secretly looking for opportunities to behave in egregiously bad ways. This is because negative reward for lying and stealing looks the same as negative reward for getting caught lying and stealing. The reward function will miss sufficiently sneaky misaligned behavior, so the AI will learn that this kind of behavior is good.

design systems. Try it — free

TIMESCOPE: HOW LONG CAN YOUR VIDEO LARGE MULTIMODAL MODEL GO?

TimeScope is a new open source benchmark for evaluating vision models on how well they handle long videos. It evaluates not just retrieval but synthesis, localization, and fine-grained motion analysis to provide a more holistic view of temporal comprehension. Using the benchmark reveals how model size isn't everything - simply scaling parameters doesn't automatically grant a longer temporal horizon - and Gemini 2.5-Pro is in a league of its own, being the only model able to maintain strong accuracy on videos longer than an hour.

VOXTRAL

Voxtral Mini and Small are new multimodal audio chat models that excel in both spoken audio and text comprehension. Voxtral Small outperforms many closed-source models, can run locally, and supports audio files up to 40 minutes.

SUBLIMINAL LEARNING IN LANGUAGE MODELS

Anthropic has explored how language models can inherit behavioral traits such as preferences or goals from other models through training data that appears unrelated. This effect only occurs when the teacher and student models share the same base architecture, suggesting hidden signals in the data can transmit unintended behavior.

HIGGS AUDIO V2

Higgs Audio v2 is an audio foundation model trained on 10 million hours of data that achieves 75.7% win rates over GPT-4o-mini-tts. The open-source model demonstrates emergent capabilities like multi-speaker dialogues and voice cloning without requiring fine-tuning.

GOOGLE DEEPMIND LAUNCHES AENEAS AI MODEL FOR ANCIENT ROMAN INSCRIPTION ANALYSIS

Aeneas is the first AI model designed to help historians contextualize ancient Roman inscriptions by identifying textual parallels and restoring damaged texts with 73% accuracy for gaps up to ten characters. DeepMind is making the tool freely available at predictingthepast.com and plans to expand to other ancient languages.

ELON MUSK SAYS XAI IS TARGETING 50 MILLION 'H100 EQUIVALENT' AI GPUS IN FIVE YEARS

xAI plans to deploy 50 million H100-equivalent GPUs within five years. The power these processors will consume is massive, estimated to be equal to the typical power generated by 35 nuclear power plants, making it unrealistic to power today. Whether xAI will be able to get power to feed such a data center by 2030 remains to be seen.

fall asleep faster and wake up feeling amazing. Get 20% off

YOUTUBE SHORTS IS ADDING AN IMAGE-TO-VIDEO AI TOOL, NEW AI EFFECTS

YouTube is following Gemini and Meta's Edits app with its own image-to-video AI feature for Shorts creators, transforming static photos into six-second videos using Google's Veo 2 model.

NEW OPEN MODELS ARRIVE IN THE VERTEX AI MODEL GARDEN

Google Cloud has added the DeepSeek R1 model to its Vertex AI Model Garden, offering more flexibility for AI developers.

TEN AI SAFETY PROJECTS I'D LIKE PEOPLE TO WORK ON

There is a real chance that AI systems capable of causing a catastrophe will be developed in the next decade - this post lists promising projects aimed at reducing catastrophic risks from transformative AI.

BUSINESS INSIDER OBTAINED AN INTERNAL LIST OF WEBSITES THAT COULD AND COULDN'T BE USED FOR TRAINING ANTHROPIC'S LATEST AI MODELS

Many of the whitelisted sources copyright or otherwise restrict their content, and the blacklist includes companies like the NYT and Reddit, which have sued AI startups for scraping without permission.

SOLANA & JITO EXECS NAMED IN EXPANDED RICO SUIT OVER PUMP.FUN

Burwick Law and Wolf Popper have amended their RICO lawsuit against Pump.fun to include the Solana Foundation, Solana Labs, Jito Labs, and multiple executives, including Anatoly Yakovenko and Raj Gokal, accusing them of facilitating an illegal digital gambling and money transmission enterprise. The complaint alleges violations of AML laws, the Bank Secrecy Act, and claims Pump enabled the laundering of stolen Bybit funds by the Lazarus Group. Plaintiffs claim Pump has generated $723 million from a system they liken to a “digital slot machine,” while Jito is accused of frontrunning winning trades via MEV.

NINTH CIRCUIT RULES NFTS CAN BE TRADEMARKED

The U.S. Court of Appeals ruled that NFTs like Yuga Labs' Bored Ape Yacht Club qualify as “goods” eligible for trademark protection under the Lanham Act, a landmark win for NFT IP rights. However, it vacated the $9M judgment against artist Ryder Ripps and ordered a trial to determine whether consumers were actually confused by his RR/BAYC project, leaving the case far from over.

TELEGRAM CRYPTO WALLET GOES LIVE FOR 87M USERS

Telegram has officially launched its built-in, self-custodial TON Wallet to U.S. users, enabling seamless crypto transfers, swaps, and purchases directly within the app. Already adopted by 100M+ users globally, the wallet offers a user-friendly experience with features like stablecoin support, MoonPay-powered fiat ramps, and access to Telegram's Mini App ecosystem. This positions it as a serious contender to platforms like Cash App and Coinbase.

TETHER PREPS U.S. INSTITUTIONAL PUSH POST-GENIUS ACT

Tether CEO Paolo Ardoino says the company's U.S. expansion is “well underway” as it readies a new stablecoin tailored for institutional clients seeking faster settlements. The move follows the signing of the GENIUS Act, which mandates full USD backing and annual audits for large stablecoin issuers. While acknowledging that U.S. banks may initially have an edge, Ardoino said Tether's tech stack and market experience remain superior and ruled out any plans for a public listing like Circle.

gluten, no sugar. Get 20% off

BEST STABLECOIN YIELDS ISSUE #13

This week's top stablecoin vaults include HyperBeat's USDT strategy on Hyperliquid at 15-35% APR and Euler's FalconStable at 35-60% APR, while ResolvLabs' RLP vault returns 27% APR (≈97% when leveraged). Avalanche stablecoins on Euler yield 47-96% APR, and BlackholeDex's ve(3,3) pools pay 30-70% APR. Other notable opportunities are Merkl's 28% APR for lending USDS, Plume's 20% APR vaults, and Gauntlet USD Alpha's 12% APY on USDC.

THE ROLEX INDICATOR

The “Rolex Indicator” tracks luxury watch prices and shows that true speculative euphoria often arrives late in crypto cycles, when new wealth starts flashing status symbols in the real world. In 2021, watch prices only really spiked after Bitcoin's second ATH and NFT mania, then crashed alongside crypto. Today, Rolex premiums remain muted despite near-record Bitcoin highs, suggesting we're in the middle rather than at the top of the cycle. When we finally see watches selling at 2x retail and countless wrist‐shots on social media, that's likely the late‐cycle signal.

WHY TWITTER IS LYING ABOUT TOKEN LAUNCHES

Crypto Twitter's favorite playbook – buying engagement, launching tiny floats, touting massive VC raises, and front-loading hype – simply doesn't move prices. Analysis of 50,000 data points from 40 major 2025 token launches shows near-zero correlation between Twitter metrics, initial circulating-supply percentages, or fundraising size and post-launch returns. The one statistically significant driver is Initial Market Cap, as larger IMCs predict weaker performance. The best predictors of success are realistic launch valuations, genuine product utility, authentic ongoing communication, and strong volume retention after the initial pump.

HOW HYPERLIQUID BREAKS WALL STREET'S GAME

Hyperliquid's HIP-3 and “Hyperps” introduce on-chain perpetual futures for pre-IPO private companies, creating a transparent, 24/7 market where anyone can trade valuations of firms like Stripe or OpenAI months before their IPO. By surfacing real-time price discovery and order flow, these permissionless markets undercut the traditional investment-bank book-building cartel that underprices shares by 20-40% and locks out retail until day one. With oracleless pricing and time-weighted funding mechanisms, Hyperps collapse the artificial private-public divide, democratize access to early growth, and strip away the “legalized front-running” banks have long enjoyed. While regulatory hurdles and initial liquidity constraints remain, Hyperliquid's approach promises to dismantle decades of IPO rent-extraction and usher in genuinely free capital markets.

YC'S CRYPTO SUCCESS

Famed startup incubator Y Combinator has only funded 72 crypto companies, but four of them – Coinbase, Axiom, OpenSea, and TRM – have become very successful.

AXIOM'S MASSIVE GROWTH

Axiom, a Telegram trading bot, has generated $179 million in revenue since launching in January.

SEC APPROVES BITWISE MULTI-ASSET CRYPTO ETF

The SEC has approved the conversion of the Bitwise 10 Crypto Index Fund into a spot ETF holding a diversified basket of digital assets.

FLUID LEADS LENDING GROWTH WITH 20.2% SURGE IN ACTIVE LOANS

Fluid tops this week's fastest-growing lending protocols with a 20.2% increase in active loans, now totaling $1.3 billion.

NETFLIX STARTS USING GENAI IN ITS SHOWS AND FILMS

POLAROID JUST DROPPED THE MOST ICONIC ANTI-AI AD OF THE YEAR

Polaroid has launched a powerful anti-AI campaign to promote its new Flip camera, emphasizing the value of tangible, real-life experiences over digital saturation. The campaign uses nostalgic Polaroid imagery paired with provocative slogans like “AI can't generate sand between your toes” to highlight the emotional depth and authenticity of analog living. Strategically placed near tech hubs like Apple and Google offices, the campaign has resonated widely online, drawing praise for its aesthetic and timely critique of tech culture.

FIGMA ANNOUNCES LAUNCH OF INITIAL PUBLIC OFFERING ROADSHOW

Figma launched its IPO roadshow, offering 12.5 million new shares, plus 24.6 million from existing stockholders, at a price range of $25 to $ 28 per share. The design platform company applied to list on the NYSE under the ticker "FIG" with significant investment banks leading the offering. Figma won't receive proceeds from existing stockholder sales, while underwriters have a 30-day overallotment option for additional shares.

AFRAID OF LOSING YOUR DESIGN JOB TO AI? THIS MCDONALD'S STORY MIGHT HELP YOU SLEEP BETTER

Fears that AI will replace creative jobs echo earlier anxieties about McDonald's touchscreen kiosks, which were expected to eliminate human roles but instead led to more jobs and higher sales by enhancing the customer experience. The key takeaway is that technology doesn't determine outcomes—organizations do. Design studios that use AI to amplify human creativity, not just cut costs, will thrive.

DESIGNING SOFTWARE FOR AI AGENTS

AI agents increasingly handle computing tasks but struggle with software designed for humans, creating opportunities to redesign systems for machine use. Web search APIs limit results and provide condensed snippets suited for human scanning, while agents could benefit from more comprehensive results and larger document summaries. AgentDB exemplifies agent-first design by utilizing UUIDs for instant database creation, storing databases as files rather than hosting them as services, and integrating with MCP servers to make data easily accessible as AI model context.

AI TOOLS + AI FLUENCY + HUMAN ADVANTAGE = AI-NATIVE DESIGNER

Product designers can thrive in the AI era by mastering three key areas: leveraging AI tools for faster shipping, developing AI fluency for designing probabilistic systems, and harnessing human advantages such as craft and creative direction. While AI handles routine tasks and generation, designers must evolve into orchestrators who provide vision, set constraints, and make high-level decisions that guide AI outputs. The future belongs to AI-native designers who combine technical literacy with uniquely human skills like strategic thinking, storytelling, and the ability to create memorable experiences that transcend generic AI outputs.

UX BITES

Find the best UX ideas, trends, and innovations at warp speed.

COPY CSS FROM ANY WEBSITE AND ITERATE WITH AI

Design UI with AI, upload a screenshot or describe the idea or UI you want, transform the design to code, and let AI generate the frontend code for you.

AI PRESENTATIONS

Workppt can create a compelling starting point for presentations in just minutes. It has everything you need to create and refine content with advanced AI quickly and efficiently.

IMPACT STORIES: HOW BLOOM REPOSITIONED GINSTERS (2023)

By 2022, Ginsters faced a perception problem: widely recognized but associated with low quality and outdated imagery, it struggled to attract new customers despite a strong sales history. A strategic rebrand by Bloom in 2023 revitalized the brand through updated design, storytelling, and a more confident tone, resulting in major sales growth, improved brand perception, and successful product innovation aimed at younger audiences.