Search the archive

with AI-powered development. Download free from DX

JONY IVE TO LEAD OPENAI'S DESIGN WORK FOLLOWING $6.5B ACQUISITION OF HIS COMPANY

OpenAI is acquiring Jony Ive's startup, io, in an all-equity deal that values the startup at $6.5 billion. Ive and his design firm, LoveFrom, which will continue to operate independently, will now lead creative and design work at OpenAI. He could help OpenAI directly compete with Apple in the consumer hardware space - Apple has struggled to develop AI features that keep up with the latest technology. io's staff of around 55 engineers, scientists, researchers, physicists, and product development specialists will join OpenAI. They will be tasked with developing AI-powered consumer devices and other projects.

TESLA POSTS OPTIMUS' MOST IMPRESSIVE VIDEO DEMONSTRATION YET

The Tesla Optimus team's latest video demonstration shows the humanoid robot performing a variety of tasks, including household using a broom and vacuum cleaner, tearing a paper towel, stirring a pot of food, opening a cabinet, closing a curtain, and picking up a Model X fore link and placing it on a dolly. The robot completed all these tasks through a single neural network. It learned its actions using data from first-person videos of humans performing similar tasks. The system could pave the way for Optimus to learn and refine new skills quickly and reliably. The video is available in the article.

FOR ALGORITHMS, A LITTLE MEMORY OUTWEIGHS A LOT OF TIME

Ryan Williams, a theoretical computer scientist at the Massachusetts Institute of Technology, recently posted a proof online on the relationship between time and memory in computing. The proof, which showed that memory was more powerful than computer scientists had previously believed, established a mathematical procedure for transforming any algorithm, no matter what it does, into a form that uses much less space. It also implies there are calculations that can't be computed in certain amounts of time, offering a new way to attack one of the oldest open problems in computer science.

THE SMALL ROBOT COMPANY WITH BIG GLOBAL AMBITIONS

TALES FROM MAINFRAME MODERNIZATION

Legacy code is full of surprises. Mainframes are chock-full of history, containing wonderful hacks that get around the limitations of systems. This post contains several examples of these hacks within a COBOL system.

PYTHON TOOLING AT SCALE: LLAMAINDEX'S MONOREPO OVERHAUL

LlamaIndex is an ecosystem consisting of more than 650 Python packages, mostly Integrations and Packs. All of these packages share a single GitHub repository, or monorepo. This article discusses LlamaDev, a tool for managing monorepos at scale, and explains the challenges with existing tooling. LlamaDev makes running tests faster and easier, creates clearer logs so contributors know exactly what happened if a check failed, and can be run locally for debugging.

TESLA'S HEAD OF SELF-DRIVING ADMITS ‘LAGGING A COUPLE YEARS' BEHIND WAYMO

Tesla's head of AI and self-driving, Ashok Elluswamy, recently admitted in an interview that the automaker's autonomous program is lagging a couple of years behind Waymo's. However, Tesla's approach is much cheaper, and its cost advantage will allow the company to scale faster. Tesla produces over a million cars a year, versus Waymo's few hundred units, and it plans to use the same vehicles for its autonomous riding service. The company is still limited to a level 2 advanced driver system, which requires constant supervision from the driver, while Waymo has been providing customers with level 4 autonomous driving rides for years. It plans to start offering level 4 autonomous rides to customers in Austin next month.

SERGEY BRIN POINTS TO WHERE GOOGLE GLASS FAILED — AND WHAT ANDROID XR GETS RIGHT

Google is getting back into the smart glasses game. The company has announced a partnership with Warby Parker - they plan to launch a series of smart glasses built on top of Google's Android XR as soon as next year. The rise of generative AI has allowed Google to revisit the idea of Google Glass, a wearable it launched in 2013 that failed partly due to a technology gap. The company has learned many lessons since the failed launch. Other companies, like Meta and Apple, are also reportedly working on smart glasses.

GOOGLE IS STUFFING EVEN MORE ADS INTO ITS AI RESULTS

Google plans to start testing ads in AI mode, a new AI-powered search feature the company just rolled out to the US.

STOREFRONT WEB COMPONENTS

Shopify's Storefront Web Components lets developers bring Shopify-powered commerce capabilities to any website.

MICROSOFT BLAMES APPLE FOR ITS DELAYED XBOX MOBILE STORE

Microsoft says it has been unable to implement linked-out payments because of Apple's anti-steering policies, which restrict Microsoft's communications to users and impose a high economic cost.

MANY IPHONES STOLEN IN THE US AND EUROPE END UP IN ONE BUILDING IN CHINA

The Feiyang Times building in Shenzhen is known as a place for phone sales and repairs - and also a marketplace for stolen devices.

SPOTIFY SAYS SUPPORT FOR EXTERNAL PAYMENTS ON IOS HAS ALREADY BOOSTED SUBSCRIPTIONS

Spotify's iOS subscriptions have significantly increased in the past two weeks, while sales have remained relatively constant on Android.

WHY EUROPE'S TECH SCENE IS SO MUCH SMALLER THAN THE US'S AND CHINA'S

China welcomes foreign companies and competition but creates conditions where local alternatives can thrive alongside them, and its provinces, despite competing against each other like countries in the EU, all have the ultimate goal of developing a Chinese champion that can successfully compete on the global stage.

Descope. Make your app AI-ready today

FORMER APPLE DESIGN GURU JONY IVE TO TAKE EXPANSIVE ROLE AT OPENAI

OpenAI has fully acquired io, its joint venture with famed Apple designer Jony Ive, for $6.5 billion in equity as it moves into hardware development.

MISTRAL'S AGENTIC LLM FOR SOFTWARE ENGINEERING

Mistral AI and All Hands AI have introduced Devstral, a new open-source LLM optimized for software engineering.

SIGNALFIRE REPORT: NEW GRAD TECH HIRING PLUMMETS AND ANTHROPIC DOMINATES AI TALENT WARS

Tech hiring for new graduates has collapsed by over 50% since pre-pandemic levels while experienced engineers remain in high demand as AI tools increasingly handle entry-level tasks. Anthropic has emerged as the clear leader in the AI talent wars, maintaining an impressive 80% employee retention rate while aggressively poaching engineers from competitors. Talent is now 8 times more likely to leave OpenAI or DeepMind for Anthropic than move in the reverse direction.

MINIMAL MCP + A2A EXAMPLE (30 MINUTE VIDEO)

A toy example demonstrating the basics of Minimum Cost Path (MCP) and Agent-to-Agent (A2A) ping checks.

BUILDING AN AGENTIC IMAGE GENERATOR THAT IMPROVES ITSELF

Large language models exhibit strong capabilities in natural language reasoning about visual imperfections, but they struggle to map these high-level insights into precise pixel-space action. The models excel when reasoning is constricted to discreet, well-scoped dimensions - performance degrades when they are asked to balance abstract aesthetic judgments with deterministic pixel-level corrections. This suggests that the models have trouble bridging symbolic understanding and spatial grounding, especially in iterative workflows that demand surgical image edits.

GOOGLE UNVEILS DEFENSE STRATEGIES AGAINST AI PROMPT INJECTION ATTACKS

Google DeepMind's research on defending Gemini against indirect prompt injection attacks reveals that more capable models aren't inherently more secure, and defenses that work in static tests often fail against adaptive attackers. Their findings show that "adversarial training" - fine-tuning models on examples of malicious prompts - significantly improved Gemini 2.5's resistance without degrading performance on normal tasks. When they combined this approach with a "Warning" defense, attack success rates plummeted from 94.6% to just 6.2%.

NEW TOOLS AND FEATURES IN THE RESPONSES API

OpenAI has added built-in tools and features to the Responses API. The API now has support for all remote Model Context Protocol (MCP) servers and tools like image generation, Code Interpreter, and improvements to file search. It has new features that improve reliability, visibility, and privacy for enterprises and developers. All of these tools and features are now available in the Responses API, which supports the GPT‑4o series, GPT‑4.1 series, and o-series reasoning models. Image generation is only supported on o3 in the reasoning model series.

LLM FUNCTION CALLS DON'T SCALE; CODE ORCHESTRATION IS SIMPLER, MORE EFFECTIVE.

Giving large language models the full output of tool calls is costly and slow. Output schemas enable developers to get structured data, which can then be processed. Using code execution as a way to process data from MCP tools opens the door to scaling the work of AI models. However, allowing an execution environment to also access MCPs, tools, and user data requires careful design as to where API keys are stored and how tools are exposed.

FUTUREHOUSE'S NEW AI SYSTEM SUCCESSFULLY IDENTIFIED A NOVEL TREATMENT FOR LEADING CAUSE OF BLINDNESS

FutureHouse combined specialized literature search agents with a data analysis agent in a continuous experimental loop to accelerate medical discovery. The system autonomously generates hypotheses from literature reviews, proposes experiments for human execution, then analyzes resulting data to inform the next research cycle. Through this process, researchers identified ripasudil, an eye drop medication that relaxes cellular tension, as a promising new treatment for the leading cause of age-related vision loss, where retinal light-sensing cells gradually deteriorate. The complete code, data, and agent interaction traces will be publicly released on May 27.

MY PROMPT, MY REALITY

AI products depend heavily on user prompts, unlike traditional software with predictable outcomes. Skilled prompting can yield varying results due to nuanced intent and context. Product teams can improve results by refining prompts and using follow-up questions to guide users.

WE'RE RELEASING V0'S AI MODEL

v0's AI model, which has specialized web-dev knowledge and an OpenAI-compatible API, is now in beta in the API, AI SDK, and AI Playground.

GEMINI DIFFUSION

Gemini Diffusion is Google's first large language model to use diffusion in place of transformers - it has the performance of Gemini 2.0 Flash-Lite at five times the speed.

LLAMA FOR STARTUPS PROGRAM

Meta has announced a new initiative to support early-stage U.S. startups using its Llama models.

LM ARENA RAISES $100M FROM AT $600M VALUATION

LM Arena, a crowdsourced AI benchmarking platform, has secured $100 million in seed funding led by Andreessen Horowitz and UC Investments, despite recent accusations from some researchers that it helps top AI labs game its influential leaderboards.

SLOW THINKING IMPROVES CONFIDENCE IN LLMS

Extended chain-of-thought reasoning helps large language models better calibrate their confidence.

SEC CHARGES UNICOIN WITH $100M `SECURITIES FRAUD`

NYC MAYOR ERIC ADAMS IS CREATING A CRYPTO ADVISORY COUNCIL

Mayor Eric Adams announced a digital assets advisory council to funnel fintech jobs and expertise into New York City, with a chairperson to be named shortly. The council will advise on blockchain and crypto initiatives, such as exploring on-chain storage of vital records, and help shape policies that serve residents. Adams framed the effort as harnessing “technology of tomorrow” and showcased participation from family offices and unicorn startups.

SP1 HYPERCUBE – REAL-TIME ETHEREUM PROVING IS HERE

SP1 Hypercube is ZK's “man on the moon moment,” as it proves over 93% of Ethereum mainnet blocks in under the current block time of 12 seconds. It achieves 5x lower latency and cost compared to SP1 turbo using a novel zkVM optimized for multilinear polynomials. A real-time prover cluster requires roughly 160 Nvidia 4090 GPUs (~$300k), with potential to drop to ~$100k on cost-efficient hardware. By delivering sub-12s proofs, SP1 Hypercube paves the way for L1 scaling, native rollups, and improved interoperability.

AGENT COMMERCE KIT

Agent Commerce Kit provides open-source protocols, including ACK-ID for verifiable agent identity and ACK-Pay for agent-native payments, that bridge AI with legacy financial systems. ACK-ID uses DIDs and Verifiable Credentials to establish cryptographic ownership and secure authentication and privacy-preserving disclosure, while ACK-Pay standardizes payment initiation, flexible settlement across traditional and blockchain rails, and cryptographic receipts with optional human oversight. It also features advanced micropayment and conditional payment features, x402 interoperability, and agent-specific reputation frameworks.

THE RISE OF ONCHAIN AI: AGENTS, APPS, AND COMMERCE

Autonomous AI agents are becoming prolific onchain participants by holding wallets, executing smart contracts, and powering AI-infused applications. Important infrastructure spans trusted execution environments, agent frameworks and toolkits, agent launchpads, multi-agent coordination protocols, Model Context Protocols for external data, and AI app stores for discovery and monetization. This “Agentic Commerce” allows agents to transact with humans and each other via crypto rails and stablecoins, exemplified by protocols like x402 enabling machine-speed, always-on economic activity. While promising rapid growth and innovation, the ecosystem must overcome challenges in agent reliability, blockchain scalability, and novel governance frameworks.

STABLECOINS ARE THE TROJAN HORSE OF CRYPTO

Stablecoin transactions hit $6.8T in Q1 2025 as platforms like Stripe launch native support and issue their own, signaling a shift toward mainstream adoption. With crypto wallets enabling direct stablecoin payments at checkout, stablecoins are becoming a default payment rail for global eCommerce. While Ethereum benefits from this growth, the stablecoin model diverges from Bitcoin's original vision of decentralized, intermediary-free payments.

INTRODUCING SIM: DUNE BEYOND ANALYTICS

Dune has launched SIM, a new language that makes onchain data analysis easier by replacing complex SQL with a more intuitive, spreadsheet-like syntax. It's designed to help users build modular, transparent data models without deep technical expertise.

COINBASE ACQUIRING CIRCLE ISN'T CRAZY

Circle trades at a $5B market cap with $155M net income and a similar 32x P/E as Coinbase, but with USDC poised to grow alongside global dollar demand and regulatory clarity via the GENIUS Act, its future earnings potential is much larger. Even with falling interest rates, a $250B–$500B USDC supply at a 3% yield could generate $7.5B revenue and $2B+ net income, matching Coinbase's current run rate and justifying a $60B market cap boost.

WORLD RAISES $135M TO SCALE ORB-VERIFIED IDENTITY NETWORK

World Assets, Ltd. has raised $135M from Andreessen Horowitz and Bain Capital Crypto through a WLD token sale to expand its Orb-verified World ID network, now used by over 12.5M people.

SOLANA PASSES 50M SOL IN LIQUID STAKING

Solana now has over 50 million SOL in liquid staking tokens (LSTs), representing 12.8% of all staked SOL (up from 6.7% a year ago).

STABLECOIN SUPPLY ON APTOS UP FROM LAST YEAR

Stablecoin supply on Aptos has jumped from ~$430M to ~$1.13B year-to-date, driven by strong growth in both USDT and USDC natively minted on the network.

SONIC LABS CONTINUES FAST GROWTH

Sonic has crossed $1M in annualized revenue, with May showing breakout momentum as the platform generates over $85K in weekly revenue for three straight weeks.

CELESTIA SCALES BY SUBSIDIZING FEES

With TVS nearing $600M, dozens of chains live, and top apps launching soon, Celestia argues it's executing a classic startup playbook: subsidize early, monetize later.

JONY IVE TO LEAD OPENAI'S DESIGN WORK FOLLOWING $6.5B ACQUISITION OF HIS COMPANY

OpenAI has acquired Jony Ive's AI hardware startup, io, in a $6.5 billion all-equity deal. The move brings ex-Apple designers to OpenAI and marks its boldest step into consumer hardware. Ive and his firm LoveFrom will lead design efforts, aiming to create AI-powered devices that move “beyond screens.” Ive will shape future ChatGPT versions and new AI products, escalating competition with Apple as generative AI hardware takes shape.

ANDROID XR IS GETTING STYLISH PARTNERS IN WARBY PARKER AND GENTLE MONSTER

Google has partnered with Warby Parker, Gentle Monster, and Samsung to develop stylish Android XR smart glasses that emphasize fashion and wearability. Integrated with Gemini AI, the glasses will offer photo capture, navigation, and live translation via built-in cameras, mics, and speakers. The move mirrors Meta's Ray-Ban strategy, highlighting how XR's mainstream success hinges on blending tech with everyday style.

VRCHAT IS LOADING INTO THE CREATOR ECONOMY WITH AN AVATAR STORE

VRChat is launching a new Avatar Marketplace to monetize its platform after 2024 financial struggles led to 30% staff layoffs. Approved creators can sell custom avatars and digital cosmetics using "VRChat Credits," keeping most of the revenue. Positioned as a meme-driven Roblox alternative, VRChat is also expanding into branded experiences, including a partnership with McDonald's Japan.