Search the archive

A SIMPLE A/B TEST ADDED $759K A YEAR

An A/B test using dollar savings and red highlights increased conversions by 11.6%.

PERPLEXITY STRIKES MULTI-YEAR LICENSING DEAL WITH GETTY IMAGES

Perplexity partnered with Getty Images to use its photos in AI search tools.

IS YOUR TEAM SHIFTING YOUR CONTENT DOWN THE FUNNEL?

Overweighting on TOFU is a mistake because AI chat and your competitors have it covered.

GET STARTED FOR FREE.

ANDROID 17 WILL PUT FULL-SCREEN APPS ON YOUR ALWAYS-ON DISPLAY

Google plans to add a new capability called Min Mode to Android's Always-On Display (AOD) in Android 17. Min Mode will use the same low-power technology as AOD to display full-screen interfaces from apps. Developers must register a 'MinModeActivity' component in their apps for them to work with the feature. Min Mode has the potential to reshape the definition of an 'idle' phone.

TIM COOK SAYS MORE AIS ARE COMING TO APPLE INTELLIGENCE

Apple plans to embed more third-party AI tools into its operating systems. The company is on track to release an AI-upgraded Siri next year. It is open to pursuing mergers and acquisitions if it thinks it will advance its roadmap. The company recently released fourth quarter earnings results that showed it raked in a record $102.5 billion over the past few months, an 8% increase compared to the same period last year.

SPACEX AND BLUE ORIGIN BOTH SUBMITTED PLANS TO GET ASTRONAUTS BACK TO THE MOON FASTER, NASA SAYS

SpaceX has pitched NASA a 'simplified mission' to put astronauts back on the moon following criticisms over delays. The company is behind schedule on building its lunar landing system for NASA's Artemis III mission, which caused NASA to reopen the landing contract for the mission. Blue Origin also submitted a plan. It has been developing a lunar lander for NASA and plans to launch a smaller-scale version of it, known as the Blue Moon Mark 1.

WORLD'S FIRST UNDERWATER HABITAT IN DECADES TO UNLOCK SECRETS OF 95% OF OCEANS

DEEP, a UK-based subsea engineering firm, has unveiled a prototype of an underwater habitat designed to allow scientists to live and carry out research beneath the sea's surface for days at a time. The Vanguard pilot human habitat is part of an ambitious effort to make long-term underwater living possible. The prototype can support up to four crew members for up to a week. Currently, most expeditions can only allow scientists to stay submerged for a few hours at a time.

ML KIT'S PROMPT API: UNLOCK CUSTOM ON-DEVICE GEMINI NANO EXPERIENCES

AI is making it easier to create personalized app experiences. The ML Kit GenAI Prompt API for Android enables developers to send natural language and multimodal requests to Gemini Nano to support custom, app-specific GenAI use cases. It allows developers to create unique features with complex data transformation. The API uses Gemini Nano on-device to process data locally, enabling offline capability and improved user privacy.

SCALE DESIGN DECISIONS ACROSS TEAMS AND PRODUCTS

The Design Tokens W3C Community Group (DTCG) is a community of designers, developers, and tooling vendors that aims to standardize how design tokens are defined and exchanged. It stewards the Design Tokens specification so that teams can share visual language reliability across design tools, code bases, and platforms. The first stable version of the DTCG JSON format is now available. The format unlocks interoperability and theming between tools.

HUMANS NEED ENTROPY

As people get older, they tend to repeat the same patterns. The solution to preventing this is to find sources of entropy. When we were kids, everything was entropy because everything was new. As adults, we need to constantly seek out the new to maximize the entropy in our lives.

TAKING MONEY OFF THE TABLE

Employees who get compensated with shares sometimes have the opportunity to sell a portion of them, but it can be difficult to choose whether to sell or to roll the dice and risk it longer. They should take the money while it's available. The amount will likely be a life-changing amount of money, and there's never a guarantee that their startup will succeed, even if it is looking good right now. The money you take today can be invested in more steady growth.

Microsoft. Claim your offer

THE REAL PROBLEM WITH AI CODING

The problem with AI coding is comprehension debt - the more AI-generated code there is in a codebase, the harder it is for human engineers to review and truly understand it.

WHAT PALANTIR SEES

An interview with Shyam Sankar, the chief technology officer of Palantir, a company building a more efficient surveillance state.

‘IT'S CULTURE': AMAZON CEO SAYS MASSIVE CORPORATE LAYOFFS WERE ABOUT AGILITY — NOT AI OR COST-CUTTING

The recent layoffs weren't triggered by financial strain or artificial intelligence replacing workers - Amazon executives believed that the company had gotten too big and layered.

Subscribe now.

THE AGI RACE IS AN ALL‑PAY AUCTION. THAT'S WHY “OVER‑INVESTMENT” IS RATIONAL

The imagined prize with AGI is monopoly-like profits across software, science, society, and the whole fabric of human civilization.

MOST OF WHAT WE CALL PROGRESS

Most progress in software is just motion, not real development.

BEYOND IP LISTS: A REGISTRY FORMAT FOR BOTS AND AGENTS

Cloudflare's new registry format for bots and agents ensures customers have control over the traffic they want to allow while fostering an open curation ecosystem that doesn't lock in customers or small origins.

Register for BUILD 2025

OPENAI NOW SELLS EXTRA SORA CREDITS FOR $4, PLANS TO REDUCE FREE GENS IN THE FUTURE

OpenAI is selling extra credits for its Sora AI video generation tool through Apple's App Store. An extra 10 video generations costs $4. The app currently has a limit of 30 free generations per day. That rate will likely decrease as OpenAI starts to monetize the platform. OpenAI reportedly plans to monetize Sora by letting entities license out their copyrighted material.

CANVA LAUNCHES ITS OWN DESIGN MODEL, ADDS NEW AI FEATURES TO THE PLATFORM

Canva has launched a foundational model trained on its design elements that generates designs with editable layers and objects. The model works across different formats, including social media posts, presentations, whiteboards, and websites. Canva also recently introduced new products and features, updates to its AI assistant, and the ability to use its spreadsheet tool alongside its app-building feature to create data visualization widgets. More details about the recent launches are available in the article.

BIG TECH IS SPENDING MORE THAN EVER ON AI AND IT'S STILL NOT ENOUGH

Big tech is planning to pour $400 billion into artificial intelligence efforts this year, but they all say it's nowhere near enough. Meta, Alphabet, Microsoft, and Amazon have all recently told investors that they will increase spending in 2026. Investors are uncertain about where the outsized spending will ultimately end. Whoever gets to AGI first will have an incredible competitive advantage over everybody else, so everyone is spending as much as they can.

HOW WELL DOES RL SCALE?

Improving AI capabilities either involves scaling the amount of compute used for RL during training or scaling the amount of compute used for inference during deployment. It's important to know where capability gains come from because scaling up the inference compute has very different implications than scaling up the training compute. While RL has provided impressive gains, we've reached a point where it is too expensive to go much further. This leaves inference-scaling as the remaining form of compute-scaling.

HOW FAST CAN AN LLM GO?

This article takes a look at how good inference software is becoming. It walks through calculations to provide estimates on what different hardware settings can theoretically achieve. Readers can adjust system configurations to change the values throughout the article.

on the right track: get the checklist.

HOW WE BUILT OWL, THE NEW ARCHITECTURE BEHIND OUR CHATGPT-BASED BROWSER, ATLAS

OpenAI built a new architectural layer called OWL (OpenAI's Web Layer) for Atlas, the company's ChatGPT-based browser, to run Chromium's browser process outside of the main Atlas app process. By moving Chromium out of the main application process and into an isolated service layer, OpenAI unlocked a simpler, modern app, faster startup, isolation from jank and crashes, fewer merge headaches, and faster iteration. Development can go faster because most of OpenAI's engineers don't need to build Chromium regularly from source. This article explains how OWL works.

INTRODUCING AARDVARK: OPENAI'S AGENTIC SECURITY RESEARCHER

Aardvark, currently in private beta, is a GPT-5-powered agent that autonomously scans code repositories to find security vulnerabilities, validate exploitability, and propose patches. It monitors commits in real-time, generates threat models for entire repositories, and integrates directly with GitHub workflows to deliver one-click patches, similar to Google's CodeMender.

IMPOSSIBLEBENCH: MEASURING REWARD HACKING IN LLM CODING AGENTS

LLM-powered coding agents have been observed exploiting loopholes in tests or scoring systems rather than solving the actual tasks specified. ImpossibleBench was created to systematically measure this behavior. Its creators took existing coding benchmarks and manipulated their unit tests to directly conflict with the natural language specifications to create impossible tasks where models must choose between following instructions or passing tests. Their 'pass rate' on these impossible tasks is a direct measure of reward hacking.

KIMI LINEAR TECH REPORT HAS DROPPED! 🚀

Kimi Linear is a novel architecture that outperforms full attention with faster speeds and better performance. It offers up to a 75% reduction in KV cache usage and up to 6x decoding throughput at a 1M context length. Kimi Linear's open-sourced KDA kernels can be used as a drop-in replacement for full attention. The two models available were trained with 5.7T tokens.

THE SMOL TRAINING PLAYBOOK: THE SECRETS TO BUILDING WORLD-CLASS LLMS (TEXTBOOK)

Hugging Face released an exhaustive, behind-the-scenes technical guide documenting what it's really like to train an LLM from scratch. It covers the entire lifecycle from small-scale ablations and infrastructure debugging to post-training and the decision to restart after 1 trillion tokens.

THE SECRETS OF CLAUDE CODE FROM THE ENGINEERS WHO BUILT IT (1 HOUR VIDEO)

Claude Code creators, Cat Wu and Boris Cherny, discuss the product philosophy and technical workflows behind Anthropic's coding agent. They cover how their engineers use competing subagents for cleaner results, the team's "unshipping" approach to balance simplicity with power, and future form factors to make the tool more autonomous and accessible to non-technical users.

MICROSOFT OUTLOOK IS GETTING AN AI OVERHAUL UNDER NEW LEADERS

Microsoft is overhauling Outlook with AI, aiming to transform it into an assistant-like tool.

PATENT RESEARCH AGENT BY PERPLEXITY

Perplexity Patents is a new tool designed to simplify patent research using natural language queries and AI-powered citation search.

NVIDIA RELEASED OPEN MODEL FAMILIES FOR SPECIALIZED AI

NVIDIA has introduced four open model families aimed at accelerating development across language, biology, and robotics: Nemotron (digital AI), Cosmos (physical AI), Isaac GR00T (robotics), and Clara (biomedical).

MASTERCARD POISED TO ACQUIRE ZEROHASH FOR ~$2 BILLION

Mastercard is in late-stage talks to acquire crypto infrastructure startup Zerohash for $1.5-2 billion, following the company's $104 million funding round in September at a $1 billion valuation. The potential acquisition comes after Mastercard lost a bidding war for stablecoin startup BVNK to Coinbase, which entered exclusivity for a deal reportedly worth around $2 billion. Zerohash, founded in 2017, provides broader infrastructure than stablecoin-focused competitors like Bridge and BVNK, offering APIs for crypto trading platforms and tokenization alongside stablecoin payment rails.

HONG KONG REGULATOR WARNS OF RISKS IN DIGITAL ASSET TREASURIES

Hong Kong's Securities and Futures Commission (SFC) has raised concerns over Digital Asset Treasuries (DATs) trading at inflated premiums and warned investors about related risks. The regulator is monitoring DAT operations and may introduce guidelines to ensure investor protection, citing cases in the US where DAT stocks traded far above the value of their crypto holdings. The Hong Kong Stock Exchange has already rejected several DAT proposals, maintaining restrictions against firms holding large liquid crypto reserves as DAT momentum builds globally.

X402 STARTER KIT

The x402 starter kit allows developers to build paid APIs requiring USDC micropayments before processing requests, with three core components: an Express server, the Merchant Executor for payment verification/settlement, and a service demonstrating OpenAI/EigenAI integration. The kit supports both facilitator-based settlement and local direct settlement modes across multiple networks, including Base, Polygon, Avalanche, and Solana, with payment verification using EIP-3009 authorization signatures. Developers can deploy custom frontends and can configure AI providers (OpenAI or EigenAI), network parameters, and settlement modes through environment variables, with the full payment flow handling 402 responses, signature verification, blockchain settlement, and service execution.

SCALING BASE: DOUBLING CAPACITY IN 30 DAYS

Base plans to double its gas limit from 75 to 150 Mgas/s by the end of Q4 2025 and reach 400-500 Mgas/s by early 2026 by migrating infrastructure to Reth and addressing four scaling bottlenecks that previously constrained growth. The Pectra hardfork doubled Ethereum blob space in May, with PeerDAS (Fusaka fork) launching December 3, followed by Blob Parameter Only forks increasing blob targets from 6 to 14 by January 2026, while TrieDB database restructuring promises 8-10x execution performance improvements for storage operations. Base successfully handled 1,500 transactions per second in June with median fees under 5 cents, targeting sub-cent, sub-second transactions for billions of users while achieving 99.9% of blocks built in under 2 seconds.

HOW A2A, AP2, AND X402 ENABLE AGENTIC COMMERCE

A2A, AP2, and x402 form the communication and transaction stack that lets AI agents autonomously interact, agree, and pay one another. A2A defines the base grammar for agent communication, AP2 extends it for auditability with cart mandates detailing price and terms, and x402 (by Coinbase) executes verified onchain payments. In a working example, Bankless's agent hired an AI debate judge, sending a request via A2A, receiving a price quote via AP2, attesting and paying through x402, and receiving the debate output stored on EigenDA. Together, these protocols define how agents talk, agree, and pay, forming the foundation for autonomous agentic commerce.

THE “STABLECOIN TAX” AND THE CREDIBLE NEUTRALITY DEBATE

Major chains like Solana, BSC, and Arbitrum are leaking over $1.1B annually in stablecoin yield to issuers like Circle and Tether (40% more than their combined fee revenue). On Solana alone, roughly $15B in stablecoins (two-thirds in USDC) generates about $500M yearly for Circle, part of which goes to Coinbase via its revenue-sharing deal, funding Base, Solana's direct competitor. Developers would like to launch an aligned, native stablecoin and redirect yield toward SOL buybacks or burns. However, others warn that if L1s compromise credible neutrality to capture yield, they risk turning from property rights systems into fintech companies, sacrificing the trust that gives blockchains long-term legitimacy.

THE BROKEN INCENTIVES BEHIND DEFI VAULTS

Most DeFi yield vaults are structurally flawed, prioritizing opaque incentives and TVL growth over user safety or transparency. Many vaults obscure wallet activity, misreport APYs, and recycle deposits to inflate their own metrics, turning “DeFi” into centralized yield farming. Whitelisted vaults with predefined, verifiable yield sources enforced by smart contracts (like STEX AMM), and competitive vaults where strategists must post insurance capital and compete for deposits, bearing losses if risks backfire (as seen in Cap Money Strategist), are both safer options. In both models, incentives are realigned toward defined, transparent risk exposure, rather than the current standard, where users unknowingly shoulder all the risk.

ETHEREUM'S FUTURE TECH STACK IS WHAT INSTITUTIONS ARE BUILDING ON

Dismissing Ethereum for high gas fees is like rejecting the internet in the dial-up era, as banks are adopting today's Ethereum stack with tomorrow's upgrades. Institutional builders aren't choosing 2015 Ethereum but the 2026+ ecosystem: Fusaka scaling to 12k+ TPS, zk-L2s like Lighter nearing Stage 2 decentralization, and ERC-7683 introducing seamless interoperability. With MegaETH targeting 100k+ TPS and zk tech soon pushing Ethereum's L1 beyond 10k TPS, the network's secure base layer and scalable rollup stack make it the most institution-ready platform for the onchain economy.