Search the archive

Secure your spot today.

APPLE'S NEW ‘ANSWERS' TEAM EYES CHATGPT-LIKE PRODUCT IN AI PUSH

Apple quietly formed a new team called Answers, Knowledge, and Information earlier this year. The group is exploring a number of in-house AI services with the goal of creating a new ChatGPT-like search experience. Apple recently started advertising job openings for the team, with several listings specifically mentioning experience with search algorithms and engine development.

REDDIT PLANS TO UNIFY ITS SEARCH INTERFACE AS IT LOOKS TO BECOME A SEARCH ENGINE

Reddit has ambitions to become the go-to search engine people use. Its search function has more than 70 million weekly users, and Reddit Answers, its AI-powered Q&A product, is now used by 6 million people. Reddit Answers is currently available in a dozen countries. The company wants to eventually unify the core search function with Reddit Season certificates and redesign its app to have the search box be more prominently visible.

THESE DRONES DROP BURNING BALLS IN THE FOREST TO CONTROL WILDFIRES

One of the oldest methods of preventing wildfires is doing controlled burns, which preemptively remove leaves, dead branches, and other dry materials. Nebraska-based startup Drone Amplify uses small quadcopter drones carrying hundreds of ping-pong ball-sized combustible eggs to ignite small, trackable, and contained fires. Its drones are currently operating more than 200 systems in the US, Canada, Germany, and Australia. Using drones enables better monitoring as well as nighttime operations.

SYNTHETIC E. COLI WITH COMPRESSED DNA OUTSMARTS VIRUSES IN RADICAL GENETIC LEAP

Researchers at Cambridge have created synthetic E. coli bacteria with just 57 codons - virtually all known life has 64. Syn57 has the most radically compressed genetic code ever built. The technique used to create the strain opens up the possibility of future organisms that can create unnatural materials and virus-resistant biofactories. Syn57 may be immune to many viruses, as viruses rely on the host's genetic code to replicate.

Schedule a demo to learn more

DEEP AGENTS

Deep Agents is a Python package that implements a planning tool, sub agents, access to a file system, and detailed prompts in a general purpose way so that developers can easily create deep agents for their applications. The project was primarily inspired by Claude Code and was initially an attempt to see what made Claude Code general purpose.

HOW TO WRITE A GOOD DESIGN DOCUMENT

A design document is a technical report that outlines the implementation strategy of a system in the context of trade-offs and constraints. The goal of a design document is to convince the reader that the design is optimal given the situation. The act of writing a design document helps the author think through designs and add substance to otherwise vague intuitions. This post walks through how to write high quality design documents.

APPLE CEO TELLS STAFF AI IS ‘OURS TO GRAB' IN HOURLONG PEP TALK

Apple CEO Tim Cook recently rallied employees around the company's artificial intelligence prospects at a rare all-hands meeting following earnings results. He told staff at the meeting that the AI revolution is as big or bigger than the internet, smartphones, cloud computing, and apps. Apple was late to AI, and its AI tools fell flat when they were finally released. However, Cook noted that while Apple is typically late to promising new technologies, it invents the modern versions of those product categories.

ANTHROPIC CUTS OFF OPENAI'S ACCESS TO ITS CLAUDE MODELS

Anthropic has revoked OpenAI's access to its Claude family of AI models due to a direct violation of its terms of service. OpenAI was connecting Claude to its internal tools to compare its performance to OpenAI's own models in categories like coding, writing, and safety. Its technical staff were also using Anthropic's coding tools. An OpenAI spokesperson described the company's usage as 'industry standard'. OpenAI's API remains available to Anthropic.

GOOGLE ROLLS OUT GEMINI DEEP THINK AI, A REASONING MODEL THAT TESTS MULTIPLE IDEAS IN PARALLEL

Google Ultra subscribers will gain access to Gemini 2.5 Deep Think, Google's most advanced AI reasoning model, in the Gemini app on Friday.

FULL-BREADTH DEVELOPERS

Developers with both technical and product capabilities stand to gain as clear winners in the age of AI.

THE BEST ADVICE I GOT AT PAGERDUTY

Seven bits of advice gained over seven years.

HIDING SECRET CODES IN LIGHT PROTECTS AGAINST FAKE VIDEOS

Researchers at Cornell have developed a way to watermark light in videos, making it possible to detect if a video is fake or has been manipulated.

BING IS GROWING WITH THE HELP OF AI AS GOOGLE SLIGHTLY SHRINKS

Bing is the search backend for some AI agents, such as ChatGPT and Meta AI, giving its numbers a bit of a boost.

ELON MUSK IS BRINGING BACK VINE'S ARCHIVES THAT WERE THOUGHT TO BE LOST MEDIA

Twitter bought Vine almost a decade ago.

READ THE AWS CASE STUDY

GOOGLE RELEASES GEMINI 2.5 DEEP THINK FOR AI ULTRA SUBSCRIBERS

Google's most powerful AI model generates multiple solution approaches simultaneously before selecting the best answer. A specialized version that can work for hours on complex problems recently achieved IMO gold medal status and is being shared with select mathematicians.

ANTHROPIC CUTS OFF OPENAI'S ACCESS TO ITS CLAUDE MODELS

Anthropic revoked OpenAI's access after discovering the company was using Claude for internal benchmarking and coding assistance ahead of GPT-5's launch, violating terms that forbid using its AI to create and improve competitor models. OpenAI will retain access for safety evaluations and research collaboration on harmful capabilities testing.

OPENAI RAISES $8.3 BILLION AS PAID CHATGPT BUSINESS USERS REACH 5 MILLION

The round was part of the broader $40 billion fundraise and completed ahead of schedule, driven by OpenAI's annual recurring revenue jumping from $10 billion to $13 billion since June and projected to hit $20 billion by year-end. Dragoneer led with a $2.8 billion investment, 10% of its entire fund.

WILL DATA CENTERS CRASH THE ECONOMY?

Big Tech's $102.5 billion quarterly AI spending now exceeds dot-com era investment as a share of GDP, but the real risk lies in financing through "private credit" funds that operate outside traditional banking oversight. Banks have increased their exposure to these shadow lenders 14x since 2013, creating conditions where a synchronized AI market downturn could trigger defaults on data centers, cascading through private credit funds to their bank and insurance company lenders.

TOKENS ARE GETTING MORE EXPENSIVE

Models will always be cheaper next year, but users will expect more of them. The goalpost is always moving ahead. Even Anthropic, which has the most vertically integrated application layer on the planet, can't make a flat subscription with unlimited usage work. Companies without a plan are headed to the graveyard fast, as there are no investors writing checks for negative margin businesses. There's no figuring it out later when your AWS bill is larger than your revenue.

ANDREW NG ON CHINA'S AI MOMENTUM

In a detailed thread, Andrew Ng argued that China's open-model ecosystem, rapid semiconductor advances, and aggressive business dynamics give it a credible path to surpass the U.S. in AI.

→ Register now

DOES THE BITTER LESSON HAVE LIMITS?

General methods that leverage computation are ultimately the most effective by a large margin. The eventual success of approaches based on scaling computation is often tinged with bitterness because it is success over a favored, human-centric approach. Developers need to be aware that their work might be overtaken by compute, but this shouldn't stop them from applying custom, human-crafted logic to get the job done.

PERSONA VECTORS: MONITORING AND CONTROLLING CHARACTER TRAITS IN LANGUAGE MODELS

Researchers extracted "persona vectors" by comparing neural network activity when models exhibit specific traits versus when they don't, revealing that personality changes follow predictable mathematical patterns. This discovery enables a counterintuitive "vaccination" approach that purposefully steers models toward undesirable traits during training to compensate for their tendency to “overfit” to negative traits.

COHERE'S COMMAND A VISION

Cohere has launched Command A Vision, an open-weight 111B parameter vision-language model designed for business use cases. It reportedly surpasses GPT-4.1 and Llama 4 Maverick on enterprise benchmarks.

SEED-PROVER: DEEP AND BROAD REASONING FOR AUTOMATED THEOREM PROVING

ByteDance's Seed-Prover proved 5 of 6 problems at IMO 2025, the third AI system to do so, and the first Chinese model, in recent weeks. The system combines "lemma-style" whole-proof reasoning that breaks problems into modular components with three-tiered inference strategies from quick refinements to days-long reasoning that accumulates thousands of mathematical facts.

HOW WINDSURF'S FOUNDERS GOT PAID FROM THE GOOGLE DEAL

New details show Google's $2.4B licensing deal with Windsurf split evenly between investors and compensation for 40 employees hired by Google, including co-founders Varun Mohan and Douglas Chen.

AI'S OVERLOOKED $97 BILLION CONTRIBUTION TO THE ECONOMY

The US economy grew at an annual rate of 3% in the second quarter. Americans enjoyed roughly $97 billion in consumer surplus (the difference between the maximum a consumer is willing to pay for a good or service and its actual price) from generative AI tools in 2024. Around 40% of US adults are regular users of generative AI. As more digital goods become available for free, measuring benefits as well as costs will become increasingly important.

now

AI REFERRALS TO TOP WEBSITES WERE UP 357% YEAR-OVER-YEAR IN JUNE, REACHING 1.13B

AI referrals to websites surged 357% year-over-year in June, reaching 1.13 billion, with ChatGPT responsible for over 80% of these referrals.

BUILD YOUR OWN AI APP BUILDER WITH THE V0 PLATFORM API

The v0 Platform API provides developers with a text-to-app interface, enabling automation of web app creation and code generation using LLM-generated UIs.

XAI TESTS TEAM ACCOUNTS FOR GROK AS COMPETITION HEATS UP

xAI appears to be testing a Teams plan for Grok that will let users toggle between personal and dedicated Team accounts.

RELEASE: POWER ATTENTION

Manifest AI has open-sourced Power Attention, which replaces Flash Attention in any transformer.

CIRCLE FACES MARGIN SQUEEZE AS USDC PAYOUTS RISE, SAYS MIZUHO

Circle earned $625M in Q2 interest income from USDC reserves but paid out $332.5M to Coinbase, raising concerns about shrinking margins as Circle adds new distribution partners like Binance. Despite USDC's growing supply (~$63B), analysts flagged stagnant circulation, looming rate cuts, and rising partner payouts as medium-term risks. They also warned that the GENIUS Act's regulatory clarity could invite more competition from giants like JPMorgan and Bank of America.

$3.5B LUBIAN BITCOIN THEFT WENT UNDETECTED FOR 5 YEARS

Arkham Intelligence claims to have uncovered evidence of the theft of 127,426 BTC from Chinese mining pool LuBian in December 2020, which would represent the largest cryptocurrency heist in history at $3.5 billion at the time of theft, now valued at $14.5 billion. The breach went undetected for nearly five years, with neither LuBian nor the suspected hacker ever publicly acknowledging the incident, though Arkham discovered over 1,500 onchain messages from LuBian pleading for the funds' return. The theft likely resulted from weak private key generation that enabled a brute-force attack, with the stolen Bitcoin remaining largely dormant except for a wallet consolidation in July 2024. LuBian controlled nearly 6% of Bitcoin's hash rate in 2020. The scale of this theft exceeds other notorious hacks like Mt. Gox and Bitfinex by nominal value at the time of loss.

AAVE V4 INTRODUCES UNIFIED LIQUIDITY HUB AND MODULAR SPOKES

Aave V4 will revamp DeFi lending with a “Liquidity Hub” that consolidates all assets into one central pool per chain, eliminating V3's fragmented market inefficiencies. Users interact through modular “Spokes” tailored for use cases like stablecoins (E-Mode), new assets (Isolation), tokenized RWAs, or vault collateral. Each Spoke has customized risk settings and authorization control, while the Hub enforces global accounting and withdrawal limits. The design allows anyone to build custom Spokes that plug into Aave's liquidity boosting composability, efficiency, and rate competitiveness across lending markets.

10-YEAR 'LEAN ETHEREUM' VISION: 10K TPS, 100% UPTIME, EVM 2.0

Justin Drake of the Ethereum Foundation laid out a “lean Ethereum” ten-year blueprint targeting 10,000 TPS at 1 gigagas/sec on Layer 1 and 1 million TPS at 1 teragas/sec on Layer 2, while preserving 100% uptime and decentralization. The plan calls for “bold upgrades” across all L1 sublayers, including Beacon Chain 2.0 for consensus, post-quantum Blobs 2.0 for data, and an RISC-V-based EVM 2.0 for execution, plus hash-based cryptography for quantum resistance. Emphasizing minimalism, modularity, formal verification, and provable security, the vision aims to deliver “performance abundance” under unforgiving continuity constraints.

FINANCE IS HAVING A NERVOUS BREAKDOWN

Financial systems are struggling under pressure. JPMorgan has begun charging for open banking amid regulatory disarray, and autonomous AI agents are trading stocks via browser tabs. A crypto treasury company holding BTC staged 2025's largest IPO, outpacing even Circle, as TradFi investors scramble for indirect exposure. Meanwhile, Ramp raised $500M at a $22.5B valuation just 45 days after its last round, betting on AI-powered finance teams. JPM's partnership with Coinbase reopens card rails, but only through closed deals, not open protocols. It all adds up to a market “mania moment” where innovation is real, incentives are distorted, and systemic risks are compounding fast.

BITCOIN TREASURYCOS: LESSONS FROM THE 1929 CRASH

STABLECOIN ADOPTION ISN'T A TECH PROBLEM, IT'S AN INCENTIVE PROBLEM

Stablecoin adoption is less about lower fees or faster settlement, and more about solving a multiplayer coordination game. Each transaction touches users, banks, and processors, none of whom want to move first due to switching costs like lost float or operational retraining. To succeed, builders must find wedge use cases where a single actor (like a cardholder) can adopt unilaterally and extract clear value. Stablecoin cards, for instance, grow faster than merchant rails because the user doesn't need to wait for ecosystem buy-in. Incentives, not just infrastructure, ultimately determine stablecoin success.

ONHAIN ANALYSIS SUGGESTS PARADIGM LIKELY HOLDS $765M+ IN HYPE

Paradigm may be the largest $HYPE holder, with ~19.1M tokens worth $765M at $40 p/HYPE. The analysis links a 4.18M HYPE transfer from Wintermute in December, publicly confirmed by Paradigm's Matt Huang, to 10 newly split wallets. Additional bulk withdrawals via FalconX and Gate.io, grouped by timing and structure, further support the theory. The estimated cost basis is ~$16.46/HYPE, implying ~$450M in unrealized gains. Though unconfirmed, the pattern of wallet activity, stealth setups, and ties to the Paradigm-affiliated entity “Sonn” strengthen the attribution. If accurate, Paradigm controls ~2% of total HYPE supply and ~6% of circulating supply.

CURVEDAO MEMBER PROPOSES HALTING L2 EXPANSION

A CurveDAO member has proposed ending future Curve deployments to new Ethereum Layer 2s, arguing they offer minimal revenue and drain developer resources.

HYPERION BECOMES FIRST PUBLIC COMPANY FARMING HYPERLIQUID POINTS

Nasdaq-listed Hyperion (HYPED) is farming Kinetiq's K-Points on Hyperliquid via iHYPE, a liquid staking derivative.