Search the archive

THE FUTURE OF PYTHON WEB SERVICES LOOKS GIL-FREE

Free-threaded Python has the potential to be a huge quality-of-life improvement for those managing thousands of ASGI and WSGI Python containers on vast infrastructure.

X IS LAUNCHING A MARKETPLACE FOR INACTIVE HANDLES

X Premium Plus and Premium Business customers will soon be able to browse and request inactive usernames on the X Handle Marketplace.

THE FIRST NEW COMMERCIAL BANANA IN 75 YEARS IS COMING TO THE MARKET NEXT YEAR

UK-based researchers used CRISPR technology to delete the specific gene responsible for browning in Cavendish bananas.

FIRST LONG-ACTING HAIR REGROWTH PILL HAILED AS 'NEW BENCHMARK'

VDPHL01, a re-engineered extended-release version of oral minoxidil, could be the first new treatment for pattern hair loss in nearly 30 years.

GENAI TRAFFIC SHARE UPDATE

ChatGPT continues to lose share, and Perplexity has exceeded a 2% threshold.

AS ANTHROPIC TRIES TO KEEP PACE WITH OPENAI, IT'S ALSO TAKING ON THE US GOVERNMENT

David Sacks, the Trump administration's AI and crypto czar, has publicly criticized Anthropic for running a sophisticated regulatory capture strategy based on fear-mongering.

Implement MCP Auth with WorkOS →

GROUNDING WITH GOOGLE MAPS: NOW AVAILABLE IN THE GEMINI API

Google Maps integration is now available in the Gemini API, enabling developers to incorporate Maps data into geospatial-aware AI applications. The tool enhances use cases in travel, real estate, retail, and logistics by providing precise, location-based responses. The tool's combination with Google Search further boosts response quality, offering structured data and timely information.

APPLE EMPLOYEES HAVE 'CONCERNS' OVER SIRI PERFORMANCE IN EARLY IOS 26.4 BUILDS

Apple engineers testing the early iOS 26.4 build have concerns about Siri's performance despite extended development time. Internal teams are working on two Siri models, one using on-device models and the other powered by Google Gemini. Concerns persist about Apple's AI strategy, potentially leading to reliance on external solutions.

ALIBABA CLOUD CLAIMS TO SLASH NVIDIA GPU USE BY 82% WITH NEW POOLING SYSTEM

Alibaba Cloud's new Aegaeon system reduces Nvidia GPU usage by 82%, cutting the number of GPUs needed for large AI models from 1,192 to 213. Beta tested in their model marketplace, Aegaeon addresses resource inefficiency by pooling GPU power. This pooling allows one GPU to serve multiple models, optimizing concurrent LLM workload management.

SUMMONING GHOSTS WITH ANDREJ KARPATHY (146 MINUTE VIDEO)

In a wide-ranging conversation, Andrej Karpathy discusses AGI's economic impact, the limits of reinforcement learning, lessons from self-driving research, and his vision for the future of education.

ANDREJ KARPATHY ON AGI TIMELINES, ANIMALS VS GHOSTS, AND MORE

Andrej Karpathy was recently on a podcast where he discussed AGI timelines, Animals vs Ghosts, Reinforcement Learning, AI agents, and job automation. Some of the explanations on the podcast were botched. This post expands on the topics that Karpathy discussed on the podcast to clear things up. It includes links to previous discussions Karpathy has had on these topics.

AN OPINIONATED GUIDE TO USING AI

The biggest mistake people make with paid AI subscriptions is staying in auto mode, where the system picks weaker models instead of manually selecting agentic modes that are far more capable and less error-prone for real work. This guide covers practical choices like when free models suffice versus paying $20 monthly, why prompting techniques that worked on older models no longer matter, and why explicitly telling AI to act as a critic is necessary to avoid talking to a sophisticated yes-man.

Atlassian, and EY SMEs. See how to make agentic AI work for you.

EXPLORING OPENAI'S DEEP RESEARCH API MODEL O4-MINI-DEEP-RESEARCH

o4-mini-deep-research costs $2/$8 per million for input/output tokens. It is only available via the Responses API. OpenAI recommends running them using its background mode. This post takes a quick look at the model's abilities.

CORAL NPU FOR EDGE AI

Google's Coral NPU is an open-source platform designed for low-power, always-on AI at the edge. It tackles key issues like fragmentation, performance, and privacy for edge devices and wearables.

CONFORMAL PREDICTION ON TEMPORAL GRAPHS

NCPNET adapts conformal prediction to dynamic graphs by using a diffusion-based non-conformity score and efficiency-aware optimization.

WE ARE IN THE "GENTLEMAN SCIENTIST" ERA OF AI RESEARCH

Discoveries were simple in the early days of science. Many were discovered in the course of pursuing a hobby. To win the 2026 Nobel Prize in physics, you have to be a physicist fully immersed in the world of physics. AI research is still very much in the 'early days of science' category. While many papers look complicated, AI research discoveries are often simpler than they look. Many things we learn about AI capabilities come from informal user experimentation.

NVIDIA SHOWS OFF ITS FIRST BLACKWELL WAFER MANUFACTURED IN THE US

Nvidia has revealed the first Blackwell wafer made in TSMC's semiconductor manufacturing facility in Phoenix, Arizona. The Blackwell architecture is now ready for volume production. Nvidia is still working on expanding its manufacturing footprint across the US. It plans to funnel half a trillion dollars towards building AI infrastructure in the US through partnerships with TSMC, Foxconn, and other companies.

GLOBAL SECTOR TRENDS ON GENERATIVE AI

This report looks at global sector trends in generative AI, with the data ending on October 10. It covers sectors such as general AI tools, character and chat, design and image generation, writing and content, EdTech AI, customer support, and more. The report looks at total visits at the domain level and does not capture API usage or integrations.

GOOGLE AI STUDIO UPDATES

Google AI Studio has rolled out new UX improvements to reduce friction and unify the developer workspace.

OPENAI'S ‘EMBARRASSING' MATH

Meta's Yann LeCun and DeepMind CEO Demis Hassabis publicly mocked OpenAI after VP Kevin Weil claimed GPT-5 solved 10 previously unsolved Erdős problems, but the mathematician maintaining the list clarified they were only marked “open” because he personally didn't know of solutions.

BROADCOM'S 5TH CUSTOMER ($10B) ISN'T APPLE OR XAI

Anthropic is buying TPUs from Broadcom.

WHATSAPP CHANGES ITS TERMS TO BAR GENERAL PURPOSE CHATBOTS FROM ITS PLATFORM

Meta won't allow AI model providers to distribute their AI assistants on WhatsApp starting on January 15.

👉 Get in at $0.81 while the window is open

BITCOIN TREASURY COMPANIES STRUGGLE TO HALT PLUNGE

Bitcoin treasury companies attempting to replicate MicroStrategy's success have suffered massive investor losses ranging from 38% for Strategy to 94% for KindlyMD over the past three months, with declines beginning well before BTC's recent peak above $126,000 on October 3rd and subsequent 11-day plunge. Executive teams have been active on social media defending business models, with Metaplanet CEO Simon Gerovich explaining preferred stock issuance benefits, KindlyMD CEO David Bailey denying "FTX vibes" claims as the stock trades below $1 facing Nasdaq delisting, and Strive CIO Ben Werkman arguing valuations have reached "deep value territory" as companies move to a "second phase" leveraging intact balance sheets. While Strategy shares rose from $30 during 2022's crypto winter to $290 today despite recent declines, demonstrating potential for BTCTC comebacks, the current negative sentiment and widespread share price carnage suggest replicating MicroStrategy's first-mover success will require more than just rising bitcoin prices.

OPENSEA'S SEA TOKEN LAUNCH AND REVENUE BUYBACK PLAN

OpenSea CEO Devin Finzer announced that the platform's long-awaited SEA token will launch in Q1 2026, with 50% of the supply going to OG users and rewards participants, and half of the platform's revenue at launch allocated to token buybacks. The release coincides with OpenSea's pivot into a multi-chain crypto trading hub supporting perps, NFTs, and tokens, alongside a forthcoming mobile app rollout.

PICO PRISM: 99.6% REAL-TIME PROVING FOR ETHEREUM ON CONSUMER HARDWARE

Brevis achieved 99.6% real-time proving coverage for 45M gas Ethereum blocks using 64 consumer-grade RTX 5090 GPUs across 8 servers, with 96.8% of blocks proven under the Ethereum Foundation's 10-second standard and 6.9-second average proving time. The system delivers 3.4x better performance efficiency than Succinct's SP1 Hypercube state-of-the-art, achieving 98.9% versus 40.9% coverage under 10 seconds for 36M gas blocks, 6.04-second versus 10.3-second average proving time, and $128k versus $256k GPU costs using half the hardware with 64 GPUs instead of 160. Pico Prism's breakthrough comes from completely redesigning the computational pipeline into embarrassingly parallel phases with computation-heavy workloads shifted to GPUs while setup tasks remain on CPUs, delivering near-linear acceleration when scaling across multi-machine, multi-GPU setups. The team's roadmap targets 99% real-time proving with fewer than 16 RTX 5090 GPUs in the coming months while maintaining the Ethereum Foundation's goals of sub-$100k hardware capex and sub-10kW power draw for on-premises proving.

UNISWAP ADDS SOLANA SUPPORT

Ethereum's largest decentralized exchange Uniswap has integrated Solana support into its web app, providing easy access to Solana, Ethereum, and 13 other chains in one interface and potentially unifying the two largest DeFi ecosystems. The integration simplifies user experience by allowing traders to connect Solana wallets and trade SOL tokens alongside Ethereum assets without using complex bridges or switching between different wallets and applications, removing barriers that previously affected both novice and experienced traders. Uniswap processes over $3.7 billion in 24-hour trading volume according to DeFiLlama, and the cross-chain functionality enables users to access Solana's speed and low costs through the familiar Uniswap platform while addressing the DeFi fragmentation problem.

TOP FIVE 20%+ APY STABLE FARMS

High-yield stablecoin strategies include Reservoir's wsrUSD/USDC loop on Morpho delivering 21.21% net APR from 8.25% collateral yield minus 6.81% borrow cost with $17.64 million available liquidity plus Season 2 points, Maple Finance on Arbitrum offering 20.5% net from 6.73% collateral yield minus 5.2% borrow cost with $18.57 million liquidity plus Drips rewards, and Stream DeFi LPing xUSD with Euler-lent USDT on Plasma for 23% APR plus Stream Points without leverage exposure. Additional opportunities include Avant Protocol on Avalanche with senior tranches exceeding 20% APR that can be leveraged on Euler for higher returns, and Pendle's short-dated AIDaUSDC Principal Tokens delivering 30% actual APR with 1% ROI on $100k deposits, though cap is hit, requiring purchases around peg on secondary markets.

STATE OF CRYPTO FUNDRAISING: Q3 2025

Total capital raised in Q3 2025 increased 47% quarter-over-quarter to $8.21 billion, reversing a two-year downtrend and marking the highest quarterly fundraising since Q1 2023, though deal count declined 16% to 480 transactions while investor participation rose 13% to 1,200 unique investors from Q2's record low. The top ten funding rounds raised $4.27 billion, led by Bullish's $1.11 billion IPO, surpassing Circle as the largest crypto IPO since Coinbase, Figure's $788 million IPO, pump.fun's $600 million public token sale at $4 billion valuation, Kraken's $500 million round at $15 billion valuation, preparing for 2026 IPO, and Rapyd's $500 million Series F, demonstrating a shift toward public market participation over token-based fundraising. M&A activity remained robust with 80 deals despite 6% quarterly decline from Q1's record 85 deals, with top acquisitions including Solowin's $350 million AlloyX purchase, Ripple's $200 million Rail acquisition for stablecoin payments, LayerZero's $120 million Stargate Finance acquisition after four-way bidding war, Polymarket's $112 million QCEX purchase for US market re-entry, and Talos's $100 million Coin Metrics acquisition backed by Citi. Newly launched funds raised $2.31 billion led by Sora Ventures' $1 billion Bitcoin Treasury Fund, 50T Fund V's $500 million growth equity vehicle, and Ondo Catalyst's $250 million RWA-focused fund, while Payments, Gaming, and DEX sectors led deal activity with AI Agents and RWAs showing steepest accelerator growth despite overall accelerator activity declining 56% to 88 deals.

STABLECOINS ARE 'CENTRAL BUSINESS DIGITAL CURRENCIES'

Privately-issued stablecoins are "central business digital currency" featuring all the surveillance, backdoors, programmability, and controls of CBDCs, plus unique risks, and that issuers like JPMorgan could freeze funds and unbank users through the Patriot Act or future regulations. Overcollateralized stablecoins backed by cash and short-term government securities face bank run risks if too many holders attempt simultaneous redemptions, while algorithmic and synthetic stablecoins carry counterparty risks and de-pegging vulnerabilities from volatility or flash crashes in crypto derivatives markets.

THE TWO-STABLECOIN FUTURE

Every fiat currency will eventually have two stablecoins. One is a “spending” version for daily commerce, and the other, a “savings” version backed by tokenized government bonds. This dual model will mirror power laws in FX, where most coins never rival USDT or USDC but still enable frictionless participation in local economies. Governments will back the shift to prevent dollarization, while onchain currencies will cut out FX middlemen, keeping more value with workers and businesses.

ETHEREUM RESEARCHER DANKRAD FEIST JOINS STRIPE-BACKED TEMPO

Ethereum researcher Dankrad Feist, best known for co-creating Danksharding and advancing L1 scaling, has left the Ethereum Foundation to join Tempo, a Stripe- and Paradigm-backed Layer 1 focused on real-world payments and stablecoin infrastructure.

RIBBIT CAPITAL'S “TOKEN LETTER” DEFINES THE NEXT DECADE

Ribbit Capital's Token Letter lays out a sweeping vision for the “Token Revolution,” where tokens become the building blocks of intelligence, identity, and value.

ETHEREUM SHOULD NOT COMPETE WITH TEMPO

Ethereum must not compete head-on with Stripe's Tempo chain, as centralized payment-focused L1s will outpace it on speed, integrations, and regulatory compliance.

THE ULTIMATE AAVE ANALYTICS PORTAL

Blockworks published an expansive Aave analytics dashboard featuring 113 charts across 9 panels, covering financials, lending markets, interest rates, and borrower behavior.

PINTEREST'S ‘TUNER' LETS YOU DIAL DOWN THE AMOUNT OF AI CONTENT — BUT NOT ENTIRELY

Pinterest has introduced a "tuner" feature that allows users to reduce AI-generated content in their feeds across categories like beauty, art, fashion, and home decor, though the platform doesn't guarantee complete elimination of AI content. The tool, found in settings under "refine your recommendations," is now available on Android and desktop, with iPhone access coming soon. This follows Pinterest's May rollout of "AI modified" labels on pins, which the company says it's making more noticeable alongside enhanced detection capabilities.

PUSH MAKE PROJECT TO GITHUB

A new feature allows Figma users to push Make projects directly to new GitHub repositories, enabling code backup and version history tracking. The functionality, rolling out this week, lets users continue development in their preferred tools. Users can push ongoing updates from Make to GitHub whenever they make changes to their projects.

THESE PRETTY NEW NEST LEARNING THERMOSTAT FACES WILL TURN YOUR TEMPERATURE CONTROLLER INTO A WORK OF ART

Google has released a new set of animated faces for the Nest Learning Thermostat, including a botanical series created by illustrator Mathilde Loubes that changes monthly to reflect the seasons. These new designs, along with minimalist and classic options, aim to make the thermostat blend more seamlessly into home decor while maintaining core functionality like displaying temperature and time.

WHAT PERPLEXITY'S AI BROWSER REVEALS ABOUT UX'S FUTURE

Perplexity's Comet browser demonstrates a fundamental shift from navigation-based to intention-based interfaces, where users state goals in natural language rather than manually navigating to information. Testing revealed users mentally adapt to this paradigm within days, even though current AI reliability remains at 60-70%. The research suggests that the competitive future lies not in monolithic AI systems, but in specialized AI services that collaborate through shared intention contexts. This requires designers to prioritize transparent failure communication and cross-platform intention continuity over traditional engagement metrics.

DEFAULT ISN'T DESIGN

Framework monoculture persists partly through System Justification Theory—a psychological reflex where people defend existing arrangements as legitimate even when harmful, driven by desires for certainty, safety, and group belonging. This manifests in front-end development through dismissive language about standards and false tradeoffs between shipping and interoperability, quietly accumulating costs like reduced portability, slower innovation, and user-shifted burdens. A standards-first approach offers a pragmatic path that preserves developer experience while ensuring work remains lean, portable, and future-proof across teams and platforms.

THE ILLUSION OF ALIGNMENT

People often believe they're aligned when using the same words, but each person imagines something different. Visual thinking—through sketches and prototypes—reveals these hidden differences early, turning abstract ideas into shared understanding and preventing miscommunication before it derails collaboration.

CREATE REMARKABLE CLIENT EXPERIENCES

Give customers a branded portal that lets you unify your client experience, from first sale to full scale.

HUMAN MADE VECTOR ILLUSTRATIONS

Enhance designs with unique illustrations. Access a vast collection with a free trial now.