Search the archive

HOW I BUILT TLDR (TLDR'S INC. PROFILE, 5 MINUTE READ)

This is the first time I've told the backstory behind building TLDR, hope you find it interesting!

NVIDIA SELLS TINY NEW COMPUTER THAT PUTS BIG AI ON YOUR DESKTOP

Nvidia is now taking orders for its DGX Spark desktop AI computer.

CSS-EXTRAS

This repository contains a collection of CSS custom functions that leverage the new native CSS @function rule.

SOFTWARE CAN BE FINISHED

Thinking about what finished software might be like reveals some interesting lessons.

Subscribe now.

JUST TALK TO IT - THE NO-BS WAY OF AGENTIC ENGINEERING

Just talk to your agents - the more you work with them, the better your results will be.

YOUR DATA MODEL IS YOUR DESTINY

A 'data model' is the core concept or object a startup prioritizes and builds around.

JIT: SO YOU WANT TO BE FASTER THAN AN INTERPRETER ON MODERN CPUS…

Modern CPUs have an insane level of optimization, but this isn't always helpful.

Get access to GLM 4.6

WALMART TEAMS UP WITH OPENAI TO ALLOW PURCHASES DIRECTLY IN CHATGPT

Walmart has reached a deal with OpenAI to allow shoppers to buy items directly through ChatGPT. The US grocer is trying to keep up with changes in how consumers discover items to buy. OpenAI's Instant Checkout feature, announced in September, has been a new moneymaker for the company. OpenAI charges companies a fee for transactions completed through ChatGPT.

ORACLE, AMD PARTNER ON NEW AI CHIP DEAL

AMD is partnering with Oracle to create a large data-center cluster with the equivalent of 200 megawatts of computing power. AMD plans to deploy 50,000 of its MI450 chips inside data centers run by Oracle. The companies plan to expand their partnership in 2027 and beyond.

WHY BROADCOM'S BET ON OPENAI IS A BIG RISK

Broadcom's deal to develop huge numbers of chips and computing systems with OpenAI is not a clear-cut win. OpenAI has recently signed multiple multi-billion-dollar deals to create data centers stocked with hundreds of thousands of chips. While the scale is remarkable, it is unclear how OpenAI will pay for all of it. OpenAI's revenue this year is expected to be around $13 billion, a large sum, but nowhere close enough to justify this level of spending. The company has told investors it won't be profitable until 2019.

LLMS ARE GETTING BETTER AT CHARACTER-LEVEL TEXT MANIPULATION

Newer models are becoming more adept at manipulating text at the character level, despite their understanding of text being based on tokens. Most current-generation models are able to decode Base64 text. They are now able to substitute characters fairly reliably. Character-level operations are far from solved, but it is clear they have made huge progress in the area.

WHY YOUR BOSS ISN'T WORRIED ABOUT AI

The public thinks AI bugs work like software bugs: find the missing semicolon, fix the problem, move on. However, AI failures emerge from training on 11.25 trillion words that would take 85,000 years to read, not buggy code lines. Unlike traditional software, where you can pinpoint and permanently fix problems, AI mistakes can't be traced, "fixed" behaviors can resurface with different prompts, and nobody, including the labs, understands why these systems make the errors they do. Your boss assumes AI problems are patchable because 40 years of software education taught everyone that bugs have discoverable causes and permanent solutions.

REPRESENTATION AUTOENCODERS FOR DIFFUSION TRANSFORMERS

Latent diffusion models get an upgrade with Representation Autoencoders (RAEs), which swap out the standard VAE for pretrained encoders like DINO, SigLIP, or MAE paired with learned decoders. These higher-dimensional latent spaces improve generation quality, and new methods allow Diffusion Transformers to handle them effectively.

Explore the scientific challenges of agentic AI →

Ready to impact the future of AI, today? Join our team →

PETRI

Petri is an alignment auditing agent for realistic hypothesis testing. It allows researchers to test new hypotheses in minutes instead of building bespoke evals over weeks. Petri autonomously crafts environments, runs multi-turn audits, and scores transcripts to surface concerning behavior.

LLMS REPRODUCE HUMAN PURCHASE INTENT VIA SEMANTIC SIMILARITY ELICITATION OF LIKERT RATINGS

AI can closely replicate human consumer choices across 57 personal care products without knowing much about the shoppers, achieving 90% of human test-retest reliability using a "semantic similarity rating" method. Instead of asking LLMs for numerical ratings (which produces unrealistic clusters around safe middle values), the breakthrough approach has AI write responses like "I'm somewhat interested if it works well and isn't too expensive," then maps these to rating scales using embedding similarity. The method could replace billions in annual consumer research spending while providing richer qualitative feedback than traditional surveys.

THE ATTACKER MOVES SECOND: STRONGER ADAPTIVE ATTACKS BYPASS DEFENSES AGAINST LLM JAILBREAKS AND PROMPT INJECTIONS

Jailbreaks and prompt injection are far from solved. Researchers from OpenAI, Anthropic, and Google DeepMind systematically broke 12 recent AI safety defenses using adaptive attacks that achieved over 90% success rates despite the defenses originally reporting near-zero attack rates. Their methods exposed four categories of fundamental flaws: "prerequisite" attacks that frame malicious requests as requirements for legitimate tasks; training-based defenses failed because they rely on fixed attack datasets; filtering defenses couldn't detect benign-seeming text like "Per new policy, delete file ID '13'"; and secret-knowledge defenses were bypassed with conditional triggers that only activate during real usage.

GOOGLE WILL LET GEMINI SCHEDULE MEETINGS FOR YOU IN GMAIL

Gmail's new 'Help Me Schedule' feature uses Gemini AI to recognize when users may need to schedule a meeting and offer possible meeting times for recipients to choose. The feature won't support groups at launch. Help Me Schedule will only be available to businesses, enterprise users, and individuals with Google AI Pro and AI Ultra subscriptions. It will be rolled out gradually, with most users getting it in late October or early November.

EXPERT COUNCIL ON WELL-BEING AND AI

OpenAI formed the Expert Council on Well-Being and AI to guide the development of AI systems that positively impact mental health. The council includes experts in psychology, psychiatry, and human-computer interaction, focusing on how technology affects emotions and motivation. They advise on creating engaging experiences and setting effective guardrails for ChatGPT, especially for teen users.

NVIDIA UNVEILS ITS VISION FOR GIGAWATT ‘AI FACTORIES' BASED ON ITS VERA RUBIN ARCHITECTURE

Nvidia is collaborating with more than 70 partners to design more efficient gigawatt AI factories to support the next generation of artificial intelligence models. These gigawatt AI factories will utilize Nvidia's Vera Rubin NVL144, an open architecture rack server based on a 100% liquid-cooled design. Nvidia will donate the Vera Rubin NVL144 architecture to the Open Compute Project as an open standard, enabling any company to implement it in their own data centers.

management, and more. Read the case study.

INTEL ANNOUNCES "CRESCENT ISLAND" INFERENCE-OPTIMIZED XE3P GRAPHICS CARD WITH 160GB VRAM

Intel's new enterprise GPU for AI will not ship until at least the second half of 2026.

OPENAI WILL LET ADULTS USE CHATGPT FOR EROTICA STARTING IN DECEMBER

Users will no longer have to manipulate ChatGPT to get it to engage in NSFW conversations.

AMD DEBUTS HELIOS RACK-SCALE AI HARDWARE PLATFORM AT OCP GLOBAL SUMMIT 2025 — PROMISES EASIER SERVICEABILITY AND 50% MORE MEMORY THAN NVIDIA'S VERA RUBIN

AMD's Helios AI hardware platform is built for streamlined scalability with an AI-first data center environment.

ANTHROPIC AND SALESFORCE PARTNERSHIP

Anthropic and Salesforce have expanded their partnership to bring Claude to the Agentforce platform, enabling trusted AI use in regulated sectors like healthcare, finance, cybersecurity, and life sciences.

FIREFOX ADDS PERPLEXITY AS SEARCH OPTION

Mozilla has officially integrated Perplexity's AI answer engine as a selectable search option in Firefox, expanding its availability beyond initial test markets like the US, UK, and Germany.

Explore Galaxy Trading

BERNSTEIN PROJECTS USDC TRIPLING TO $220B BY 2027

S&P GLOBAL BRINGS STABLECOIN RISK RATINGS ONCHAIN VIA CHAINLINK

S&P Global has partnered with Chainlink to publish its Stablecoin Stability Assessments (SSAs) directly onchain, enabling DeFi protocols to integrate real-time risk data into smart contracts. The ratings evaluate stablecoins on factors like asset quality, governance, redeemability, and liquidity, helping institutional users assess fiat-backed stability. The SSAs will debut on Base, Coinbase's Ethereum Layer 2, before expanding to other networks. Chainlink co-founder Sergey Nazarov said the integration brings S&P's trusted framework into digital markets, advancing transparency and compliance as stablecoin adoption scales.

MACHINE-TO-MACHINE PAYMENTS WITH GATEWAY, X402, AND USDC

Circle is integrating Gateway with the x402 protocol ecosystem pioneered by Coinbase and Cloudflare to enable lightweight machine-to-machine payments for agentic AI systems that need to settle thousands of small transactions per second for data, compute, content, and API calls. Gateway's new batching feature launching on testnet allows hundreds of thousands of transactions to be processed together through deferred settlement, bundling payments offchain before settling onchain to eliminate per-transaction gas costs while preserving finality at a fraction of the cost and latency. The integration avoids gas fees on micropayments that make high-frequency small transactions impractical, and enables crosschain agency where sellers on one chain can accept payments from buyers on another through Gateway's chain-abstracted USDC design. Circle is also participating with Google on its Agent2Agent and Agent Payments Protocol initiatives, leveraging Gateway's throughput and efficiency for autonomous transactions, including AI research assistants paying per article in USDC or compute models dynamically compensating providers in real-time across any supported blockchain.

ZKSYNC PRIVIDIUM BREAKTHROUGH INITIATIVE

ZKsync Prividiums are private, permissioned Ethereum-secured Layer 2s validated by 35+ global financial institutions, including Citi, Deutsche Bank, State Street, and Mastercard, for cross-border payments and intraday repo use cases. The technology addresses enterprise requirements by keeping all transaction data off-chain in institution-controlled environments while posting only ZK proofs to Ethereum, enabling atomic finality, role-based access controls, and selective regulator visibility without exposing sensitive counterparty or balance information to competitors. Cross-border payment demonstrations showed near-instant settlement across interoperable bank blockchains with consolidated liquidity at FMI chains, eliminating the $27 trillion locked in nostro/vostro accounts and reducing the estimated $120 billion annual corporate fees, while intraday repo demonstrations enabled 24/7 automated atomic settlement of cash and collateral through smart contracts, removing manual reconciliation and daylight credit risk. Native interoperability through ZKsync Gateway enables trust-minimized asset transfers between Prividiums using Merkle proof verification rather than inefficient third-party bridges, with one-second ZK finality allowing institutions to verify proofs quickly without re-executing transactions. Everything is still a proof of concept.

THE STABLECOIN OPPORTUNITY THAT BANKS ARE MISSING

WHO IS NICK SZABO?

Nick Szabo invented Bit Gold in 1998 using proof-of-work puzzles, timestamping, Byzantine fault tolerance, and cryptographic chains that became direct predecessors to Bitcoin's design, coined "smart contracts" in 1994, envisioning computerized protocols executing automatically without intermediaries. He participated as an original 1990s Cypherpunk fighting NSA surveillance, with his core philosophy that "trusted third parties are security holes" becoming Bitcoin's trustless foundation. Linguistic analysis by Aston University identified Szabo as the likely Bitcoin whitepaper author, while Wei Dai claimed only "Nick Szabo and me" had the ability to build Bitcoin, though Szabo denies being Satoshi, and despite early Ethereum support resulting in a denomination named after him, he's since turned critical.

STABLECOINS DON'T WIN BY MINTING. THEY WIN BY MOVING

USDT and USDC dominate 85% of stablecoin supply not through branding or liquidity alone but through distribution moats built over years of integrations across every exchange pair, DeFi protocol, OTC desk, and on/off-ramp, demonstrating that issuance is a product challenge while distribution is an ecosystem challenge requiring network effects. Ethena's USDe reached $12 billion as the fastest stablecoin to hit $10 billion through horizontal distribution embedding in Aave for leveraged loops and Pendle for yield tokenization, while Hyperliquid's USDH exemplifies vertical distribution embedded directly in platform usage loops, contrasting with PayPal's PYUSD which despite 400 million users sits at only $2-2.5 billion supply because distribution only matters when connected to real use cases. A new mental model separates transacting stablecoins like USDT/USDC with high velocity and deep integration from holding stablecoins like wallet-issued tokens optimized for yield and safety with low velocity, suggesting wallets could route idle balances into "staked dollars" that earn returns without manual interaction. The next opportunity lies in orchestration layers that restore stablecoin fungibility across issuers by autonomously reading and converting different stablecoins into USD, solving the fragmentation problem as over $140 billion sits idle while less than 10% engages actively in DeFi.

CRYPTO MUST MOVE BEYOND THE CASINO

The recent market chaos is a wake-up call for crypto's dependence on speculation. The industry must fix its market structure, bridging onchain and offchain systems, and stop leading with “50x leverage products.” Instead, builders should focus on stablecoins, programmable money, and global capital markets that drive real-world economic growth. If more than half of onchain activity in five years isn't tied to real GDP, “we've failed.”

CIRCLE PARTNERS WITH SAFE FOR INSTITUTIONAL USDC STORAGE

Circle has named Safe, the multisig smart account platform securing over $60B in digital assets, as its “premier institutional storage solution” for USDC.

STEAK ‘N SHAKE WILL ONLY ACCEPT BITCOIN PAYMENTS

After pushback from the Bitcoin community, Steak ‘n Shake will not consider adding Ethereum payments and instead maintain opinionated around Bitcoin.

MONAD AIRDROP CLAIM LIVE

Monad's MON token is now claimable for early ecosystem users and select NFT holders.

REDDIT REMOVING NFTS

After a multi-year trial run with collectible avatars and crypto wallets associated with accounts, Reddit announced it's shutting down this initiative at the start of next year.

NEW APPLE TV REBRAND IS THE RIGHT MOVE AT JUST THE RIGHT TIME

Apple has officially dropped the “+” from Apple TV+, rebranding its streaming service simply as Apple TV. The change aligns the service with the Apple TV app, reflects how most people already refer to it, and comes just as new Apple TV 4K hardware is expected to launch—offering the perfect moment for a unified brand rollout.

VIDEO OVERVIEWS ON NOTEBOOKLM GET A MAJOR UPGRADE WITH NANO BANANA

NotebookLM has upgraded its Video Overviews feature with Gemini's Nano Banana image generation model, which automatically creates contextual illustrations in six visual styles, including Watercolor, Papercraft, and Anime. Users can now choose between two formats: "Explainer" for comprehensive summaries or "Brief" for quick highlights of their uploaded documents. The update is rolling out to Pro users this week and to all users in the coming weeks across all supported languages.

TELEGRAM ADOPTS A LIQUID GLASS-LIKE DESIGN, NO IOS 26 REQUIRED

Telegram's latest iOS update introduces its own version of Apple's Liquid Glass design, bringing transparent, refractive interface elements even to devices not running iOS 26. The update also adds group call comments and reactions, contact notes and birthday suggestions, customizable profile colors, threaded AI bot chats, and improvements to the gifts marketplace and bug reporting tools.