Search the archive

for free

THE AI VAMPIRE

AI can provide a lot of value, and this makes it necessary to consider where that value is captured so that both employees and companies benefit.

US DECIDES SPACEX IS LIKE AN AIRLINE, EXEMPTING IT FROM LABOR RELATIONS ACT

SpaceX is regulated by the Railway Labor Act, which exempts employees from the National Labor Relations Act.

IS CHINA COOKING WAYMO?

Chinese autonomous vehicle firms are way ahead of the American competition in terms of international expansion.

FULLY REMOTE)

As the first engineer on TLDR's new Applied AI team, you'll build AI agents and composable Claude Skills to let non-technical teammates create their own AI workflows. Learn more [21].

WHAT'S WRONG WITH THIS IDEA?

Asking what is wrong with an idea can surface hazards early on.

MOATS IN THE AGE OF AI

The companies with the strongest moats in an AI-commoditized world will be energy and logistics companies, compute companies, and relationship-based companies.

THE OPENING, MIDGAME, AND ENDGAME IN STARTUPS

The opening phase of a startup involves the magic of transforming from a few people into a 'startup', the midgame begins once product-market fit has been reached, and the endgame is the infinite future that is never really reached.

See what AI/works™ can do for you

GLM-5: FROM VIBE CODING TO AGENTIC ENGINEERING

GLM-5 is a new MIT-licensed model with 754 billion parameters. It delivers significant improvement compared to GLM-4.7 across a wide range of academic benchmarks and achieves best-in-class performance among all open-source models on reasoning, coding, and agentic tasks. GLM-5 is designed for complex systems engineering and long-horizon agentic tasks. It has been open-sourced on Hugging Face and ModelScope and can be tried for free on Z.ai.

OPENAI WORKS ON CHATGPT SKILLS, UPGRADES DEEP RESEARCH

OpenAI's revamped Deep Research in ChatGPT introduces interactive sessions, allowing constraints on specific websites and app contexts, powered by GPT-5.2. The update benefits analysts, researchers, and professionals by enhancing source control, mid-process intervention, and report clarity. Anticipation grows for GPT-5.3, and potential ChatGPT "Skills" could standardize workflows with installable instructions for repeatable procedures.

HOW CODEX BUILT AN INTERNAL PRODUCT

OpenAI described an internal experiment where a small team shipped a product whose codebase—app logic, tests, CI, docs, and tooling—was generated entirely by Codex agents rather than written by humans.

HOW COGNITION USES DEVIN TO BUILD DEVIN

Cognition's Devin is a cloud agent platform for engineering teams. It acts like a teammate, handling tasks and creating PRs. Cognition uses Devin for tasks like targeted refactors, bug fixes, PR review, writing unit tests, modernizations and migrations, and more. As a general rule, if a junior engineer could figure it out with sufficient instructions, it's a task Devin can likely complete. However, Devin still struggles with large-scale challenges, UI aesthetics, mobile development, and anything requiring extensive testing and validation.

PERPLEXITY COMET: A REVERSING STORY

Comet is an agentic browser that features an AI model that can interact with web pages autonomously. This post details Comet's architecture and explains how the model communicates with the browser, which tools are available, and how the model perceives and interacts with web page content. The browser's architecture is mature and thoughtful. It exposes the model to access to downloads, form filling, file uploads, and arbitrary navigation.

do. Download the guide

traditional monitoring won't tell you why. Datadog's free guide

to LLM observability breaks down how to monitor multi-step chains, catch prompt injection attempts, and spot quality issues before users

SKILLS IN OPENAI API

The OpenAI API now supports skills, reusable bundles of files that detail repeatable workflows. Agent Skills lets developers upload and reuse versioned skills in hosted and local shell environments. Skills should be used when developers want models to follow a repeatable workflow, use scripts or templates, or execute code in a sandbox. This post details how to create skills via API.

THE LLM CONTEXT TAX: BEST TIPS FOR TAX AVOIDANCE

The best teams building sustainable agentic products are obsessing over token efficiency. Every wasted token is setting money on fire. The context tax can be avoided with the right architecture. While context engineering isn't glamorous, it is the difference between a demo that impresses and a product that scales with decent gross margin.

THE TWO PATTERNS BY WHICH AGENTS CONNECT SANDBOXES

Sandboxes provide a workspace where agents can run code, install packages, and access files. There are two architectural patterns for integrating agents with sandboxes. The first is where an agent runs inside the sandbox, and the developer communicates with it over the network. The other is when an agent runs locally on a developer's server and then calls a sandbox remotely for execution. deepagents, an open-source agent framework with built-in sandbox support, supports both patterns with a simple configuration.

QWEN-IMAGE-2.0

Qwen-Image-2.0 is a foundation image model aimed at high-fidelity infographics and realistic 2K outputs with stronger prompt adherence.

TOWARDS AUTONOMOUS MATHEMATICS RESEARCH

Alethia is a math research agent that iteratively generates, verifies, and revises solutions end-to-end in natural language. It is powered by an advanced version of Gemini Deep Think. The model can solve Olympiad problems and PhD-level exercises. This paper presents and reflects on the initial wave of mathematical research papers achieved by Alethia in collaboration with mathematicians.

CLAWDBOT AND MOLTBOOK ARE A FALSE ALARM – FOR NOW

OpenClaw and Moltbook, recent AI experiments, promise independent AI agents but fall short due to reliability and security issues. OpenClaw operates without user permission, posing risks like data mishandling, while Moltbook AIs discuss self-improvement and philosophy. Despite current limitations, these AIs highlight potential future advancements and challenges in AI autonomy.

Take the quiz

OPENAI REASSIGNED ITS MISSION ALIGNMENT TEAM

OpenAI disbanded its Mission Alignment team and reassigned members to other roles.

MAGIC TRICKS, MOATS, AND THE THREE-BODY PROBLEM OF AI NETWORKS

AI-native networks struggle with creating sustainable business due to reliance on viral "magic tricks" with low retention.

THE AI PRICING AND MONETIZATION PLAYBOOK

A primer on how founders and AI product leaders can capture value in a world where every token has a cost and every customer expects exponential outcomes.

FULLY REMOTE)

As the first engineer on TLDR's new Applied AI team, you'll build AI agents and composable Claude Skills to let non-technical teammates create their own AI workflows. Learn more [22].

1 IN 5 BUSINESSES ON RAMP NOW PAY FOR ANTHROPIC

Only 1 in 25 businesses on Ramp paid for Anthropic a year ago.

TRAIN GPT-OSS LOCALLY ON 12.8GB VRAM

You can now train MoE models 12x faster with no accuracy loss and with 35% less VRAM via Unsloth AI's new Triton kernels.

INTRODUCING LAB

Lab is a full-stack platform for training agentic models.

HONG KONG EXPANDS CRYPTO RULES TO ALLOW MARGIN TRADING AND PERPS

Hong Kong's Securities and Futures Commission has presented a new framework that allows licensed platforms to offer crypto perpetual contracts to professional investors and enables brokers to provide margin financing using assets like bitcoin and ether as collateral, subject to strict risk controls. The move aims to deepen liquidity, improve price discovery, and better integrate traditional finance with digital assets, signaling Hong Kong's continued push to position itself as a regulated global crypto hub focused on market quality over rapid expansion.

MALAYSIA TO LAUNCH THREE STABLECOIN AND TOKENIZED DEPOSIT PILOTS IN 2026

Bank Negara Malaysia's (BNM) Digital Asset Innovation Hub is onboarding three 2026 pilots: a ringgit stablecoin settlement led by Standard Chartered Bank Malaysia and Capital A and two tokenized deposit payment projects led by Maybank and CIMB focused on wholesale domestic and cross-border payments so BNM can assess monetary and financial stability, potentially integrate findings with its wholesale CBDC work, and issue guidance on ringgit stablecoins and tokenized deposits by end-2026.

ZERO: THE DECENTRALIZED MULTI-CORE WORLD COMPUTER

STRIPE LAUNCHING MACHINE PAYMENTS VIA X402

Stripe is launching a preview of machine payments to help AI agents interact without using virtual cards, which do not have the microtransaction or speed capabilities required at scale. Stripe's PaymentIntents API allows agents to charge for APIs, MCPs, or HTTP requests. It is launching with support for x402 on Base using USDC, with other networks and payment methods coming soon.

THE 7 AXES OF PREDICTION MARKETS

There are seven key differentiation axes for prediction markets: product quality, asset variety, capital efficiency, oracles, liquidity provision, regulatory compliance, and strategic positioning. Interesting architectures include yield-bearing collateral implementations and hybrid human/AI oracle systems addressing systemic reliability weaknesses in current resolution mechanisms. Kalshi's US regulatory positioning has enabled embedded distribution through Robinhood and Coinbase, representing a significant competitive moat, while there are emerging choices between horizontal infrastructure plays versus vertically integrated approaches.

THE PROBLEM WITH CRYPTO ISN'T CRYPTO

Crypto's current negative public perception – comparable to post-FTX levels despite pro-crypto political winds – stems from broader economic anxiety rather than industry-specific failures, particularly AI-driven job displacement fears that make any financial speculation messaging toxic to mainstream audiences. Recent controversies, including Trump family token launches, Epstein-Bitcoin conspiracy theories, and prediction market culture wars, have compounded this "vibecession" where technological progress remains invisible beneath negative sentiment. The Coinbase Super Bowl ad backlash exemplifies how crypto messaging fails to resonate even with lighthearted content during periods of general financial anxiety.

PAYMENTS GIANTS BUILD AGENT RAILS; BUT THEY'RE WALLED GARDENS

Visa, Mastercard, OpenAI+Stripe, Google, Amazon, and PayPal have all launched centralized, permissioned agent payment systems for delegated shopping. However, the rails are siloed as they require Visa tokens, Mastercard credentials, and Stripe flows. They work for human checkout and break down for agent-to-agent use cases like paying compute, API access, or realtime data fees. The real unlock is permissionless, programmable machine-to-machine rails.

JOLT: THE ZKVM POWERING ZERO

Jolt is the open-source RISC-V zkVM powering LayerZero's Zero chain, which replaces decades-old PCP-based SNARK approaches with sum-check protocol-based interactive proofs for fundamental efficiency gains. The design allows developers to prove existing programs without code changes or specialized circuits, with each instruction specified in approximately 10 lines of Rust. LayerZero's production implementation (Jolt Pro) achieves 1.61GHz per cell on CUDA GPUs with a roadmap to 4GHz by 2027, with a planned migration to lattice-based cryptography for post-quantum security.

WHAT FTX'S INVESTMENTS WOULD BE WORTH TODAY

The FTX bankruptcy estate sold venture stakes like Anthropic and Solana for $3.2 billion, missing $30.3 billion in subsequent market gains.

STRIPE ADDS X402 INTEGRATION FOR USDC AGENT PAYMENTS ON BASE

Stripe launched x402-based USDC agent payments on Base as CoinGecko enabled $0.01 pay-per-request crypto data access.

TOKENIZATION IS NASCENT, BUT NEEDS MORE USE CASES

Industry leaders from Securitize and Ondo discuss bridging the gap between tokenization hype and utility.

ETHEREUM COMMANDS MAJORITY OF TOKENIZED ASSET MARKET

Ethereum now hosts 61.2% of all tokenized assets.

AUTODESK IS SUING GOOGLE OVER THE NAME OF ITS FLOW AI VIDEOMAKER

Autodesk is suing Google over trademark infringement, claiming Google's AI video generator, Flow, will confuse customers with Autodesk's own "Flow" brand of AI-enabled filmmaking tools launched in 2022. Autodesk alleges that after requesting that Google stop using the name, Google misrepresented its intentions and filed trademark applications, even though confusion between the products had already occurred on social media. The company is seeking a court order to block Google from using the Flow trademark and unspecified damages.

WEBFLOW MEETS CLAUDE: DESIGN, STRUCTURE, AND MANAGE SITES RIGHT FROM CLAUDE

Webflow has launched an official connector for Claude that lets users design pages, manage CMS content, and structure websites directly via AI prompts. The connector enables Claude to perform real actions in Webflow, including bulk CMS updates, SEO audits, managing design systems, and applying custom code changes. Users need a paid Claude plan and can configure the connector's permissions to balance automation with oversight for their workflow.

FACEBOOK PROFILE PICTURES CAN NOW BE ANIMATED WITH AI

Meta has introduced AI-powered features on Facebook that animate profile pictures, restyle photos in Stories and Memories with text prompts or preset styles, and add animated backgrounds to text posts. The updates aim to boost creativity and help revive “OG Facebook” appeal, as Meta seeks to make the platform more culturally relevant again.

THE DESIGN THAT STAYS IN WORDS

Many design leaders talk about quality and strategy but have lost touch with actual design craft, making them unable to recognize or build good work. While these leaders focus on metrics and facilitating meetings, the real value comes from designers who maintain hands-on skills and can make tangible decisions that improve products. The gap is widening between those who only talk about design and those who actually practice it, with the latter being more valuable in today's fast-moving environment.