Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
ChatGPT Apps allow developers and product teams to embed apps directly into ChatGPT, enabling users to complete tasks like book hotels, search for properties, and shop online. With a platform of 800 million weekly active users, ChatGPT's app platform may be the next great distribution wave. This article walks readers through how to build ChatGPT apps. It includes real-world examples, code, and links to courses with more material.
The ultrawealthy buy shares of the most in-demand companies via invite-only transactions long before the shares are listed on public stock exchanges. This small, privileged group can obtain shares of companies still in their early growth stages, while the rest are left with older, slower-growing names. This dynamic is exacerbating the wealth disparity in the US. It could even become an existential threat to the US economy.
Apple was found to be in willful violation of a 2021 injunction intended to open up iOS App Store Payments in April. That contempt of court finding has now been almost entirely upheld by the Ninth Circuit Court of Appeals. Apple may still be able to charge a 'reasonable fee' for payments made outside its App Store based on actual costs to ensure user security and privacy. The ruling means that there will likely soon be rapid adoption of outside payment processors.
The Nex Playground active game system is a $249 motion-controlled game console with an $89 annual subscription marketed to families with young children.
A discussion on the problems with large language models and the workarounds people use to try to make them produce what they want.
AI accelerates the learning that makes the bet on juniors worth it.
Elon Musk will personally oversee the production lines at Samsung's Taylor facility to speed up the feedback process.
Coding agents can be used to delete complexity from codebases at a reasonable cost.
Apple locked out this developer's account after they attempted to redeem an Apple Gift Card purchased from a major brick-and-mortar store.
Sponsored
Skills support is quietly showing up in OpenAI's Codex CLI tool and ChatGPT. The skills folder can be accessed by prompting, 'Create a zip file of /home/oai/skills'. So far, the skills cover spreadsheets, docx, and PDFs. A link to a repository with a copy of the skills is available in the article.
Tinker is open to everyone now, featuring a new reasoning model, Kimi K2 Thinking, and an OpenAI API-compatible interface for seamless integration. Vision input capabilities have been added with Qwen3-VL models, allowing processing of images and text together. These updates enhance Tinker's utility in image classification, outperforming traditional models with limited labeled data.
Claude uses on-demand tools and selective retrieval for its memory system. This post explores Claude's memory system through conversations with the bot. Claude appears to be cooperative, transparent, and willing to share information about its internal structure, tools, and prompt format. However, it is worth noting that Claude can hallucinate, so some of the information may be inaccurate.
Diffusion language models generate code at a faster rate than large language models. Increased structure tends to correlate with reduced entropy, which leads to higher confidence token predictions, which directly means more tokens decoded in parallel per step. Tests suggest that it really is the structuredness of the output, not memorization, that matters.
Mintlify's AI-powered assistant helps end users get answers from docs with clear citations and useful code examples. This article walks through how the team analyzed and improved the assistant after they decided that it wasn't performing the way they wanted. The team rebuilt its feedback pipeline, moved conversation data into ClickHouse, and categorized negative interactions at scale. Its analysis surfaced that search quality was the assistant's biggest weakness, while most other responses were strong.
While large language models are useful tools, it's hard to look at the frontier models and see something approximating human-level intelligence when there are such evident cognitive gaps. These models aren't being fine-tuned to make accurate judgments about even small datasets. There needs to be more efficient ways to attach data that doesn't burn tokens while still allowing models to pass data sets through to efficient tools. This would make tooling a great deal more efficient.
Sponsored
Most companies get stuck tinkering with prompts and wonder why their agents fail to deliver dependable results. This guide from You.com [11] breaks down the evolution of agent management, revealing the five stages for building a successful AI agent and why most organizations haven't gotten there yet.
The initial version of Sora's production Android app was built in 28 days using OpenAI Codex. It took a lean engineering team and roughly 5 billion tokens to ship the project. The app has a crash-free rate of 99.9%. This article describes how OpenAI used GPT-5.1-Codex - the same version available to any developer or business - to build the app.
Claude Code uses boundary-aware queuing, where new messages are inserted at natural break points, which changes the model's course of action smoothly without stopping ongoing generation. OpenAI Codex uses post-turn queuing, where user messages wait until the current action finishes completely before they are handled. Agentic tools should implement both types of queuing and let users choose which to use. Having that option would make a difference in agentic workflows where users are running three to four agents in parallel.
The easiest way to work with tools is to ask agents to write their own tools as a skill. This leaves control of the tool largely to the user. Whenever it breaks or needs modification, the user can just ask the agent to adjust it. Dynamic tool loading with MCP will likely become a thing, but it will probably take quite a few protocol changes to bring in skill-like summaries and built-in manuals for the tools.
Google used its video generation model Veo to build a world simulator that predicts how robotics algorithms will perform in novel environments without physical testing. The system accurately ranked eight policy checkpoints and identified safety vulnerabilities—like a robot knocking over a laptop or grabbing a bottle too aggressively—through 1,600+ simulated rollouts that correlated strongly with real-world results.
OpenAI has ended a compensation policy that required employees to work at the company for at least six months before their equity vests. The change is designed to encourage new employees to take risks without fear of being let go before accessing their first chunk of equity. OpenAI had shortened its vesting period for new employees to six months from the industry standard of 12 months in April. xAI also made a similar change in the late summer.
Claude Code's capabilities have grown tremendously. This means that developers have a lot more to learn. Claude Code is currently optimizing hard for the power user while trying not to lose everyone else. While the learning curve is manageable, every new capability adds weight. The risk is that Claude Code becomes so capable that you need to learn Claude Code to use it.
The new Kimi K2 1T model (4-bit quant) runs on 2 512GB M3 Ultras with mlx-lm and mx.distributed.
The most expensive frontier approaches will likely be much cheaper within a year.
Google's AI Mode appears to retrieve content from a proprietary content store separate from the search index.
Liang Wenfeng is a founder who created his own success in his own way, a prime example of a modern educated youth changing his own destiny through entrepreneurship.
Google is integrating Gemini's advanced translation capabilities into Google Translate, including a beta for live speech-to-speech translation with headphones and new language options for practice and learning.
This post explores how to build a universal engine for any schema-based integration, without manual plumbing.
Sponsored
RAD Intel has 7-figure recurring contracts, sales set to 2x in 2025 [4], a valuation up 5000% in about 5 years, $60M+ raised from 14K+ investors, and a reserved Nasdaq ticker $RADI [4] while still private.
DTCC received an SEC no-action letter approval for a three-year pilot program starting H2 2026 to tokenize Russell 1000 stocks, major ETFs, and US Treasury securities onchain while maintaining identical rights and protections as original securities. Potential benefits include collateral mobility, 24/7 trading access, and programmable assets. The move signals growing acceptance of blockchain in traditional finance, with over $33 billion in real-world assets now onchain. DeFi protocol Ondo Finance urged the SEC to delay Nasdaq's tokenization plans until DTC settlement processes are disclosed publicly to ensure fair competition and prevent larger institutions from gaining advantages.
Do Kwon received a 15-year prison sentence for his role in Terra's $50 billion collapse, exceeding prosecutors' 12-year request and tripling the defense's five-year proposal, which the judge called "utterly unthinkable and wildly unreasonable." District Judge Paul Engelmeyer cited the "eye-popping" magnitude of the crime, which affected hundreds of victims, Kwon's flight to Serbia and Montenegro on false passports, and "despicable" conduct during UST's de-peg advising retail to stay invested while insiders exited positions, with Kwon's trollish Twitter presence particularly galling to the judge. Kwon pleaded guilty in August to conspiracy and wire fraud after extradition in Montenegro, reducing the original nine-count indictment, which carried 135 years maximum, to two counts. He must serve at least half his sentence before applying for a South Korea transfer, where additional charges await.
Bhutan has launched TER, a Solana-based gold-backed token issued through Gelephu Mindfulness City with custody by DK Bank, the kingdom's first licensed digital bank, offering international investors blockchain-based physical gold ownership with digital portability. The announcement follows Kyrgyzstan's $50 million USDKG gold-backed stablecoin launch days earlier, demonstrating a pattern of smaller nations using blockchain to fuse traditional asset reserves with regulated digital finance. Gelephu Mindfulness City operates as a special administrative region using digital assets for financial reserves and innovation as part of Bhutan's broader blockchain strategy to diversify its economy.
Coinbase has released x402 V2, an updated stablecoin-based AI agent payments protocol that features wallet-based identity, eliminating per-call repayment, automatic API discovery, dynamic payment recipients, and multi-chain/fiat support via Chain Agnostic Improvement Proposals. The modular SDK enables custom networks and schemes, moving the protocol from "what does it do" to "which services plug in next" as developers combine payments with secure wallet access. The update expands on the original open-source x402 protocol and x402 Bazaar discovery layer released earlier this year, positioning the system as future-proof infrastructure for unified payment models between agents and humans.
Fogo cancelled its $20 million token pre-sale, originally planned for December 17, instead airdropping the 2% allocation to better distribute tokens and reward early users ahead of its mainnet launch on January 13, with 38.98% of supply unlocked. The Solana Virtual Machine-based Layer 1 adjusted tokenomics to give institutional investors, Distributed Global and CMS Holdings, 8.77% while burning an additional 2% from core contributors' genesis allocation, with the team taking a snapshot of testnet Fogo Fishers and Portal Bridge users for Fogo Flames points redeemable for tokens. Fogo is the first blockchain implementing Jump Crypto's Firedancer client.
The persistent ~6% premium for USDT in India isn't a retail quirk, it's a shadow exchange rate driven by capital controls and gold smuggling economics. India legally taxes imported gold at 6%, and historically, smugglers bypassed this using the hawala system to move money offshore. Today, hawala has upgraded to USDT as people with unreported cash buy Tether at a premium in India to pay gold suppliers abroad, pushing USDT above the RBI-managed INR/USD rate. The premium tracks the gold duty almost perfectly, as when the tax was 15%, the USDT premium was ~15% and when the tax dropped to 6%, the premium fell to ~6%.
Consumers have an unfair relationship with banks. People earn almost nothing on deposits while banks take large risks, keep the upside, and rely on taxpayer bailouts for the downside. Historically, this was controlled by Glass-Steagall, which forced banks to separate boring deposit-taking from speculative investment banking, creating a safer and more socially aligned system. Deregulation in the 80s–90s removed those guardrails, letting banks earn huge profits while lobbying to restrict competition from yield-bearing stablecoins. There are two paths we may pursue: one is to allow free competition between banks and stablecoins, or consider bringing back Glass-Steagall to realign banks with public interest.
Dismissing crypto cards because they're “not the endgame” misses how infra transitions actually happen, as we won't jump from SWIFT to self-custodial ZK payments overnight, and cards exist because merchant acceptance rails haven't moved yet. Commoditizing Cards-as-a-Service is an accelerant (not a weakness) that pushes differentiation up-stack into custody, yield routing, tax/FX transparency, UX, and compliance. Most crypto cards will die, especially in bear markets, but the winners will be self-custodial, multi-stablecoin native, yield-safe, low-FX, niche-focused, and genuinely user-first.
Tether, a sponsor and partner of Juventus Football, submitted a $1.3 billion bid to acquire a majority stake in the football club, which was quickly declined.
Meme coins aren't just about gambling or “financial nihilism”, they're the first large-scale experiment in turning attention itself into a tradeable asset.
Citadel Securities sent a letter to the SEC urging it to regulate DeFi technology as if it were traditional financial intermediaries, arguing that validators, wallets, and smart contract developers should be treated like brokers.
Stripe's stablecoin volume on Polygon has grown from under $1M per month in late 2024 to peaks of $8–9M in 2025, signaling rising merchant confidence in Polygon as a payments layer.
Disney signed a three-year partnership with OpenAI and invested $1 billion, allowing Sora and ChatGPT Images to generate content featuring over 200 characters from Disney, Marvel, Pixar, and Star Wars. Users can create videos and images using iconic characters like Mickey Mouse and Darth Vader, though talent likenesses and voices are excluded from the agreement. Despite previously suing Midjourney and sending cease-and-desist letters to Character.AI over IP violations, Disney will become a major OpenAI customer to build new products.
Instagram's Edits app now includes pre-built templates, storyboards, advanced text tools, and an iPhone lock screen widget for instant camera access and quick content capture. The platform is introducing "Your Algorithm" in the US, allowing users to view and modify the topics Instagram uses to curate their Reels feed by adding or removing interests. These updates aim to streamline content creation for creators while giving regular users more control over their algorithmic feeds, with similar Explore page features planned.
Apple's iOS 26.2 adds another control to reduce Liquid Glass transparency—this time for the Lock Screen clock—continuing Apple's rollback via user-controlled settings after complaints that the new glassy UI hurt readability. The update also brings AirDrop codes, Reminders alarms, offline lyrics in Apple Music, AI features in Podcasts, a Sleep Score on Apple Watch, and critical security patches across Apple devices.
A study of 340 product designers found widespread overload and burnout driven by chaotic management, constant multitasking, unclear expectations, and poor cross-team communication. Career growth was often blocked despite designers taking responsibility for their own development. While most aren't afraid of AI and see it as a helpful companion, job searches are getting significantly longer, feedback is scarce, education feels disconnected from real work, and the profession's challenges are increasingly seen as systemic rather than individual.
Design systems often stall when perfectionism overtakes practicality, as teams polish components endlessly while product teams build their own solutions rather than wait. The most critical elements—accessibility, semantic structure, token consistency, stable APIs, and sensible defaults—can't be compromised, but everything else is a judgment call that shouldn't delay shipping. Components gain adoption by arriving when teams need them at 80% completeness, then improving through real-world use rather than theoretical perfection.
Most design tokens lack the context AI needs, leading tools like Claude to misuse primitives rather than semantic tokens. Making tokens AI-readable requires three elements: semantic naming that conveys intent rather than appearance, descriptions explaining when to use each token, and relationships showing which tokens connect or pair together. Teams can start by adding descriptions to their top 10 most-used tokens, or by creating companion files that provide usage guidelines without restructuring existing token systems.