Search the archive

LET'S TALK ABOUT LLMS

Large language models might be another vaporware hype cycle, or it might be a dot-com-style bubble, but everyone seems to agree we're in the middle of something.

ANTHROPIC LAUNCHES CLAUDE COWORK IN GENERAL AVAILABILITY

Claude Cowork is now in general availability with new organization controls designed to give admins more oversight and flexibility.

WILL I EVER OWN A ZETTAFLOP?

The biggest bottleneck to owning a zettaflop is power.

✅ Truly serverless → Run your first query in ~4 minutes

✅ Up to 95% cheaper than traditional vector/text search databases

CHATGPT $100 PRO PLAN INTRODUCED

OpenAI introduced a new $100/month ChatGPT Pro tier aimed at power users, expanding its pricing lineup between the $20 Plus and $200 Pro plans. The company confirmed the $200 plan still exists despite not being listed on the pricing page.

MAKING CLAUDE COWORK READY FOR ENTERPRISE

Claude Cowork is now enterprise-ready, enhancing work with organization controls like role-based access, group spend limits, and expanded observability features. Admins can manage adoption with detailed usage analytics and integrate tools like Zoom for seamless workflows. Companies like Zapier and Airtree have already leveraged these features for improved project management and operational efficiency.

ALIBABA CLAIMS VIRAL HAPPY HORSE AI MODEL IN LATEST BREAKTHROUGH

The Happy Horse video AI model that sent ripples across China's AI industry this week was created by Alibaba. The model hit the top spot on the text-to-video leaderboard of Artificial Analysis this week. Alibaba plans to provide API access to the model in the near future.

PERPLEXITY LAUNCHES PERSONAL FINANCE POWERED BY PLAID

Perplexity has expanded its financial services with Plaid integration, transitioning from portfolio tracking to a comprehensive finance dashboard. Users can link checking, savings, credit card, and loan accounts to analyze spending, track liabilities, and calculate net worth. Targeted at AI-savvy consumers, it combines data insights with Plaid's infrastructure, transforming from a portfolio add-on to a full personal finance hub without executing trades.

AGENTIC INFRASTRUCTURE ON VERCEL

Vercel argued that coding agents were reshaping how software gets built and deployed, with agent-initiated deployments rising sharply and accounting for more than 30% of weekly deployments. The post framed this shift as a need for new infrastructure designed for agents to deploy software, run AI systems, and increasingly operate infrastructure autonomously.

COREWEAVE TAKES AS MUCH FINANCIAL ENGINEERING AS IT DOES DATACENTER DESIGN

CoreWeave's revenue backlog is up to $87.8 billion, with Meta represents 40.1% of that backlog. OpenAI's deals represent 25.5% of extended backlog. CoreWeave only has two dozen named customers, but it likely has more. The company achieved $5.13 billion in sales in 2025, up 2.7x from 2024, and posted a net loss of $1.17 billion. It recently held a private offering of $1.75 billion in notes to cover its data center expansion costs.

RESEARCH-DRIVEN AGENTS: WHAT HAPPENS WHEN YOUR AGENT READS BEFORE IT CODES

Coding agents generate better optimizations when they read papers and study competing projects before touching code. Coding agents working from code alone generate shallow hypotheses. Adding a research phase improves results.

Try IronClaw Free [13]→

ANTHROPIC LAUNCHES ADVISOR TOOL FOR CLAUDE PLATFORM API USERS

The new advisor tool on Anthropic's Claude Platform API provides developers with the ability to use Opus as an advisor alongside Sonnet or Haiku as executors. This allows agents to access advanced reasoning capabilities while maintaining operational costs at the level of the more efficient executor models. The advisor tool is now publicly available through a configuration in the Messages API request.

MULTIMODAL EMBEDDING & RERANKER MODELS WITH SENTENCE TRANSFORMERS

The v5.4 update of Sentence Transformers introduces multimodal embedding and reranker models that enable encoding and comparing texts, images, audio, and videos within a shared embedding space for tasks like cross-modal search and retrieval-augmented generation. These advanced models map different modalities into a consistent embedding space and assess relevance across mixed-modality pairs, though lower cross-modal similarity scores may occur due to modality clustering. Notably, multimodal rerankers provide high-quality scoring by evaluating individual pairs for more accurate retrieval results.

PROCESS DRIVEN IMAGE GENERATION

Meta has introduced a multi-step framework for image synthesis that alternates between textual reasoning and visual updates. This enables models to iteratively plan, draft, reflect, and refine outputs.

EFFICIENT ROLLOUT SCALING FOR DIFFUSION RL

NVIDIA's Sol-RL introduces a two-stage framework that separates exploration and training, using FP4 rollouts to generate large candidate sets and BF16 for selective policy updates. This approach reduced compute costs while improving alignment and accelerating convergence in diffusion model post-training.

DARIO SAYS CONTINUAL LEARNING IS SOLVED. IS IT?

Continual learning research is aimed at breaking past the feasible horizon of current techniques. If models can't learn new things while performing tasks, they will struggle when the task horizon grows very long. Context management and other engineering improvements can push task horizons to weeks, or even months, but very long-horizon tasks remain out of reach. Human-level continual learning may be 'solved', but the revolution is not yet complete.

INTRODUCING KELLYBENCH

KellyBench evaluates sequential decision-making in sports betting by simulating the 2023-24 English Premier League season. Models like Claude Opus 4.6 and GPT-5.4 struggled, with none achieving positive returns, highlighting their limitations in adapting to long-term strategies. This underscores a need for complex environments enabling agents to learn from experience under uncertainty.

FULLY REMOTE)

TLDR's Applied AI team is tasked with making every process at TLDR legible to code, runnable by anyone, and composable into larger workflows. Join a small, fast moving team using the latest AI tools with an unlimited token budget. Learn more [21].

THE GEMINI APP CAN NOW GENERATE INTERACTIVE SIMULATIONS AND MODELS

The Gemini app now supports generating interactive visualizations, like 3D models and simulations, directly in chat.

GOOGLE'S PAPERORCHESTRA AI CONVERTS LAB NOTES INTO PUBLICATION-READY RESEARCH PAPERS

Google Cloud AI's PaperOrchestra converts disorganized lab notes into submission-ready academic papers using five specialized agents.

OPENAI PROJECTS STEEP ADVERTISING GROWTH, TARGETING $100 BILLION BY 2030

OpenAI expects its advertising business to generate $2.5 billion in revenue this year.

PICSART LAUNCHES "EARN WITH PICSART", A MONETISATION PROGRAMME

Earn with Picsart is a monetization program open to all 130+ million users that pays creators based on content performance rather than follower count. Creators browse campaign prompts, create content using Picsart's tools, post to social media platforms, and earn money based on engagement metrics like views and shares. This launch, along with their AI agent marketplace, signals Picsart's transition from a design tool to a full creator economy platform.

HIGGSFIELD BETS ON AI-GENERATED SERIES WITH VIEWER-PICKED ORIGINALS PUSH

Higgsfield Original Series is a streaming platform for AI-generated shows. It is debuting with the pilot episode, "Arena Zero," and several other pilot concepts. The platform uses a crowdsourced model in which viewers vote on which pilots become full series, bypassing traditional studio decision-making. This approach aims to reduce production risks by testing audience demand before committing to full development, while also opening submissions for creators to pitch their own AI-generated pilots.

POKE MAKES USING AI AGENTS AS EASY AS SENDING A TEXT

Poke is a new AI assistant that works through text messaging apps like iMessage and SMS, letting users automate everyday tasks such as scheduling, health tracking, and controlling smart devices without installing an app. It stands out by being easy to use and by choosing the best AI model for each task, while also offering shareable automation “recipes.” Backed by major investors and growing quickly, the company is focused on reaching a wide audience with flexible pricing rather than making profits right away.

THE INVISIBLE LAYER OF UX MOST DESIGNERS IGNORE

Most designers focus on visual polish but overlook how their interfaces sound when announced by screen readers. Common issues like vague link labels, missing heading structure, and icon buttons without accessible names create frustrating experiences for assistive technology users. Thinking in terms of each element's role, name, and state—and testing with a screen reader—can close the gap without requiring code expertise.

EX-APPLE ENGINEERS CREATE AN AI BUTTON THAT LOOKS LIKE AN IPOD SHUFFLE – BUT CAN'T EXPLAIN WHY

Dedicated AI hardware devices are seen as unnecessary because they replicate functionality that smartphones already handle, much like trying to reinvent the iPod in the age of the iPhone. Although earlier AI wearables failed due to poor performance and unclear purpose, even newer concepts like the “Button”—which improves privacy with push-to-activate AI—still struggle to justify why they shouldn't simply exist as smartphone apps.

'HUMAN ARTISTS CHARGE TOO MUCH, AI CAN DO JUST AS GOOD A JOB': AI LOGO CONTROVERSY DIVIDES CREATIVES

A restaurant in Santa Cruz drew backlash after using an AI-generated logo, leading to negative reviews and eventually a redesign, and igniting a broader debate about the ethics and value of AI in design. Opinions were split: some argued that small businesses shouldn't be expected to hire designers and that AI offers an accessible, cost-effective alternative, while others felt the logo's quality was poor and emphasized the importance of supporting human creatives and craftsmanship. The controversy highlights ongoing tensions between affordability, creative labor, and the growing role of AI in everyday design work.

ONLINE GRAPHIC DESIGN TOOL FOR E-COMMERCE BUSINESS OWNERS

Glorify is an online graphic design tool specifically created for e-commerce business owners to create marketing visuals that convert browsers into buyers.

BUILD YOUR OWN COMPONENT LIBRARY WITH AI

Bearnie is a CLI tool for building custom component libraries with Astro and Tailwind CSS.

UNICODE WIREFRAME DESIGN TOOL

Wiretext is a Unicode-based wireframe design tool that allows users to draw wireframes using text characters and symbols.

HOW DESIGNERS SURVIVE

Designers are shifting from being specialized "violin players" to generalist "conductors" who orchestrate AI tools and cross-functional teams. This change creates more opportunities for designers since good design is becoming a differentiator, with teams needing more design expertise than other functions. Job candidates now need experience orchestrating projects and leading teams with AI tools, rather than just being skilled individual contributors.

ETHICAL UX PATTERNS: BUILDING TRUST WITHOUT MANIPULATION

Mental health apps often use engagement tactics from the attention economy, but these manipulative design patterns can exploit vulnerable users and undermine the trust essential for therapeutic effectiveness. Standard engagement techniques like streaks and notifications that work for general apps can cause harm when applied to users managing depression, anxiety, or PTSD. The challenge is designing for consistent usage while building genuine therapeutic relationships rather than addictive behaviors that contradict mental health goals.

A NEW LEAF: AN EXCLUSIVE LOOK INSIDE THE NEW YORK TIMES MAGAZINE REDESIGN

The redesign of The New York Times Magazine responds to a media landscape where stories span print, digital, audio, and video, requiring a more flexible system that works across formats while maintaining a distinct editorial identity. It introduces updated typography, layouts, and repeatable story formats, strengthens collaboration between design and editorial, and aims to preserve what makes magazine journalism unique—especially the immersive, visually rich experience of print—while adapting content to feel meaningful in every medium.

BLENDING ABSTRACT GEOMETRY WITH BEAUTIFUL COLORS: ORGANIC 3D ARTWORKS OF ROMAN BRATSCHI

Roman Bratschi is a 3D freelance designer and animation director who has collaborated with major brands such as Apple, Spotify, and Samsung throughout his 15-year career.

ADOBE FIREFLY'S CUSTOM AI MODELS "PRESERVE THE UNIQUE SOUL OF YOUR WORK"

Adobe Firefly's custom AI models, now in public beta, allow users to train AI on their own images to maintain consistent brand aesthetics and creative styles, preserving the unique "soul" of creative work while enabling faster iteration.

THE JOY AND POWER OF RISOGRAPH: RISOTTO SHOWCASES 400 PRINTED POSTCARDS FROM ARTISTS ACROSS THE GLOBE

Snail mail is experiencing a revival as a tactile, slower alternative to digital overload.

MOVING FROM WORDPRESS TO JEKYLL (AND STATIC SITE GENERATORS IN GENERAL)

DemandSphere migrated its website from WordPress to the static site generator Jekyll to resolve issues with speed, ease of modification, and general development limitations. This extensive effort was significantly supported by AI-assisted development using Claude Code to build custom tools for comprehensive auditing, resulting in improved performance, a robust SEO architecture, and faster, higher-quality content execution.

HOW SPOTIFY SHIPS TO 675 MILLION USERS EVERY WEEK WITHOUT BREAKING THINGS

Spotify maintains a 95% success rate for weekly releases to 675 million users by using a "rings of exposure" model and feature flags that decouple code deployment from feature activation. By integrating a centralized management dashboard with an automated "Robot" service to handle predictable transitions, the company has a streamlined release pipeline while allowing human engineers to focus on high-stakes architectural judgment.

I STILL PREFER MCP OVER SKILLS

MCP is a better standard for AI service integration through its API abstraction and authentication. Shifting "Skills" toward a focus on pure knowledge and workflows while utilizing MCP for connectivity creates a more scalable and less fragmented ecosystem for LLMs.

CLEAN CODE IN THE AGE OF CODING AGENTS

Maintaining a clean, modular codebase is necessary in the age of AI because organized structures reduce token costs and prevent cognitive overload for coding agents. By prioritizing readability and simplicity, developers make sure that LLMs can implement features more accurately without getting lost in a mess of unnecessary context.

BUILD WITH INTENT

CLAUDE MANAGED AGENTS OVERVIEW

Claude Managed Agents offer a pre-built, configurable agent harness running in managed infrastructure, providing a fully managed environment for Claude to operate autonomously. This service removes the need to build custom agent loops, allowing Claude to securely read files, run commands, browse the web, and execute code with built-in performance optimizations.

A BACKEND FOR AI-CODED APPS

Instant is an open-source backend that simplifies full-stack development for AI agents by providing real-time sync, authentication, and file storage through a reactive multi-tenant database. By abstracting infrastructure hurdles, it allows devs to push high-performance, never-frozen applications to production with reduced code complexity.

WHAT GAME ENGINES KNOW ABOUT DATA THAT DATABASES FORGOT

Game servers face a dilemma, needing both the raw performance of game engines and the transactional safety and durability of traditional databases. Typhon is an embedded .NET database engine that solves this by speaking the native language of game servers and synthesizing best practices from both fields. It uses game engine principles like cache locality and zero-copy access, while incorporating database features like ACID transactions with per-component MVCC, advanced indexing, and configurable durability.

REALLOCATING $100/MONTH CLAUDE CODE SPEND TO ZED AND OPENROUTER

Frustrated by usage limits on a $100/month Claude Code subscription, this author came up with a new strategy that involves reallocating funds to $10/month for the Zed editor and $90 for OpenRouter API credits. The approach provides access to more AI models and a more adaptable coding workflow using non-expiring credits.