Search the archive

Read the guide

see coming. Datadog's LLM Observability Best Practices Guide

walks through how to trace, evaluate, and secure every step of the chain - so you can improve output quality and reduce security risk.

58% OF PRS IN OUR LARGEST MONOREPO MERGE WITHOUT HUMAN REVIEW

One of Vercel's oldest and largest Next.js apps is a monorepo that contains multiple critical properties. The repository sees over 400 pull requests per week on average. Until recently, this required human approval before merging, but now, an agent reviews and merges 58% of those pull requests without a human reviewer. This has dropped the average merge time by 62%.

HOW META USED AI TO MAP TRIBAL KNOWLEDGE IN LARGE-SCALE DATA PIPELINES

Meta's AI agents weren't making useful edits quickly enough when pointed at one of the company's large-scale data processing pipelines. The company fixed this by building a pre-compute engine consisting of a swarm of over 50 specialized AI agents that systematically read every file to produce context files that encode the tribal knowledge that previously lived only in engineers' heads. As a result, its AI agents now have structured navigation guides for 100% of the company's code modules. The system works with most leading models because the knowledge layer is model-agnostic.

CONFESSIONS OF A MILLENNIAL IN TECH

Millennials may be the last generation that builds their careers around software as a medium. They drove the digital transformation, but unfortunately, they are the sandwich generation that has to explain software to their parents and AI to their kids. It is unknown what will happen to the people whose edge was being good at creating software when software started getting built, configured, and operated by AI. However, it is clear that the advantage is shifting.

SOCIAL MEDIA HAS BECOME A FREAK SHOW

Today's social media ecosystem is particularly unhealthy. It is small and continuing to get smaller. When there's a lack of competition in an isolated environment, strange things start to happen. All sorts of things that might not be survivable in a more competitive environment can actually become fitness advantages in these scenarios.

A CRYPTOGRAPHY ENGINEER'S PERSPECTIVE ON QUANTUM COMPUTING TIMELINES

Current predictions about quantum computing timelines could turn out to be wrong, but the risk that they could be right is unacceptable, and people need to start acting now.

AMAZON AND US POSTAL SERVICE REACH DELIVERY DEAL

GOOGLE AI EDGE GALLERY

Google AI Edge Gallery is Google's official app for running Gemma 4 models directly on an iPhone.

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 [24].

AGENTIC CODING AND MICROSERVICES

AI enables developers to work more monolithically.

WHAT NEXT FOR THE COMPUTE CRUNCH?

The next 18 to 24 months are going to be defined by compute shortages.

TEARDOWN OF UNRELEASED LG ROLLABLE SHOWS WHY ROLLABLE PHONES AREN'T A THING

The internal complexity of the Rollable made it extremely expensive to manufacture, and durability was a big concern.

Add the latest episode to your Spotify / Apple / Youtube queue

Listen to learn

* Why product-market fit is incomplete without pricing-product-market fit * How usage-based models demand a completely different operating model across product, sales, and finance * The pricing innovations brands like HUGGINGFACE and NVIDIA are adopting

OPENAI TESTS NEXT-GEN IMAGE V2 MODEL ON CHATGPT AND LM ARENA

OpenAI is testing its next-gen Image V2 model on ChatGPT and LM Arena in three variants. Early tests show improvements in UI design rendering, prompt adherence, and compositional understanding. The outcome of these tests could bolster OpenAI's competitive position against Google's offerings.

GOOGLE TESTS JULES V2 AGENT CAPABLE OF TAKING BIGGER TASKS

Google is developing the Jules V2 ("Jitro") coding agent, designed to autonomously manage high-level development goals rather than specific tasks. Launching via waitlist, it aims to redefine AI software development by shifting focus from task-based commands to KPI-driven outcomes. The approach might benefit teams handling large codebases, but it faces challenges of unpredictable changes and trust issues.

ANTHROPIC EXPANDS PARTNERSHIP WITH GOOGLE AND BROADCOM FOR MULTIPLE GIGAWATTS OF NEXT-GENERATION COMPUTE

Anthropic has signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity expected to come online starting in 2027. The capacity is necessary to serve the exponential growth in Anthropic's customer base. It will also enable Claude to define the frontier of AI development. The vast majority of the new compute will be sited in the US.

SOME OF META'S NEW AI MODELS WILL EVENTUALLY BE OPEN-SOURCE

Meta is close to releasing its new AI models. Some of the models will eventually be released with an open-source license. Meta plans to focus on consumers rather than the enterprise market. It previously embraced open-source AI with its Llama models, and CEO Mark Zuckerberg has written a manifesto declaring open-source AI as the past for ared. The company will be pursuing a hybrid strategy of proprietary and open-source models going forward.

AIS CAN NOW OFTEN DO MASSIVE EASY-TO-VERIFY SWE TASKS AND I'VE UPDATED TOWARDS SHORTER TIMELINES

There is now an almost 2x higher probability of full AI R&D automation by the end of 2028. Strong performance in easy-and-cheap-to-verify SWE tasks that don't require new ideas will allow AI to substantially speed up AI R&D. A lot of the work in AI R&D involves these types of tasks, and AI can complete them significantly cheaper and faster than humans.

ANATOMY OF MERCOR'S DATA BREACH

Mercor is an AI-powered talent marketplace that connects software engineers, AI data labelers, and knowledge workers with companies seeking contract labor. The company claims it was recently breached due to a supply-chain attack on a popular open source package. However, there are several reasons to doubt that claim. Saying that it was affected by the supply-chain attack may be an attempt to deflect from deeper, more embarrassing security failures.

OPENAI'S $122B “VC ROUND” IS VENDOR DEALS, CONTINGENT CAPITAL, AND A GUARANTEED RETURN IT ARGUABLY CAN'T AFFORD

OpenAI closed a $122 billion investment round at an $852 billion post-money valuation at the end of last month. Amazon, Nvidia, and SoftBank accounted for $110 billion of that investment, while the remaining $12 billion came from a broader pool of institutional and individual investors. This article breaks down the numbers of the deal. Completely separate from the round, OpenAI is negotiating a joint venture with several private equity firms valued at roughly $10 billion pre-money with a guaranteed minimum return of 17.5%, plus early access to OpenAI models before general availability.

THE ANATOMY OF AN AGENT HARNESS

This post breaks down the agent harness, the full software infrastructure that transforms an LLM into a goal-directed, tool-using agent. Production harnesses use layered memory, verification loops, subagent orchestration, and structured tool execution to prevent context loss, errors, and silent failures. Frameworks like Claude Code, OpenAI Agents SDK, and LangGraph show that harness design alone can shift agent performance by 20+ ranks, making it the critical engineering layer around any LLM.

TURNING THE WEB INTO A FILESYSTEM

Code hallucinations are mostly a data problem, caused by outdated or fragmented training data. Nia's approach mounts documentation sites as virtual filesystems, letting agents grep, cat, and tree through docs in real time, so they code against current APIs rather than stale embeddings. This client-side, in-memory shell works with Claude Code, Codex, Copilot, Gemini, and OpenCode, providing a universal, low-latency interface for agents to browse structured information without complex tool schemas or container overhead.

GITNEXUS

GitNexus indexes entire codebases into knowledge graphs, making AI agents like Cursor and Claude Code aware of every relationship, from dependencies to call chains. It offers a CLI for deep integration with editors and a web UI for visual exploration, both providing comprehensive context and reducing the likelihood of blind edits and breaking changes.

WHAT TO KNOW ABOUT OPENAI'S IDEAS FOR A WORLD WITH ‘SUPERINTELLIGENCE'

OpenAI recently released policy proposals for a world with AI that far surpasses human capabilities. The proposals suggest creating an AI-centric tax system, a public investment fund focused on AI that could regularly distribute returns to Americans, and more. OpenAI has embraced President Trump's AI strategy, including his approach of limited guardrails for the industry, but it still wants to cooperate with the rest of the world in identifying dangerous AI systems and containing them. It supports giving the government a bigger role in evaluating models.

AI IS BECOMING AN OPERATING SYSTEM LAYER

AI must become an operating system layer across entire companies, reshaping product development, service delivery, and internal operations. This approach involves integrating AI into core functions rather than treating it as a mere feature. Companies not adopting this model risk inefficiency and losing competitive edge, making AI integration crucial for investment appeal.

Some AI tools remember your conversations. Littlebird

understands what you've been working on. It observes your screen and meetings, building a private memory that grows with you. Try it free.

OPENAI SAFETY FELLOWSHIP PROGRAM

OpenAI announced a new fellowship running from September 2026 to February 2027 that will support external researchers working on AI safety, alignment, and related technical challenges with mentorship, compute, and stipends.

OFFLINE AI DICTATION APP FROM GOOGLE

Google released an iOS app using on-device ASR models for dictation that offers live transcription, filler-word removal, and optional cloud-based editing via Gemini models.

OPENAI ASKS CALIFORNIA, DELAWARE TO INVESTIGATE MUSK'S 'ANTI-COMPETITIVE BEHAVIOR' AHEAD OF APRIL TRIAL

OpenAI urges California and Delaware to investigate Elon Musk's alleged anti-competitive behavior ahead of an April trial.

THE TERM “AGI” IS ALMOST USELESS AT THIS POINT

The gaps between the different conceptions of the term 'AGI' are starting to damage our ability to think together about how to navigate the transition to a world with extremely advanced AI.

HIGGSFIELD BRINGS ART-DIRECTED QUALITY TO AI IMAGE GENERATION

Higgsfield's Soul 2 model brings art-directed AI image generation to professional creators, eliminating the synthetic look of typical AI outputs, powered by a team with luxury fashion and brand communications backgrounds. Built on human feedback from art directors and photographers, the model leverages fashion history and cultural context to accurately represent diverse aesthetics, skin tones, and hair textures. Soul ID extends this further, letting users train the model on as few as 20 photos for campaign-ready personalization, while integrated tools like OpenAI's Sora and Google's Veo keep it a one-stop production workflow.

YOUTUBE TESTS AI SUMMARIES ON HOME SCREEN

YouTube is testing an Android home feed layout that replaces or hides video titles with collapsible AI-generated summaries, aiming to speed up content understanding but significantly changing how users choose videos. Users and creators are concerned this adds friction, reduces accuracy and accessibility, and could hurt key metrics like click-through rate and watch time—so unless implemented carefully (keeping titles visible, labeling AI, and giving control), it risks degrading the core browsing experience.

MICROSOFT LAUNCHES NEW HIGH-SPEED VOICE AND IMAGE MODELS

Microsoft launched three AI models — MAI-Image-2, MAI-Transcribe-1, and MAI-Voice-1 — optimized for image generation, speech transcription, and synthetic voice, respectively. MAI-Image-2 produces images up to 1,024×1,024 pixels and runs twice as fast as its predecessor, while MAI-Transcribe-1 transcribes speech 2.5 times faster and achieved a 3.9% word error rate across 25 languages. All three models are available via Microsoft Foundry and are being rolled out across products such as Bing, PowerPoint, and Copilot Audio Expressions.

INSIDE ANTHROPIC'S BIGGEST DESIGN CHOICES

Anthropic's head of design, Joel Lewenstein, describes Claude's pushback behavior and distinct personality as entirely deliberate, aiming to create a collaborative "sparring partner" rather than a passive tool. Practical constraints like cost and response time explain why Claude doesn't automatically self-verify its outputs. Despite fears that AI will flatten design roles, Lewenstein says he is expanding his design team, with every product group requesting more human designers.

AGENTS, MEET THE FIGMA CANVAS

Figma is integrating AI agents with its design canvas to combine the power of code and visual design tools. Claude Code to Figma represents one approach to giving designers and developers more flexible building options. This integration aims to support the future of design by bridging the gap between coding and canvas-based creation.

DO LESS WITH AI

The rapid growth of AI tools is creating “productivity paralysis,” where trying to explore every new platform, automate everything, and stay competitive leads to scattered focus, unfinished projects, and a false sense of progress. Tools can quickly generate outputs that look nearly complete, but the final stretch still requires significant effort, clarity, and human judgment—something often overlooked. The takeaway is to do less, but more intentionally: focus on finishing meaningful work, define clear goals and outcomes before using AI, and treat these tools as support rather than distractions. Real productivity comes from consistent execution and prioritization, not from chasing every new capability.

GOOGLE ADS REMOTE MCP SERVER

Launch and manage Google Ads straight from your MCP client (e.g., Claude) with a hosted, remote MCP server.

TURN VIDEOS INTO VIRAL SHORTS

Create TikToks, Reels, and Shorts from your long videos in just one click.

MD UI: A TYPEFACE FOR INTERFACING

MD UI is a neo-grotesque typeface five years in the making, conceived by the author alongside MD IO in 2020 and developed with Luke Charsley, built specifically to optimize word and sentence-level legibility for everyday interfacing contexts. It comes in three sizes, unified in a single variable font spanning 48 styles, with features such as ink traps, widened apertures, and adjustable italic angles. Rather than imitating Swiss Style classics like Helvetica or Univers, it aims to advance the neo-grotesque tradition by correcting its known legibility shortcomings while preserving the familiarity that made it dominant.

MAKING EMOJIS AND ICONS SCREEN READER ACCESSIBLE

APPLE'S LATEST ACQUISITION IS TELLING SIGN OF WHERE THE COMPANY IS HEADED

Apple Inc.'s acquisition of MotionVFX suggests a renewed focus on pro software, likely tied to monetization through services like Apple Creator Studio and subscriptions. More broadly, it signals a shift toward investing where there's clear revenue potential—raising questions about whether future features like advanced AI (including Siri or iMessage improvements) could become tied to paid tiers such as iCloud+.

CUTE ILLUSTRATIONS AND CHARACTER DESIGNS THAT WILL INSTANTLY BRIGHTEN YOUR DAY

Olya Kovalenko is a talented digital artist from Belarus, currently based in Poland, who specializes in bright and expressive 2D and 3D character designs.

THIS INTERACTIVE TIMELINE SHOWS EVERY IPHONE SIZE, COLOR, SPEC, AND MODEL EVER RELEASED

An interactive project by sheets.works showcases all 52 Apple iPhone models released over 19 years, featuring specs, colors, pricing trends, and key milestones like Retina displays and larger screen sizes.