Search the archive

LOCAL AI NEEDS TO BE THE NORM

Prioritizing on-device AI over cloud-hosted dependencies leads to greater user privacy, improved software reliability, and more efficient use of modern hardware.

IT'S A BUYER'S MARKET | AND A SELLER'S NIGHTMARE

AI has flipped the leverage in software deals because more tools are sliding into the "nice-to-have" bucket. One vendor in the piece was charging $50K a year right up until a competitor released a free module that did 80% of the same work. The renewal playbook now works because buyers have a credible threat that the seller can't easily dismiss. "We'll build it ourselves with Claude." That sentence, three months before renewal, is what gets pricing to move.

THE NEXT BIGGEST MOAT IN AI

The most important companies are actually organizational inventions. They create a new kind of institution around a new kind of work, which makes a new kind of person possible. Great companies are structures that let a certain kind of talent finally express themselves. The opportunity now is in companies that have not been possible before, and the people who have been waiting for it to exist.

WHY ATTACKING YOUR COMPETITORS ONLINE IS DUMB

Public competitor attacks never work. Naming a competitor only gives them more exposure, and most people don't care what's going on between two companies. Instead, make fun of trends, publish details and honest comparisons, bid on competitor keywords, and correct factual errors privately, if they actually matter. Use the energy instead to outship competitors and build a better product.

HOW AI PRODUCTIVITY FAILS

Most AI users are only about 10% to 20% more productive, despite how 'game-changing' they claim the technology is. However, it is theoretically possible for 2x or even 10x+ productivity gains. Real transformation requires organizational refactoring and a change in personal practice. This article looks at personal and organizational pitfalls that bottleneck AI productivity. AI amplifies what you already do well and whatever your organization already enables.

MOMENTUM IS THE SAFEST MOAT

Building momentum safeguards businesses against AI disruption by enabling rapid execution and innovation. Quick, cohesive shipping, as seen with Claude Code, provides a strong competitive edge. Companies slow down when scaling, so maintaining speed and focus is crucial to outpace giants with traditional moats like economies of scale.

LOCAL-FIRST COST TRACKER FOR AI CODING AGENTS

budi is a local-first cost analytics tool for AI coding agents. It allows developers to see where tokens and money go across most popular AI coding tools, breaking down costs by repository, branch, ticket, and file. budi runs everything locally by default. It has an optional team dashboard that syncs aggregated metrics.

GOOGLE LAUNCHES GEMINI 3.1 FLASH-LITE IN GENERAL AVAILABILITY

Gemini 3.1 Flash-Lite is now generally available through Google Cloud platforms. The model provides ultra-low latency and can handle high-volume processing. It is the most cost-efficient Gemini 3 model. Gemini 3.1 Flash-Lite has multimodal capabilities and supports both text and image processing. The model can handle agentic tasks like tool calling and orchestration.

FIGMA'S COLLAPSE SHOULD TERRIFY SAAS LEADERS

Figma has lost roughly 85% of its market value since its IPO in August last year. The market has been pricing design-tool disruption for months, and the question now is not whether agents will replace design workflows but rather how fast the transition will happen. Figma's vulnerability was that its product was the workflow itself, something that AI agents excel in. This post discusses a three-layer framework for repositioning products over the next few years to help companies avoid succumbing to the same fate as Figma.

IN THE TRENCHES WITH 50 MIDWEST CIOS

The story in enterprise AI right now is that the bottleneck has moved off the model. Fifty CIOs from Midwest enterprises met in Chicago last week, and when asked who had agents at scale, no one raised a hand. Of 25 in a smaller breakout, only 5 had agents in production at all. The reason isn't model quality. It's that nobody has actually written down how the work gets done inside these companies, so there's nothing for an agent to operate on.

SLOP VS TASTE

Most people use "slop cannon" as an insult for a person making AI slop. The piece flips it. The slop cannon is a person using junky AI output as raw material to think with, then editing hard. Move fast in produce mode, slow down in selection mode. The piece also bets against the new OpenAI and Anthropic consulting joint ventures, on the grounds that most enterprise data isn't ready for the work being sold.

THE HARD THING ABOUT EASY THINGS

Management is more unpleasant when you're not defending humans, and when the blame you're absorbing is not on behalf of humans.

WHAT DO YOU HAVE TO BELIEVE?

If you can't describe what you'd do to kill your own company, you don't fully understand what you're building.

AI CHAT BOTS SHOULD BE DEPRECATED

Chatbots essentially offload interface design to users.

IS META REALLY DYING?

Claims that Meta is in decline overstate a short-term drop in daily active users and ignore the scale of its business. Facebook usage has shifted as social media moves from connection to entertainment. Users are spending more time on TikTok and Instagram and using Facebook mainly for updates from friends and family. This reduces per-session engagement and supports the view that Facebook is less central than before. Even so, Meta still reaches billions of users and continues to grow revenue, using that position to invest in AI, AR, and wearables.

THE GREAT $110 TRILLION WEALTH TRANSFER WON'T HAPPEN ANY TIME SOON

The “great wealth transfer” is expected to be slow rather than sudden as older Americans live longer and keep accumulating wealth. Americans 55 and older control about $110 trillion, with boomers adding over $1 trillion in wealth in a recent quarter. Bequeathable wealth has risen to 424% of GDP, with 97% of gains driven by households over 55. Much of this wealth is first passed to spouses, while spending on longevity and care reduces what is left for heirs. As a result, Gen X and younger generations will see a gradual transfer instead of a windfall.

WHAT TO DO IF YOUR CUSTOMER COUNT IS FLAT BUT YOUR BASE IS FALLING APART

Flat customer growth can mask a weakening base and declining profitability. In one case, retained customers fell 18% while net-new customers rose 31%, yet margin-per-customer dropped 14%. To surface this early, segment customers into retained, reactivated, and net-new groups, then track margin-per-customer over time. Monitor CLV decile migration to see whether high-value customers remain loyal and mid-tier customers move up. Margin-per-customer is the key leading indicator, often signaling base erosion 6-9 months before revenue declines.

AUTHORSHIP NOW DRIVES SEARCH VISIBILITY IN GOOGLE

options. Get 3 months of Cuey PRO free with code TLDRMARKETING

HOW PERPLEXITY WINS THE 10X KNOWLEDGE WORKER

Perplexity Computer is not a chatbot or search tool, but a system for building and running workflows end to end. Users describe what they want, and it assembles the stack and executes it with no code. It turns multi-step tasks into automated systems that run in the background, like pulling Reddit trends, ranking posts, generating hooks, and syncing outputs into Notion on a schedule. Its advantage is orchestration, combining multiple models and tools into one workflow that handles real work.

COURSE CORRECTION: GOOGLE TO LINK MORE SOURCES IN AI OVERVIEWS

Google is adding several new link formats to AI Overviews and AI Mode in an effort to drive more traffic back to the websites its AI summarizes. Changes include a "Further Exploration" section at the bottom of AI answers linking to relevant articles, an "Expert Advice" block surfacing snippets from news outlets and forums, and preview pop-ups when users hover over source links. Google is also testing a subscription integration that would surface a user's subscribed sites more prominently in AI results, with early tests showing higher click-through rates.

EXPERIMENT ESTIMATION STUDY: HOW WELL CAN 1,391 GUESSES PREDICT A/B TEST EFFECTS?

A/B test impact is often overestimated, making gut-based prioritization unreliable. Across 1,391 predictions, the median error was 2.7 percentage points, and 56% of guesses were too high. Experience had a limited effect, with teams of 10+ years performing similarly to newer practitioners. Familiarity with similar tests improved accuracy, raising directional correctness to 65.1% and reducing median error from 3.2 to 2.5 points. Small groups performed better than individuals, with crowds of five cutting error to 2.3 points, though a consistent positive bias remained.

THE WORK OF KNOWLEDGE IN THE AGE OF AI REPRODUCTION

Walter Benjamin's idea of mechanical reproduction is extended to knowledge, where AI collapses the scarcity of knowledge work in the same way mass media did for art. Consulting decks, medical analysis, and legal drafts can now be produced instantly, reducing the value of the artifact while leaving judgment and accountability as the main sources of value. This creates a barbell effect where low-cost AI output floods the market and value concentrates in institutions and credentials that can be held responsible for decisions. As skills become easier to replicate, signaling and provenance rise in importance, similar to the sheepskin effect, where degrees matter more than marginal skill gains.

GOOGLE TO NO LONGER SUPPORT FAQ RICH RESULTS

Users will no longer see FAQ rich results in Google Search results.

IKEA TURNS ITS FRAKTA BAG INTO A CAMERA FRAMING THE WORLD

A playful IKEA campaign that turns its iconic blue bag into a Point-Of-You frame.

80% OF COMPANIES CUT JOBS FOR AI. IT DIDN'T IMPROVE THEIR RETURNS. NOT EVEN A LITTLE

There's no correlation between AI-driven layoffs and improved ROI.

APPLE'S CAMERA-EQUIPPED AIRPODS REACH LATE TESTING IN AI DEVICE PUSH

Apple is in the late stages of developing new AirPods with built-in cameras. The prototypes feature a near-final design and capabilities. The device will be Apple's first foray into AI-enhanced hardware. While the hardware is nearly ready, there are still concerns about the AI elements, which could further hold back a launch if the quality of the visual intelligence features isn't good enough.

GOOGLE UNVEILS SCREENLESS FITBIT AIR AND GOOGLE HEALTH APP TO REPLACE FITBIT

Google's newest Fitbit Air has a screen. It contains a suite of health sensors that pipe data into the new Google Health app, which has an AI-powered health coach that can tell users what that data means. The device is a small plastic puck that fits into various bands. There will be plenty of colors and style options available. Pictures of the devices and bands and screenshots of the new Google Health app are available in the article.

WHAT THE HELL IS HAPPENING IN CHINA?

China has several early-stage biotechs with over 10 development candidates. It is much easier to develop drugs in China. Chinese biotechs lacked access to later-stage development resources in the past, so they have always leaned toward a breadth-first strategy. This has resulted in an industry with heavy competition that will likely surpass its Western rivals within the next few years.

THE AI REVIVAL OF THE THREE MILE ISLAND NUCLEAR PLANT

The buildout of AI infrastructure in the US has transformed the country's energy needs. A new crop of companies has developed nuclear reactor designs that they claim to be cheaper, safer, and easier to build than the ones currently in operation. However, it will take many years before these technologies will meaningfully contribute to the US energy supply. This means that the country will have to rely on much older technology until the new plants come online.

and docs. That's why this realistic enterprise RAG benchmark

tests 12 AI tools against 500 questions and 500k docs across 9 company sources. Onyx is the only system that beats OpenClaw. Check out the

benchmark on GitHub

BEHIND THE SCENES HARDENING FIREFOX WITH CLAUDE MYTHOS PREVIEW

Mozilla recently announced that it had identified and fixed an unprecedented number of latent security bugs in Firefox with the help of Claude Mythos Preview and other AI models. This post goes into detail about how the team approached this work, what it found, and advice for other projects on using emerging capabilities to harden against attacks. Many of the bugs discovered would need to be combined with other exploits to achieve a full-chain compromise.

OPENAI LAUNCHES NEW REALTIME VOICE AND TRANSLATION AI MODELS

OpenAI's API now supports live voice agents, instant translation, and streaming transcription. Its new GPT-Realtime-2 model offers GPT-5-class reasoning for spoken conversations. GPT-Realtime-Translate, designed for live multilingual voice products, supports speech input in over 70 languages and output in 13 languages. GPT-Realtime-Whisper can transcribe audio as people speak, making it ideal for live captions, meeting notes, and other situations where speech needs to be converted into structured text during the conversation.

CLOUDFLARE TO SLASH 1,100 JOBS DUE TO AI-DRIVEN RESTRUCTURING PLAN

Cloudflare plans to slash 1,100 jobs as part of a restructuring plan that it claims will define how a world-class, high-growth company operates and creates value in an agentic AI era. The company says it will become even faster and more innovative by embracing an agentic AI-first operating model. The layoffs are expected to be substantially complete by the end of the third quarter. Cloudflare expects to incur charges between $140 million and $150 million for the layoffs.

ELON MUSK TRIED TO HIRE OPENAI FOUNDERS TO START AI UNIT INSIDE TESLA

OpenAI claims that Elon Musk was always happy to commercialize the lab as long as he remained in charge. Musk had tried to hire OpenAI's founding team to lead a new AI lab within Tesla in 2018. The deal would have brought Sam Altman, Greg Brockman, and Ilya Sutskever to the company, with Altman becoming a board member or OpenAI becoming a Tesla subsidiary. The claim contradicts Musk's accusations of Altman stealing the charity by converting the company into a for-profit.

DOWNLOAD THE 4 STANDARDS.

AI LOAD BREAKS GITHUB – WHY NOT OTHER VENDORS?

The fact that Microsoft's competitors seem to be keeping up with increased load due to AI suggests that the company has not been responding to its growth like a world-class engineering organization.

CODEX NOW WORKS DIRECTLY IN CHROME ON MACOS AND WINDOWS

The agent can work in parallel across tabs in the background without taking over the browser.

THE AGENT PRINCIPAL-AGENT PROBLEM

'Contributors' submitting slop PRs is an unmanageable disaster - reviewers could be more productive by prompting agents themselves.

attention. Learn more about advertising opportunities.

TOKENMAXXING, PROMOMAXXING, AND MISALIGNED INCENTIVES IN TECH

When a measure becomes a target, it ceases to be a good measure.

MARKETS IN EVERYTHING?

Properly implemented and regulated, markets are the best fundamental arrangement of society for maximizing human flourishing.

MATT MULLENWEG ASSEMBLES TRUSTED GROUP TO OVERHAUL WORDPRESS.ORG AND FIVE FOR THE FUTURE

Matt Mullenweg has given a small group of trusted contributors the authority to overhaul WordPress.org without approval from any team, committee, or stakeholder other than himself.