Hi, I'm Luis 👋

Today, we’re excited to introduce the state of enterprise AI report⁠. For the first time, we’re sharing a comprehensive look at how enterprises are adopting AI, what workers say they’re gaining, and how organizational leaders are turning experimentation into measurable productivity and new capabilities.

I became addicted to my smartphone within about a week. Having perpetual, low-resistance access to email, Signal, news, browser and podcasts and having my smartphone in constant physical proximity resulted in a notable acceleration in frequency of checking these apps. After a day or two, I began to find it difficult to leave my phone alone. I would put the phone in sleep mode, yet find myself picking it again, sometimes before even a minute was up. I invented a stream of reasons to check my phone and I found these impossible to resist, establishing a pattern of consistent little failures in discipline throughout the day.

I have a similar dedicated device setup but over the past few months have had to shift most of my computing onto my smartphone. After the shift, I've had a similar experience where I find myself reaching for my phone more often than I'd like. It's way too convenient to just "look that thing up" or "add that thing I need for the house to the cart". Personally, I don't mind carrying around or using multiple dedicated devices. I find the friction helps slow the impulses and makes me more intentional. Ultimately, my plan is to eventually transition back to dedicated devices. I've already done it to an extent with my Innioasis Y1

WordPress has been around for over twenty years. It powers 43% of the web. It’s not going anywhere. And if we build foundational systems that adapt to whatever the AI landscape becomes, WordPress will be the center of the open web for the next twenty years too.

The primitives we now have fit together like legos – and now WordPress really is a great place to build AI.

The open web is going to change. The way people consume content, the way they interact with services, the way they build and create – much of it will be mediated by LLMs and AI. The question is whether that future is open or closed. Controlled by a handful of platforms or distributed across millions of independent sites.

I know which future I’m building for.

In two new papers, Titans and MIRAS, we introduce an architecture and theoretical blueprint that combine the speed of RNNs with the accuracy of transformers. Titans is the specific architecture (the tool), and MIRAS is the theoretical framework (the blueprint) for generalizing these approaches. Together, they advance the concept of test-time memorization, the ability of an AI model to maintain long-term memory by incorporating more powerful “surprise” metrics (i.e., unexpected pieces of information) while the model is running and without dedicated offline retraining.

The MIRAS framework, as demonstrated by Titans, introduces a meaningful shift toward real-time adaptation. Instead of compressing information into a static state, this architecture actively learns and updates its own parameters as data streams in. This crucial mechanism enables the model to incorporate new, specific details into its core knowledge instantly.

As the proverbial creatures of habit, people tend to repeat the same behaviors in recurring contexts. This review characterizes habits in terms of their cognitive, motivational, and neurobiological properties. In so doing, we identify three ways that habits interface with deliberate goal pursuit: First, habits form as people pursue goals by repeating the same responses in a given context. Second, as outlined in computational models, habits and deliberate goal pursuit guide actions synergistically, although habits are the efficient, default mode of response. Third, people tend to infer from the frequency of habit performance that the behavior must have been intended. We conclude by applying insights from habit research to understand stress and addiction as well as the design of effective interventions to change health and consumer behaviors.

“Too computerized? Too cold? The world of tomorrow”, intones the narrator, “will be as cold as sunlight tuned through photochromic windows”.

Interesting view of the future as envisioned in 1967.

On January 1, 2026, books published in 1930 enter the U.S. public domain.

Books by William Faulkner, Franz Kafka, Agatha Christie, and Langston Hughes enter the U.S. public domain. In addition, The Maltese Falcon, perhaps the best-known noir book—and film—of all time, and books by Evelyn Waugh, Dorothy L. Sayers, and more, become free for anyone in the U.S. to read, use, and re-use.

Today I’m excited to debut Anti-viral, a new series where I talk to creatives about the meaningful work that algorithms overlook, and what they would do more of if attention was no object.

...we grow not only up—not only skyward—but down, into the roots, back to that from which we came and to which we will, one day, return. We become, in time, more rooted and resilient, more capable of surviving the storm, less easily shaken away from ourselves by idle wind or rain. When I think about growing down instead of up, I think about becoming centered, about knowing what work is ours to do (and, critically, what work is not), about a slow, steady power rather than a rash and inconstant one. After all, as anyone who’s ever lived among city trees can tell you, neither brick nor concrete nor iron can stop a root as it seeks out water. We should be as steady in our search for that which nurtures our own lives.

The metaphor of growth is one of hunger, consumption, acquisition—to acquire more pips on your collar, more titles after your name, more people under your domain. But craft asks instead, what are you doing? What reality comes into being with your shaping and working? What is in your hands and in your heart?

Org Social is a very peculiar decentralized social network, as it works from text files in Org Mode format. Each user interacts through their own social.org file, which they modify with their favorite editor. In this plain text file, you create posts, participate in conversations, leave comments in groups, react, create polls, vote, and much more. All this without depending on a centralized server, without algorithms deciding what you see, without registration, always with total control of your content.

As an org-mode user and fan of static sites, I find Org Social very interesting

1. CLAUDE.md is for onboarding Claude into your codebase. It should define your project's WHY, WHAT, and HOW.
   
   
2. Less (instructions) is more. While you shouldn't omit necessary instructions, you should include as few instructions as reasonably possible in the file.
   
   
3. Keep the contents of your CLAUDE.md concise and universally applicable.
   
   
4. Use Progressive Disclosure - don't tell Claude all the information you could possibly want it to know. Rather, tell it how to find important information so that it can find and use it, but only when it needs to to avoid bloating your context window or instruction count.
   
   
5. Claude is not a linter. Use linters and code formatters, and use other features like Hooks and Slash Commands as necessary.
   
   
6. CLAUDE.md is the highest leverage point of the harness, so avoid auto-generating it. You should carefully craft its contents for best results.

The 2025 IndieWeb Gift Calendar is the 9th annual group effort to gift (ship) one or more IndieWeb-related thing(s) each day of December that others can use to improve their IndieWeb experience.

This is such a cool initiative. I'm thinking of making my GitHub CMS project the gift this year. Technically, its usable today but still needs some polish and I need to find some time to write about it, architecture, and thinking behind it.

STARFlow-V is the first normalizing flow-based causal video generator demonstrating that normalizing flows can match video diffusion models in visual quality while offering end-to-end training, exact likelihood estimation, and native multi-task support across T2V/I2V/V2V generation.

HTML offers many of the conventional publishing idioms for rich text and structured documents, but what separates it from most other markup languages is its features for hypertext and interactive documents. This section introduces the link (or hyperlink, or Web link), the basic hypertext construct. A link is a connection from one Web resource to another. Although a simple concept, the link has been one of the primary forces driving the success of the Web.

Roundabout is a community space, built from the ground up with community leaders and neighbors.

The ultimate goal is to build for social trust — every decision, every design, optimized to build bonds and increase belonging.

I love the Fediverse but there are costs to federation and the cost of running a single user instance isn't amortized. I love having a static site and while I've added a ton of new features to my site, it's all still a bunch of static files. I know if I want to participate in the Fediverse, I'll have to host something to connect and interact with the ecosystem. That something though should be lightweight, cost-effective, and easy to maintain. My preference would be for my website to be my central digital hub that also knows how to "talk" different protocols like ActivityPub or AT Proto. Whatever I end up hosting would act as the bridge between my site and those ecosystems. Ultimately though, when possible, I'd prefer for a majority of my site components to remain static files.

I decided to launch another newsletter called Dealgorithmed. Will start on January 1st, delivered every 1st and 15th of every month. It’s gonna be a discovery newsletter focused on the personal/independent/whimsical/indie web.

The fact that people keep browsing the same 3 sites, day after day, getting served content by algorithms controlled by 3 companies is such a shame. Because there is so much interesting content out there ready to be discovered. And discovering new content also means connecting with new people, getting exposed to new ideas, different cultures. That’s by far the best quality of the web if you ask me.

The problem many people are facing is how to find that content, how to escape the algorithmic bubble. I think the only answer to that is curation.

The goal with Dealgorithmed is to provide interesting content gathered from all around the web in a convenient package delivered in your inbox twice a month. Content that you can then use as a starting point for your own internet explorations.

Dealgorithmed is a newsletter for the people who still love and care about the web, for those who are tired of doomscrolling the same three sites day after day.

A newsletter about the small web, the poetic web, the quiet web, the web many say we lost years ago, yet it's still here, ready to be rediscovered by those who care.

Subscribed! Really looking forward to this newsletter from Manu. If you're interested in more of his work, I recommend reading through the People & Blogs catalog.

Bloomberg’s Mark Gurman is reporting that Johny Srouji, senior vice president of hardware technologies, told Tim Cook he is “seriously considering” leaving Apple for another company in the near future. It was reported in October that Srouji was “evaluating his future at the tech giant.” While nothing is confirmed, it seems the executive is leaning towards not having a future at Cupertino.

Penn State and Matt Campbell reportedly came to an agreement earlier tonight, with the now former Iowa State head coach headed to Happy Valley on an eight-year deal. The first order of business for Campbell if you asked ANY Penn State fan was: ask and/or beg Terry Smith to stay at Penn State. Well, I’m not sure on the “ask” or “beg” part, but according to multiple reports (first by Audrey Snyder), Terry isn’t going freaking anywhere.

Love this. Happy for Terry.

You are on your own blog, your own corner of the web, powered by the platform you’re the CEO of, a blog that also serves content via RSS, the thing you’re building a tool for, and you’re telling people to follow the progress on fucking Twitter? Come on John.

😂 looks like I'm not the only one who thought this was a missed opportunity.

The intro to this version of Eyes is so good.

If 2025 was about the fediverse — with new activity from Ghost, Flipboard, WordPress, and also newer platforms — I think 2026 will be a partial reset to the IndieWeb. More blogs. More independent voices, as diverse as the web. More platforms and bridges that span multiple protocols.

Looking forward to it.

This is exactly what I need for my Podcast TUI app. I need to give this a try. Thanks @hanselman

Explore physical games tagged journaling and One-page on itch.io

Listeners have requested a list of LibriVox recordings of books recommended by Ambleside Online for use with homeschooling. Anyone looking for audio books for children will find plenty of ideas on this page.

The "Staff Picks" are a monthly feature of the LibriVox homepage and news. Every month, 10 books from our catalog are chosen according to a theme and are displayed on our homepage. The main goal is to show the variety of audiobooks we are offering, and to help listeners find the hidden gems in our catalog.

I want to leave them the web we had in the 90s and 00s, but doing a lot of the new things we've learned how to do since, without the silos.

Everything built on the web, everything replaceable, choice for users. And all the writing features of the web show through.

When this done, the writer's web will be as open as podcasting, something I had a hand in developing. Any time you want to switch platforms, you can, and lose absolutely nothing.

It's not decentralized, it's uncentralized.

Today, we are pleased to announce Fara-7B, our first agentic SLM designed specifically for computer use.

Unlike traditional chat models that generate text-based responses, Computer Use Agent (CUA) models like Fara-7B leverage computer interfaces, such as a mouse and keyboard, to complete tasks on behalf of users. With only 7 billion parameters, Fara-7B achieves state-of-the-art performance within its size class and is competitive with larger, more resource-intensive agentic systems that depend on prompting multiple large models. Fara-7B’s small size now makes it possible to run CUA models directly on devices. This results in reduced latency and improved privacy, as user data remains local.

Fara-7B is an experimental release, designed to invite hands-on exploration and feedback from the community. Users can build and test agentic experiences beyond pure research—automating everyday web tasks like filling out forms, searching for information, booking travel, or managing accounts. We recommend running Fara-7B in a sandboxed environment, monitoring its execution, and avoiding sensitive data or high-risk domains. Responsible use is essential as the model continues to evolve."

Racket 9.0 is here!

A major release is always exciting and Racket 9.0 is no exception in that it introduces Parallel Threads. While Racket has had green threads for some time, and supports parallelism via futures and places, we feel parallel threads is a major addition.

Today, we’re introducing WorldGen: a state-of-the-art end-to-end system for generating interactive and navigable 3D worlds from a single text prompt. WorldGen is built on a combination of procedural reasoning, diffusion-based 3D generation, and object-aware scene decomposition. The result is geometrically consistent, visually rich, and render-efficient 3D worlds for gaming, simulation, and immersive social environments.

"Today, we’re releasing Nano Banana Pro (Gemini 3 Pro Image), a higher-fidelity model built on Gemini 3 Pro for developers to access studio-quality image generation. This follows our release of Nano Banana (Gemini 2.5 Flash Image) just a few months ago. Since then, we’ve loved seeing the community put its key features to work — from character consistency to photo restoration, and even using its capabilities to make local edits in an infinite canvas."

"Today we’re introducing SIMA 2, the next milestone in our research creating general and helpful AI agents. By integrating the advanced capabilities of our Gemini models, SIMA is evolving from an instruction-follower into an interactive gaming companion. Not only can SIMA 2 follow human-language instructions in virtual worlds, it can now also think about its goals, converse with users, and improve itself over time. This is a significant step in the direction of Artificial General Intelligence (AGI), with important implications for the future of robotics and AI-embodiment in general."

...we present DS-STAR, a new agent designed to solve data science problems. DS-STAR introduces three key innovations: (1) a data file analysis module that automatically extracts context from varied data formats, including unstructured ones; (2) a verification stage where an LLM-based judge assesses the plan’s sufficiency at each step; and (3) a sequential planning process that iteratively refines the initial plan based on feedback. This iterative refinement allows DS-STAR to handle complex analyses that draw verifiable insights from multiple data sources. We demonstrate that DS-STAR achieves state-of-the-art performance on challenging benchmarks like DABStep, KramaBench, and DA-Code. It especially excels with tasks involving diverse, heterogeneous data files.

Paper

We’re excited to share that the Y2 is already in development!

Already Planned for Y2:

1）Hardware upgrades: New chipset, increased internal storage, and improved Bluetooth 2）Dedicated DAC for enhanced audio quality 3）New physical buttons: Power and volume controls 4）SD card slot for expanded storage 5）Improved file management 6）Better playlist support, especially M3U playback and easier playlist editing 7）Refined music playback logic with fewer menu layers 8）Support for more audio formats and improved metadata display 9）PC-created playlist compatibility

Yes! I really like what's planned for the new device and plan to get it on day one because I love my Y1. Wi-Fi is something I could do without.

Also, TIL under the hood it the Y1 is running Android.