Working with omp and OpenCode Zen is a pleasure. Fixing up a KenKen game in this session. It's amazing: I can spend a couple of bucks and get a game with zero advertising, runs on Android and iOS in any browser, is about 50k, smaller than any app available on any app store. It's also free, so you can take it and ask *your* AI to add the feature or variant you want, and you can share it, if you like. There's a bright interesting future for small tools built with AI.
I talked about Brave Origin a few days ago (see https://rpdillon.net/now/post/22 ), and after a week using it on mobile and desktop I thought it was worth jotting down some notes. Origin has a bunch of quality of life improvements, and is truly user-focused. It's really, really nice. Reminds me of how I felt about Kagi when I first picked it up. Origin is fast, featureful, and has power-user features like Manifest V2 extensions (uBlock Origin in particular), has a built in, efficient adblocker besides, has native vertical tabs, has toolbar affordances for all currently playing/paused media, and very cool features like "tab search" that search across "audio/video", "current", and "recently closed". Nicely conceived, and very handy. I'm now using it at work as well. It's extremely smooth. I present a lot in meetings I run, and Brave is handling the split screen views and multiple windows well, and has a nice UI that stays out of the way, and feels professional. That's soft praise, but I guess I'm saying the experience moving from Firefox has been overall good. brave://flags is a hardcore poweruser feature, and it's right there to tweak. That whole vibe about the browser makes me feel like it was $60 well-spent to buy it, even though I run Linux and could get it free of charge. Worth it to have it on Android and Mac as well. I really hope Brave sticks with the model.
We gotta downgrade how important politics are when it comes to interpersonal interactions. I think polarizing the populace is a good technique for gaining power, since you stop people from talking to each other. I talk daily about politics with folks that disagree with me on a lot of things, but we still really like each other, and always learn something when we talk. I watch this show, Love is Blind, where people date without seeing each other. Somewhere in the middle of season 7, this guy Ramses has all this trouble with the girl he loves, because she served in the military. He places politics above his love of the woman, and his excuse is: "It's a moral thing for me, your politics is tied to your ethics." So, he's young, and he's wrong. Politics is, 90% of the time, a disagreement about *how* to obtain a goal, rather than what that goal is. Politics is also a grift, where candidates sell one thing (making our country great) and instead do another thing (gain wealth and power by any means necessary). So when Ramses does this, he comes across as hopelessly naive. And to judge a woman, a Surface Warefare Officer in the United States Navy, with Strike Officer credentials and release authority for a tomahawk launch on a DDG, of being *immoral* is a simply wild claim. The military is a tool, and how it is wielded varys widely. In my time in the Navy, I did about as many humanitarian missions in southeast asia as I did combat missions in the gulf. And some of the "combat" missions were actually recovering US prisoners of war (https://en.wikipedia.org/wiki/Jessica_Lynch was recovered to my ship, USS Boxer, while I was on watch. I didn't know at the time.) To his credit, he works for Impact Justice, which has noble goals: > Impact Justice advances safety, justice, and opportunity through boundary-breaking work that honors and empowers people and is changing expectations about what we can accomplish together. > Sometimes you have to reinvent the wheel. > We spend billions of dollars annually on systems of law enforcement and punishment that fail to provide the safety and justice everyone deserves—systems that perpetuate racism and cycles of harm and that dehumanize and divide rather than unite us. It’s time to get creative, change the script and rules of engagement, believe in one another, and expand what’s possible. So, reading critically, there's nothing there. The goal is superb, but the means? Not so much. Let's browse the site. Oh! I see they have a whole page dedicated to "Our Impact". Let me paste *the whole thing* here: > We leverage research and other knowledge to connect the desire for change with informed action and work to shift the narrative so that decision-makers and the public understand what’s at stake and what’s possible. What the actual fuck does this mean? It's completely untethered from anything meaningful. > We engage in three interrelated areas of work: > Reducing the number of people caught up in harmful legal systems. That's an outcome, not what you do. What do you do? > Improving living conditions and life outcomes and safeguarding fundamental human rights for the millions of people still subject to mass incarceration. OK, sure. How? > Expanding opportunities for formerly incarcerated and other system-involved people to heal, fully participate in community life, and reach their potential. This isn't impact, it's a goal. What metric did you move? > In all our work, we aim to both create positive outcomes for people who are directly served by our programs, and for those outcomes to ripple out more broadly to communities and make us all safer. Glad to hear your goals. Why should I give you money? Sorry to digress there, but this kind of political virtue signaling is nauseating. He enjoys his ability to have these opinions because of the dominance of the US military and economic engine. Bite the hand that feeds you, and all that. But Ramses might still be a great guy. Just because he disagrees on stuff like whether the military is ethical, is almost orthogonal to whether he's a good person. Yes, sometimes judgment failures are part of a larger pattern of cluelessness, but often they are narrow weaknesses that you can learn to work around, and otherwise have an amazing relationship. That's why I think politics should play less of role in our interpersonal conflict, and instead inspire discussion and thought. We're all in this together.
After many years with Firefox/Librewolf I'm switching my main driver to Brave Origin. I need to put my money where my mouth is. I've always wanted software that is: * User centric * Privacy-oriented * Open-source so the user has a measure of control But funding that is hard. I've long said that if folks would just sell their open source, I'd buy it. Well, Brave is selling a great open-source browser with all the extra cruft taken out. It's actually even better: there's a dashboard where you can enable or disable whatever features you want. So far: quite happy. My biggest concern is that most folks won't pay. I'm hoping Brave can secure several ways of funding so they can preserve Origin. ---- Technically, Origin does not appear to be a separate codebase; it looks like a different build of the main Brave Browser (itself based on Chromium). Critically, Origin disables revenue-generating features like the crypto wallet, Brave Rewards, Leo AI, and telemetry. This is awesome. It does require a 1-time payment of $60. I'm on Linux, so I could get it free, but it seemed nice to pay.
I've written before about Scoundrel, my favorite solitaire game: https://rpdillon.net/scoundrel.html I recently was searching around for some variants, and ran across Regicide, a co-op multiplayer game from the same creators: https://www.pagat.com/invented/regicide.html Excited to try it soon!
Just ran across worm blossom, willow, and sneaker web, and I'm very intrigued. https://worm-blossom.org/ https://willowprotocol.org/ https://sneakerweb.org/ A lot of what I've been using AI to build in the evenings are offline, capable, highly resilient, simple pieces of software for basic information sharing and online interaction, where I use the word online perhaps a bit loosely. In particular, I'm putting the finishing touches on a tool called AppDeck, which can serve plain HTML files that are entirely self-contained applications from a directory on disk, and then make that available both on the local network and to local users for editing and update. I'm going to give sneakerweb in particular a much closer look because that seems to be highly relevant to my interests.
So I'm never buying a Kindle book again because Amazon got rid of downloads. But my friend recommended the Silo series by Hugh Howey, and so I thought I'd pick it up. I have bookshelves of old books, so I do prefer digital, so I looked around for lists of book sellers that offer books DRM free, and decided to take a plunge and spend $22 on Kobo's store to get the trilogy. Huge mistake. The entire book industry has turned to total shit, following in the footsteps of the movie industry (who I never give a dime to anymore). No ability to download the book to read on Xteink reader, and even more Calibre plugins to actually get the file (Adobe Digital Editions trash + Kobo's own DRM). Never buying from Kobo again; that was my first and last book from them. Ended up using Anna's Archive to get a copy I could actually read on my reader. Why do they insist on turning everything to shit? Just short-term thinking constantly. There are exceptions. Special shoutout to Tor Books for understanding the way: https://torpublishinggroup.com/ So if you head over to Bookshop.org, you can buy Tor books and actually download the file. Amazing.
Been messing around with https://omp.sh as a substitute for OpenCode's harness. It certainly has a lot going on, but it's going to take me some time to understand it. One interesting experience: the very first message you read when you fire up omp is "Press ? to view keyboard shortcuts". But when you press ?, nothing happens. This is a failure mode unique to AI coding. Apparently the functionality was there, and got removed, and they are spending 3 weeks trying to add it back: https://github.com/can1357/oh-my-pi/issues/1614 This is an extremely poor first impression, despite the tool otherwise being very impressive. Weird experience (and I still don't know what shortcuts to use...the best I found was https://github.com/can1357/oh-my-pi/blob/main/docs/keybin... UPDATE 27 Jun: Issue closed, this turned out to be nothing more than a dev forgetting to remove that help line. You're supposed to use /hotkeys now. Amazing how much difference attention-to-detail makes.
`du -d2 | sort -nr | less` is super helpful for tracking down disk usage.
Google tends to slowly choke freedom out of its products, with the latest installment a Chromium commit that finally removes Manifest v2 from the codebase. https://chromium-review.googlesource.com/c/chromium/src/+... I was curious what Brave's take was, since I remember someone at Brave quoted as saying they wouldn't be able to independently maintain a patch-set to retain Mv2. I checked again today, and they have instead done something quite cool! They have brought support for the key Mv2 extensions into the core directly, so they don't need to maintain Mv2 separately. Nice move! https://brave.com/blog/brave-shields-manifest-v3/ It looks like that happened almost a year ago, and I had somehow missed it. Glad to find out now!
Been using OpenCode almost exclusively (really like both the harness and OpenCode Zen), but I'm thinking of experimenting with something like https://omp.sh since I don't currently leverage features like sub-agents as much as I probably should. There's just so much to learn around all of this, even after you understand much of how the underlying systems are put together.
Extremely hyped for how the trailer for Doom Dark Ages - Revelations is looking. I haven't played Dark Ages yet because there was something about the gameplay that turned me off...it felt slower somehow than both 2016 and Eternal. Anyway, the Revelations DLC for DE is a massive change! Dashing is back! And there's the...I'm not kidding...grappling hook with full angular momentum, so you can reel in to whip around stuff as you're fighting it. And a spear. And some lore! Check it out: https://www.youtube.com/watch?v=nL_JF1ZNllI
I've always been fascinated by how societies will convince themselves they are somehow virtuous by passing, for example, labor laws, or environmental laws, and then inevitably don't like the cost increases associated with those laws, and therefore outsource to other countries that don't have those laws. This effectively punishes those that follow the rules, and rewards those that are the most ruthless. iPhone owners walking around talking about how unenlightened we were when we had slavery don't know what to say when I point out how their phone was made. When I was in the Navy, we'd pull into Thailand, and we'd have to pay a service their to pump our sewage to another ship where it could be processed, because the US Navy has rules against dumping sewage in foreign ports. But a lot of foreign ports don't care, so they'll happily take a couple hundred thousand USD to pump the sewage off the ship, and then cruise right across the bay where we were moored and dump it within a mile of shore. Easy money. Apparently something similar is happening with pesticides in the EU: "EU-banned pesticides found in rice, tea and spices" documents it, and points out: > Although these chemicals are not allowed on the EU market, they can still be exported from European Member States to third countries. From there, they can return to Europe as residues in imported food — a “toxic pesticides boomerang” that puts consumers at risk. The pattern holds: ban the thing that is bad, then have other countries use those things and import the product. Fascinating. https://www.foodwatch.org/en/eu-banned-pesticides-found-i...
The reason Cordel (the software that powers this site) and PiFeed (my simple social network) both lean heavily into RSS is that it's still the right tool for the job - high engagement feeds on Facebook, Insta, Youtube, and Tiktok drive human eyeballs, but lack the structure and control that RSS provides. That's why RSS won't die, and why AI is turning to RSS to get reliable updates! https://julienreszka.com/blog/rss-is-back-ai-agents-are-r...
This HN comment by "gormo" is the reason I visit HN. It's the only nuanced, intelligent discussion I've read about the de/merits of AI and software engineering, and it's clear the participants are intelligent and articulate. Really appreciated the back-and-forth. https://news.ycombinator.com/item?id=48357291
Comments that writing is done by AI get tiresome. I hang out on HN a fair amount (my main social media, tbh), and the accusations that articles, papers, and code are written by AI are getting out of hand. It would be interesting to have a scaffold that allowed you to take the raw prompts or initial versions of folks' writing and juxtapose it in a dynamic way with the "AI polished" version to dispel some of this commentary.
I tend to work on projects in my spare time that are small, self-contained web pages that do things like display a calendar, kanban board, web launcher, spreadsheet, document, log, world clock, etc. These are self-contained capsules of capability, containing both application logic and application data in a portable format that runs just about anywhere, and requires no special infrastructure (a simple WebDAV server is all that's needed). Since it is all file-based, the hosting, backup, restoration and duplication story is solved. I often wonder why these very simple applications are so rare; it may be because they are difficult to monetize. But since I don't care about that, but rather hackability and portability, I think they are pretty ideal.
It's frustrating that it's been more than 20 years since Katrina and we haven't done too much about this. Relocation seems like a multi-decade effort. > Coastal Louisiana faces sea level rise of around 10 to 23 feet, according to the analysis published in May in the journal Nature Sustainability. The impacts will be bleak: around 75% of its remaining wetlands are set to be lost and its shoreline could retreat inland by up to 62 miles, the scientists found. https://lite.cnn.com/2026/05/25/climate/new-orleans-sea-l...
Really excited to see this! The problem is easy to understand, but wickedly difficult to solve. > The result is also notable for how it was found. The proof came from a new general-purpose reasoning model, rather than from a system trained specifically for mathematics, scaffolded to search through proof strategies, or targeted at the unit distance problem in particular. As part of a broader effort to test whether advanced models can contribute to frontier research, we evaluated it on a collection of Erdős problems. In this case, it produced a proof resolving the open problem. https://openai.com/index/model-disproves-discrete-geometr...
Really impressed with with Antirez is doing over at https://github.com/antirez/ds4. My main gripe is that I'm running a 128GB Strix Halo machine with ROCm, and that's not a primary target for DS4. I think the endgame will be llama.cpp fixing this issue (https://github.com/ggml-org/llama.cpp/issues/22319 requests model support for DS4), and then running that via lemonade's build of llama.cpp (https://github.com/lemonade-sdk/llamacpp-rocm/releases) with an Unsloth quant of DeepSeek v4 Flash (like https://huggingface.co/unsloth/DeepSeek-V4-Flash). Just waiting for progress on llama.cpp, since everything else should be ready.
One interesting application of LLMs is to have them read through privacy policy and terms of use for you so you can look for anything particularly hazardous to your considered use cases. I did this for opencode.ai and discovered that their Zen product provides better guarantees about privacy than their Go product, so I'm switching to Zen, even though it's a little bit more expensive.
The real insight over the past 2-3 days has been a combination of: 1. Local models like Qwen 3.6 and Deepseek v4 Flash (even in a 2-bit quant!) are quite useful, and understand integrations like tool calls well. 2. A harness like OpenCode is an immense step forward in keeping local models on-task. This site was crafted almost entirely locally with Qwen 3.6 35B-A3B. 3. It's becoming clear that adding on additional skills and integrations will pay huge dividends in getting it to operate autonomously for hours. I think a project like my good friend's https://dex.rip for cross-session memory and tracking will take this even farther.
Working with some local models (Qwen 3.6 35B-A3B and Qwen 3.6 27B UD, both in Q_6K) as well as some cloud-hosted models (Deepseek v4 Flash hosted through OpenCode.ai) to build a simple feed page, can be hosted through CGI or as a server. This is the first deployment - fairly happy with how quickly it came together without any frontier model contribution!