Part of the reason posting has been a bit light recently is because Rob and I have been working on Outer Web Exploration Mode. And there are a few other related things to talk about:

1. Rob managed to squeeze keyword search out of the existing code
2. The finishing touches on embeddings for OW
3. Some other indieweb indexes and commentary on them
4. Using Exploration Mode

Rob was curious if Outer Web could help him with some devops/hosting stuff and decided that keyword search (rather than webpage peer search) should be a feature of the site. But how does one search on a few keywords when the system is designed to match long form text? The default answer in 2025, "Hey LLM, generate me a blog post about using [x] with [y] to accomplish [z]"

Rob	Please write a 150 word blog post on switching from loki to signoz for log UI

LLM

Switching from Loki to SigNoz for Log UI: A Quick Win We recently migrated our logging UI from Grafana Loki to SigNoz, and the difference has been immediate. Loki served us well, especially with its tight Grafana integration, but we found it lacking in search UX, filtering flexibility, and correlation with traces. SigNoz, by contrast, offers a modern, unified observability interface with excellent log, metric, and trace integration out of the box. Its UI feels purpose-built for debugging production systems-search is fast, filters are intuitive, and logs can be seamlessly tied to trace data, which speeds up root cause analysis dramatically. Setup was straightforward using the OpenTelemetry collector we already had in place. We still run it self-hosted, and the resource usage is comparable to our previous stack. If you're already on OpenTelemetry and want a more focused log experience, SigNoz is absolutely worth a look. We're not looking back.

He was reasonably satisfied with the results, they included:

jessitron.com Run an OpenTelemetry Collector locally in Docker - Jessitron This is an experience report for my future reference (and yours). The OpenTelemetry Collector is useful for receiving trace data in whatever format and exporting it to the back-end of your choice for storage and querying. For instance, I wanted to receive traces over HTTP/JSON from this sneaky browser extensionLINK and send them to Honeycomb....

logz.io Auto-Instrumenting NestJS Apps with OpenTelemetry | Logz.io Looking to use distributed tracing in your NestJS application? Follow this tutorial for using OpenTelemetry auto-instrumentation

nodesource.com Enhance Node.js Observability with OpenTelemetry in N|Solid Enhance Node.js performance with OpenTelemetry in N|Solid. Learn automatic instrumentation, OTLP export, and seamless integration with observability tools.

At the moment it's still BYO LLM but we're fairly confident in the process.


The Outer Web proof of concept was focused on allowing web contributors (bloggers et al) to find peer content in hopes that they will link each other. With the first product out the door, Rob and I set about using the same database to let casual browsers explore the indieweb. The result is an interface resembling a dungeon crawler and a wikihole ͥ user experience.

Exploration mode can be entered from one of two places:

1. A post query
2. A set of category entry points ('football', 'knitting', 'books', etc.)

I wrote a couple of posts on my embeddings implementation for Outer Web (as well as my ssg flavor of it). My final tweak to this code was to correct a glaring issue with compression. Quick recap:

• Rather than storing all embedding values, I only store the large ones as a list of (position, value)
• Each of these pairs is stuffed into a 32-bit value, meaning the 32-bit floating point embedding value is quantized into 16 bits

In hex, this looks like:


In binary:


This is pretty wasteful, with two bytes we can index 65k embedding array indices where embedding vectors are typically fixed at a length of 100-300. So:


Much better. Nine bits of position information, one bit of sign, and 22 bits of value. This is way higher precision (though a bit less convenient) than two shorts smashed together.

Source.

Over the years I've linked neat websites that allow you to search or stumble through the indieweb/blogosphere. There are a few big collections of RSS feeds on Github and elsewhere that have perhaps made blog indexing more accessible to the casual developer. Minifeed looks promising, though I haven't played with it enough to have much commentary. Another one, Blogscroll, garnered some lively discussion on Hacker News. The site itself is still developing, but the HN commentary had some criticisms/suggestions worth considering for OW.

mikae1

Weird. I love a good link list, but most links I clicked were just regular portfolio websites, not digital gardens in the gardens vs. streams sense. Does the creator not know what a digital garden is?

Takeaway: authenticity only goes so far, you still need to link to something informative or entertaining.

openrisk

Cool but this will grow unworkable as the number of entries increases and will require some sort of filtering / searching. Even pure signal can be overwhelming these days when so much good stuff exists. Which brings us to the (as far as I know unsolved) question of supporting large scale discovery of the web without drifting into enshittification. Some sort of decentralized index that will be distributed in a torrent-like manner might work but that requires curation too: Who and with what criteria can add an entry etc. Bottom line is that the walled gardens did not exist, they evolved because the original web was missing critical components of usability. They exploited a vacuum. To fill the vacuum with something more benevolent we need to go back and solve these problems. The rest will be history.

Takeaway: the web is large and both the database and UI needs to handle scaling. Also, don't enshittify.

klez	> that requires curation too: Who and with what criteria can add an entry etc. IMHO that's the beauty of it. Sometimes I want to be hit with everything the web has to offer. But some other times I do want curated lists of links organized by category with opinionated criteria for inclusion.

Takeaway: there's more than one use case for exploring the web. Rob('s bot development team) added static links to Exploration Mode so it's easy to share pages. But we have no user accounts and no way to customize the experience, for now.

rednafi

Love the idea, but I'm not sure how well it will scale. I've seen similar lists before and even went out of my way to include my blog there, only to never find that page again. That said, I'm loving this renewed interest in building our own little corners on the internet. I have mine too.

Dani is at a pretty good age for traveling, so I've got my eye out for trip ideas. It's easy to do another Catalina weekend or visit family or do a quick Hawaii trip, but for something more ambitious it's nice to have diverse information sources.

Social media and professional writers (Rick Steves planned my honeymoon) aren't terrible sources for this, particularly compared to whatever you'll get from a search engine. Still, I like the idea of a thoughtful, photo-filled blog post.

One example:

vienna-pyongyang.blogspot.com The forbidden railway: Vienna - Pyongyang - - ???? - ??? - ??: By train across North Korea (1/2) Download the kmz-file for GoogleEarth to follow our route inside North Korea and see, where the photos were taken. Most of my photos from in...

Well okay, this isn't inspiring me to make the same trip but it's a pretty good demonstration of the benefits of a blog post over a travel guide or social media reel. A few other posts I found on the Outer Web:

52things52weeks.com The Great American Road Trip Part II: 3,000 Miles in America - 52 Things 52 Weeks Three years ago my best friend and I piled in the car for a 2000-mile drive from Knoxville to Phoenix.In an effort to constantly outdo ourselves, this time we embarked in a 3,032-mile, 11-day, 12-state journey. A vital part of any road trip is the right playlist. Jill and I put an unhealthy amount of...

idratherberiding.com The Ducati Multistrada does the California Sierras | I'd rather be riding... The Multistrada found its way into my garage for me to open my street riding horizons. It is here to take me on long distance trips while bringing an edge to the touring side of riding, the Ducati edge I learned about with the Streetfighter. Therefore, when guys from the Ducati.ms portal invited me to...

www.oneman-onemap.com Setting sails on Lake Ashi | One Man, One Map

sam.hooke.me Holiday to Italy visiting Pompeii, Herculaneum, Vesuvius and the Amalfi Coast The home page of Sam Hooke.

rick.cogley.info Pick Me Up in Yamanashi? : Rick Cogley Central Our oldest daughter Kylie jokingly asked if we could pick her up in Yamanashi, which is the next Prefecture over from Kanagawa where we live. She was going to be at a rock concert, one of many on the Okamoto's 47 tour (they're touring all 47 Prefectures), and she was not feeling like schlepping from the venue to the train station, 30 min away on foot. The Planning On a lark, we decided to go for it, to get in a day trip down to Japan's wine country, and maybe hit an onsen hot spring.

janikvonrotz.ch Janik von Rotz - A Glimpse of China I visited China from the beginning of last december until christmas. My backpack was filled with prejudice and the expectation of having a hard time travelling there. The reality looked quite different. China was a blast! Vibrant, lively, fascinating, contradictory and quite the opposite from the european life style. Swallowed in Beijing and spit out in Hong Kong I enjoyed China all over. Definitely not the last I have been there. Again my favorite pictures:

Related / internal

Related / external