Show newer

Computer programming is about reading through 20k lines of diff to figure out WTF broke in your dependencies and how to work around it. ☝️🤓👍

@SecureOwl Yeahhhhhh. TBH undocumented user-focused APIs have been my go to. I just inspect what a web app is tying to do and strip it down. Works most of the time until they change something. :P

I opened a technical blog post and left the tab to read later, only to discover it was choking the browser to death with a sudoku-solving neural network embedded as a proof of concept ballooning to 5GB

One of my cats has taken to hanging out in my office more while I work and it's only a little bit distracting.

@SecureOwl Snap! That's a lot cleaner/more simple than I expected. I assume you extracted the auth tokens from user sessions?

Do you have a site that is abandoned or difficult to maintain? 🏚

@sutty can help you recover it and turn it into a preserved site, without exposed databases and panels, greatly reducing the risk of attacks and maintenance work. ❇️

Most importantly, thanks to the integrated @distributedpress support, the renovated website can be automatically published on the Decentralized Web!

ℹ️ More info → sutty.coop.ar/en/services/pres

@SecureOwl That's cool! Is the code published somewhere I could read? I'd be interested in setting one up too.

Anyone got experience of screen readers on fedi? I can't really see since to eye surgery I want to keep up with you all and have some mental stimulation.

*unreal tournament announcer voice* M-M-M-MULTICAST

Whenever my screen reader hits a DNA sequence it sounds just like my cat chirping at a bird.

Show thread

Love having my screen reader freak out repeating strings of ones and zeros or a string of DNA and just get to zone out for a minute trying to find any sort of pattern in the speech.

Software that modifies itself using human language is kinda nest tbh.

are you a full-stack dev with broad, modern JavaScript experience? there’s currently a rare opportunity to join my team and help us keep improving our trust & safety systems 👀 backend stack is MongoDB/Express, frontend is React/RTK jobs.lever.co/vrchat/cda69202-

Social media for the blind:
Bluesky, hardly any Alt Text. To understand an image, you must open the post, tap on the image and hope your screen reader can describe it, or spend a minute or so scanning it.
X, hardly any image descriptions. The built-in AI will try to describe the images which is a major improvement, but you need to wait until the screen reader finishes reading all the info about the post.
Mastodon, loads of image descriptions. What a pleasure it is to browse my timeline here.

Machine learning is one thing. The trouble is in getting it to forget.

Believe it or not, this isn't #Meshtastic or #Meshcore

It's #Reticulum :blobcattoot:
That's right, Reticulum is finally running on microcontrollers and there's so many ways to do it:

github.com/attermann/microReti (C++ implementation of Reticulum; finally getting close to 100% feature complete)
github.com/ratspeak/ratdeck (Pictured T-Deck firmware)
github.com/torlando-tech/pyxis (T-Deck firmware from the maker of Columba)

Introducing Rook: Sovereign Intelligence for the Solidarity Economy and Beyond 

Projects move at varying speeds: some shift rapidly from ideation to product while others take a slower and more thoughtful approach. For Rook, Hypha’s new agentic knowledge management tool, we’ve been deliberately considered about how and why we move into the AI space. This is not only because the market is a little saturated with LLM-esque tools, but also because we wanted to offer something that could be useful to organizations working within and serving the solidarity economy, as well as arts and cultural institutions, and organizations who value digital sovereignty and data privacy. Emphasizing care and consideration might be anathema to the typical AI product release cycle, but these values, along with quality, are what make Rook different.

Rook is powered by your organization’s data. You give it access to the files and folders you want it to query, and it then returns responses based on that data. Nothing is ever trained on your data, and the infrastructure is configured to be privacy-preserving. Our demo, which you can see here, gives you an idea of how Rook could work with your organization’s knowledge base.

At Hypha, we’ve been using this AI knowledge management tool internally for over a year, building it out under the open source project known as RooLLM (named after our digital office pet). And while much of this work will remain open source, for organizations lacking an internal IT department, we wanted to make the tools accessible as a product that Hypha can customize and deploy for interested organizations.

In keeping with our thoughtful approach, today’s announcement is more of a soft launch, as we’re looking for partners to help us make Rook the first choice tool for privacy-conscious social enterprises, cooperatives, cultural organizations and progressive thinktanks. Hypha is pleased to announce that we are already collaborating with jason wiener | p.c., the preeminent law firm for mission-driven companies, social impact, and environmental enterprise. Based in Boulder, Colorado, Jason and his team have already provided valuable feedback on Rook’s response quality, and on data privacy considerations.

We are actively seeking organizations interested in supporting Rook as a key piece of the solidarity stack. Get in touch if this resonates with you.

@bluemoon And if people claim that public transportation loses money, the solution is simple, stop charging people to use public transportation.

99.9 percent of all roads are free for cars to drive on them, no one talks about how the road loses money.

Show older
Mauvestodon

Escape ship from centralized social media run by Mauve.