Indeed: The industry optimizes for speed, time to market, and features, and applies the ostrich model to everything that doesn't bring short-time revenue (security considerations, accessibility, vendor lock-in, interoperability, …)

This has been going on for as long as the industry exists, and now we start to have the proper tools to assess the damage and understand the brittleness of it all.

We've tried to take a first-principles approach to our end goal of 'legal superintelligence' that has involved identifying the areas of our domain most in need of improvement and releasing models that raise the bar on quality in those areas.

We've been around for a couple months now and ended up starting with retrieval and enrichment. The models we've released to tackle those problems have indeed been smaller in size than their competitors, yet they still rank ahead on open-source benchmarks.

Them being so small also helps with their accessibility — as I mention in our post, our models can be deployed on ordinary hardware, not a supercomputer.

Next on our roadmap is reasoning and research, which will require more infrastructure to support, but again, we aim to be judged by performance at the time of release.

> As much as I agree with the message, this reads like marketing copy trying to make a big deal out of a tiny model being hosted privately.

The point of this point is really just to reaffirm our commitment to sovereignty and accessibility and contrast our approach with that of major AI labs. It _is_ possible to commercialize LLMs while still keeping them accessible. A customer using a self-hosted deployment today does not need to worry about our models no longer being available tomorrow. We think that's a good thing. And moving forward, we want to keep that option available for anything we do, instead of trying to pull up the ladder while we're ahead.

Come to think of it, another interpretation of "a love-letter to text" would be designs catering to text-mode browsers (e.g. Links, Lynx) which could also overlap with accessibility for the blind.

The broader discussion about AI and model capabilities died a couple of years ago precisely because it's so underwhelming now. People did adapt. Startups stopped hiring just to get to MVP. Coding sweatshops had huge layoffs and stopped overhiring. The corporate world got better tools for collaborations and meetings. Accessibility tools are still bad, but improving. I would argue that the a11y topic is still very ripe to be the next big thing as it continues to converge with better UI/UX instead of being an afterthought.

The layperson and tech professional alike otherwise agreed that this is a vehicle for blame game, grift, disinformation, etc. This is where all the pushback is and the topic at hand. People aren't dumb. The only people worried about "AI" are the ones who bet too big on it.

Now generalize this. Every tiny community, every person with a disability, everybody for whom the default software doesn't work right? Can now change it specifically for them. Not add peperoni, that's far too low-dimensional to capture what is happening. Actually build their own interface, be able to use something they simply didn't have access to before, and critically not depend on another programmer (there are like a dozen of us blind ones!) to build something for them.

About a year ago I remarked to people that despite all my attempts to make data more programmatically accessible, the most effective way for AI to interact with a modern computer is to use the built-in accessibility interfaces driving actual desktops with full applications. IE, the best API for an AI is the UI (mainly because that's what most humans use).

I'm also reminded by an observation of the late Robert K. Merton, on latent vs. manifest functions. Originally coined in the context of sociology, but far more broadly applicable. In discussing these, Merton makes the perceptive observation that because latent functions are not immediately apparent, obvious, or significant, they represent a greater increment of knowledge and understanding than manifest functions, which are obvious, evident, easily understood and communicated, etc.

Popular works, or opinions, tend to be more accessible, yes. But they are also frequently a lower increment of knowledge or utility.

I too am pained by book and other information collections which pander to easy accessibility at a cost to insight and significance. That isn't to say that libraries should discount popularity at all, but I cringe when it seems to be the primary consideration.

By extension, other mass-context systems (markets, mass media, etc.) also tend toward minimum viable standards (often mis-stated as "least common denominator", problematic in several ways), and discount both long-term (non-obvious, non-apparent) benefits and costs.

You absolutely don't need LLMs for that.

Its the very description of most corporate JavaScript developers, and probably most Java developers. I say that as somebody who wrote corporate JavaScript full time from 2008-2023. Most of these people had no idea what they are doing. They could throw something together using their favorite abstraction library/framework but then struggled to maintain it. If there were performance or accessibility problems that came up there were only three outputs: hostility, crying, or starting over from scratch. The insecurity was real. You can still see it today. As an experiment take React away and note the response.

I do disagree with you vehemently about Seattle’s light rail being inferior to Portland’s. That may have been true 10 years ago, but it for sure is not true today, especially after the East link opened earlier this year.

That said, Your parent is wildly off the mark (and honestly quite reactionary) in describing harassment from fentanyl addicts. Such harassment is extremely rare, and in the few cases where it does happen it is a failure of public health policy, not transit and accessibility policy.

If you need to put a camera on glasses for a legitimate reason, such as a device purely for accessibility, then you should be able to get an exception, of course.

I rarely go to a library to loan a specific work - I go there to find a work. This means going through dozens of potentially-relevant titles, taking them off the shelf, quickly browsing through them, and taking the one or two best ones home. This entire workflow becomes impossible if the book isn't readily available.

A book hidden in a box in the basement, or which arrives after only a few days, might as well not exist at all. I'm simply not going to scroll through a list, order several dozen books solely by their title alone, and come back a few days later (if this is even allowed at all): it's just not worth my time.

The whole "we keep a copy in a central archive" approach only works for historical purposes, not for actually making it available for reading. If you do that you have to also make digital scans trivially available for browsing - and in practice that rarely happens!

Given the low adherence to accessibility standards e.g. in academic publishing [3], LLM parsing needs creating a commercial incentive for comparable structured access would be marvelous.

[1] https://www.section508.gov/create/pdfs/common-tags-and-usage...

[2] https://pdfa.org/resource/tagged-pdf-best-practice-guide-syn...

[3] https://arxiv.org/html/2410.03022v1