Yeah, LLMs are useful for delegating work that is labor intensive and bullshit in nature. I bet Peter from Office Space would have no problems with his TPS reports if a LLM were helping out.

In theory, LLMs can help people do better quality work faster, but the productivity boost is modest when quality work is the goal. However, what seems to be happening to a great extent is that people are putting in a lot less effort and employers are now paying for salaries and AI tokens while getting worse quality work and slower progress in return.

And yeah, anyone who actually is trying to do a good job will get pissed at their co-workers half-assing it and pumping out low-effort AI slop while pawning work off on them. I think the only sustainable path forward with AI is instead of token usage being a badge of honor, there’s a culture of shame around AI where people put in a lot of effort to make sure nobody can tell they used AI out of fear of ridicule.

deleted by creator

Believing these apps will succeed is based on an assumption along the lines of LLMs being able to produce quality code at 100x the speed of a competent human engineer. Maybe they can produce slop that fast, but using LLMs in a disciplined fashion to produce a quality product usually means LLMs are only providing a modest productivity boost in my experience.

It’s not realistic that these apps are anything but slop given the timeframe and scope spanning multiple complex apps that each took a decade or more to develop.

In my 20+ years of experience as a software engineer, I’ve never seen a successful project that hastily pumped out a massive pile of buggy code and then went back and fixed all of the problems later without doing a complete rewrite. You have to build incrementally and bake in quality every step of the way. LLMs may make the process a little faster, but the same rules still apply.

A poorly built codebase will always eventually collapse under its own weight. What I anticipate is that they’ll play whack-a-mole for a while and then give up. If they were targeting a small subset of one app’s features over a timeframe of months, I might believe it, but several massive apps all at once and it will all be done weeks from now? Stop bullshitting yourselves.

When you first signed up for Proton would’ve been the best time to use your own custom domain instead of theirs. Today is the second best time.

I don’t suppose buying AMD is really much of a protest vote against Nvidia. I also see the 7900 XTX is going for $1800 whereas it was available for a little over half that a few months ago. 😖

In all fairness, I haven’t investigated it thoroughly myself, and my skepticism mainly comes from the proposed timeframe. I’d be less skeptical if they said they were going to focus only on, say, Photoshop and spend a few months reproducing a small subset of its features without decompiling the code while creating a public audit trail of their clean room process that a lawyer helped them create.

I use LLMs for coding every day at work and also use local LLMs at home. From my experience, LLMs can be used to create a better end product than humans alone if the proper discipline is followed.

However, my impression is that ArtCraft is a bunch of vibe coded slop where the goal is to reproduce several apps as quickly as possible by letting the LLM go wild. In that case, you usually end up with something that seems to work in many respects, but eventually collapses under its own weight.

We will see whether ArtCraft ever gets out of alpha and shapes up to be solid, but as someone who has been a software engineer over 20 years, hastily creating a mountain of code and then expecting to come back and fix all the problems later never works. It’s always better to work incrementally and add one feature at a time while building in quality every step of the way.

What I expect is that they will continue adding features and trying to fix the problems, play an extended game of whack-a-mole, then eventually get frustrated and abandon the project.

I detest companies like Adobe and I’d be delighted to have my skepticism be proven wrong and see them end up with replacements for the whole suite that work well and are on solid ground legally. I’d also feel better about this if they were using local models, since spending a fortune in tokens to support a company like Anthropic is certainly not on higher moral ground than supporting Adobe with subscription fees.

Yeah, as much as I’d like to applaud this sort of thing, ArtCraft, for example, is going to end up with some major legal woes, in addition to the fact that it’s a pile of slop that doesn’t work worth a damn. The only winner is Anthropic or OpenAI getting big money for all those wasted tokens.

were meant to make the viewer feel extremely uncomfortable and realize how racist the character was being

The Diversity Day episode of the office comes to mind as something matching this exact description that has been excluded on streaming platforms.

Depending on the use case, this is solid advice. I’d suggest checking on eBay for older, used machines that that are still pretty powerful compared to a Pi and are also quite cheap.

It’s funny the Democrats won on the promise to clamp down on big tech given their history. Next election cycle, I bet the Republicans will be running on regulating the fossil fuel industry.

I suppose the reality is that we can’t fully trust any open weight or closed model. It’s worth checking the code for any open source harness or runner you’re using with open weight models for any data collection or other potential problems and being careful about what permissions agents have.

Of course, LLMs can write bad, insecure code in general, so it’s always unwise to trust the output of any model in production without thoroughly vetting it.

That being said, I trust Qwen 3.8 27B running in llama.cpp with OpenCode more than I trust anything from Anthropic or OpenAI, mainly because I’ve seen the code for llama.cpp and a OpenCode and I’ve never seen Qwen do anything suspicious, whereas I know Anthropic and OpenAI would be collecting my data and training their models on it, along with who knows what else.

Even if it did work, Adobe’s legal team will obliterate them. I’m also not sure about the legality of using it for any paid work. It’s a moot point, though, since it’s slop, and better off using a cracked version that at least works.

Jira’s AI is mainly useful for asking how to use their shitty UI, like where the hell did that button get moved because I can’t find it anymore.

Eh, these articles are basically, “well-known gobshite once again spewing shite from his gob. Everyone be outraged.”

Nice try with “Sam Altman Crows” instead of the usual “Sam Altman Says”, but I still didn’t read it.

Just cut down on the avacado toast.

Exactly. The only people willing to pay premium prices for privacy instead of surrendering their data to surveillance capitalists are people with a low tolerance for bullshit.

deleted by creator