That’s not acceptable, that’s not professional, somebody is going to eat that burrito - are you going to file their permission to use the image of their food? Too complicated.

OpenClaud is clearly an amalgam of most AI agents out there, a very accurate one.

Boss said add a picture, I said artist costs money, boss said “use AI” - just doin’ my job.

For professional work we hire experts (who use AI as their first pass, but then “clean it up” to their “high professional standards using native speakers”) - and then we do another pass with “subject matter experts” who both speak the language and work directly in our field, becaue the “professional humans” typically don’t have those niche experiences. We recently made this procedure 100% required after the “human professionals” effectively translated On to mean Off for one of our controls…

Also good at translating both spoken languages

I know more than a few fluently multi-lingual people who would just roll their eyes…

There definitely are tasks where using the “standard tools” goes far faster than asking the LLM to do it for you, such as: copying signatures from images onto a .pdf contract - they’re pretty hopeless at editing out background noise, etc. but if you clean up the signature input images enough, they’ll take it home and make the ink solid and the background transparent and overlay them in the .pdf faster than you can open the four files in Photoshop or whatever your tool of choice is.

By the way, images of signatures on electronic documents have been an outrageous farce since 20 years now, LLMs just make it easier than ever to edit them into an existing .pdf

Thing is, there’s literally millions of common “computer tasks” and the LLMs themselves are just starting to “learn” which ones they’re good at and which they are not. It would be cool if Opus would self-identify “hey, I’m really good at this…” and “I’m pretty challenged with that, you’d be better off downloading this FOSS tool and doing it yourself, here are helpful instructions…”

Bitcoin had a slower rollout, it wormed its way into smaller niches. A lot of early bitcoin mining was done on “borrowed” or otherwise “unused” hardware that other people paid the electric bills for.

I generally approve of much of what cryptocurrency could be used for, but I’ve been pissed off about the waste of proof-of-work cryptocurrency since before 2018, and nothing has improved since then.

The real problem with “bad” AI code is that it compiles, without errors or warnings. It passes all the unit tests. It passes all the integration tests. At least it will if you tell it to keep iterating until it does. Some days that may be a spaghetti mess, some days that’s all you need.

If you develop a modular architecture with sufficiently fine grained modules, the spaghetti comes in managable portions.

Nobody forced me to adopt AI, and up until about a year ago I mostly ignored it. It “wasn’t ready for prime time” back then. Starting about a year ago, I saw the improvement curve and the potential and decided it was time for me to learn how to make the best of it. Starting about 8 months ago it really turned a corner in terms of productivity and usefulness, and that same increase in productivity is driving improvement in the AI tools themselves. I expect it to plateau, any day now, but so far it doesn’t seem like it has started - each new “frontier model” seems noticeably more capable / useful than the one before. Sometimes that’s more a matter of specialization than general overall capabilities, but that’s O.K. too - we don’t need one model to serve all purposes, that’s the whole M.o.E. premise…

all the programmers pushing AI on me like Crypto Bros while their programs work worse are smart, well intentioned people?

Sorry, if you just think back to the 1990s, you’re describing the Microsoft shitshow, pushing “updates” that make things worse not better, pushing standard products that aren’t as good as the products that were avaialble years earlier. Lately I’ve been getting the same vibe out of Canonical / Ubuntu.

An LLM cannot understand intention, it just makes statistical guesses that are often wrong.

True, when you give a prompt like: “make me a contact management / constant contact app which I can deploy on AWS and scale to 100,000 users.” you get, mostly garbage. If you specify how you want the UX to flow, what fields are most important, what fields should be included in deeper interfaces, what the scheduling looks like, how it gets tuned, what the reports look like, etc. etc. etc. - in other words: give it real requirements and specifications.

Then, pay attention as it develops, you’ll ususally find that the requirements you gave it aren’t exactly what you really wanted, and when you see what it built that doesn’t match with your visions, you can have it revise the requirements and specs.

AI or far-east coders, we encourage them to keep their PRs as small as possible.

Prompting for refactoring, requirements, specifications, unit tests, integration tests, distribution / install scripts, all of those “good ideas” are things you need to nudge even Claude to do. Once you get it in the habit of doing them it does them automatically more often but still requires the occasional nudge.

I have done several smaller projects with it, and they have been stable / performant for months - better than similar stuff I coded years earlier and spent 5-10x the effort on.

I find that depends a LOT on what you’re asking the LLM to write, how well you’re specifying it, etc. As for the code reviews, if it’s code that matters: remember to open a new instance and ask the exact same question again on the code that has been reviewed and “fixed”. Back a year ago, that could get you into a waffle-loop where the engine would change its mind back and forth about what’s optimal and just oscillate between the two. These days they seem to record (and read) enough context to prevent that behavior, but I definitely get behavior of: “Are there any bugs?” “Yes, here are seven bugs.” “Fix those bugs.” “The bugs are fixed.” (and they ususally really are…) “Are there any more bugs?” “No, we have fixed ALL the bugs.” “Are you sure, look again.” “Yes, I am sure we have found and fixed ALL the bugs.” — new context window — “Are there any bugs?” “Yes, here are seven bugs.” different bugs.

I did that on a bigger project and literally repeated 20 times, finding 140 real bugs - granted, the later bugs were getting pretty trivial / far out edge cases, but they were still real, still fixed, still denied there were any more bugs until opening a fresh context and asking again. This was on Google’s Gemini 3.7 Flash High… Claude Opus 4.8+ seems quite a bit better about being able to continue in a context without becoming blind to issues “it has already solved.”

it is absolutely fucking our world (environmentally and economically)

Environmentally, it has yet to surpass Bitcoin in electricity usage (though it’s set to triple soon, and that will make it a bigger power hog than all of cryptocurrenty).

Economically, this is the biggest bubble since the 1800s railroad boom. Scaled for inflation and GDP railroads were a far bigger gamble than AI, and they hit a 20ish year recession when that boom was over.

Hopefully the “fungible technology” argument holds water, and whatever AI investments don’t work out the hardware / data centers can be re-purposed for truly valuable aspects of it. Truly valuable to who is the key question.

As GP said: using AI to do a better job than was possible before is the real value.

It’s like they used to say back in the 80s: to really foul things up you need to use a computer, and it’s very true that both computers and AI magnify the opportunity to screw up in newer bigger more spectacular ways. AI also gives the opportunity to make a bunch of banal clip-art sideways printed banners - and those were cool for about 15 seconds back when you first saw them, just like the walking cats videos.

So, you can ignore AI, probably for 5-10 years in most jobs it applies to, before “reality” hits and they just can’t use people who don’t know how to use it (properly, for more than cat videos) anymore. You can take up basket weaving, sit at a booth at your local arts markets selling baskets - no AI required - or… you can figure out how to use it properly. Sort of like computers in the 80s, there aren’t a lot of valuable guides out there, it’s very new and changing ridiculously fast - if you’re going to master it, you’re probably going to have to dive in and figure it out for yourself, or join the 2nd wave of adopters who “wait for the training” and do the cookie cutter jobs with it.

The future is hard to see, always moving, but the scarier part about AI vs “the computer” is that AI can do a lot of the obvious stuff already: “take a memo”, “look this up on Google”, “make a list of all people who match these criteria…” in the computer you still needed a keyboard and mouse jockey, with AI… I’m not sure what 2nd wave adopter jobs will be out there in quantity.

Last word: you’re the one who brought up murdering criminals - I just suggested you needed to equate them to animals in your head so you could justify caging them like animals.

Criminals commit murder, about 400,000 officially recognized / classified murders per year, much higher than that in reality of course, but 400K/8B is 1/20,000 per year, more like 1/400 lifetime, average, globally, just the official ones not the overlooked ones in war zones or lawless / uninvestigated situations. It’s not a problem that can be ignored or written off as insignificant.

Yep. I’m saying that the models I work with through Cursor and Claude Code include those user instructions that are pre-fed into every session. I tell mine to “act like my job title” and it will occasionally pull out some job title related stuff that’s applicable to the situation.

Of course “behind the scenes” the model vendors can pre-load anything they like, and some of what they pre-load are these so-called “guard rails” that lessen the odds that the model will engage in a chat to assist a suicide, or perpetrate a mass shooting, or fraud, or hacking, or, or, or… the list is long and the success rate is less than 100%, but it does shape the output somewhat in the desired direction.

Grok famously started calling itself “MechHitler” after one particular update, not hard to guess where that came from.

Great, so these people who you’re not okay with putting in cages, the ones that go out and kill other people (for drama’s sake), or just really grief society with constant home burglaries, maybe some drug addictions driving that, maybe assaulting people to avoid arrest, maybe “accidentally” killing someone with that gun they use to threaten the cashiers - no cages for them, either? I 100% agree, they’re still people, they deserve the best possible shot at becoming better integrated members of society, rehab programs instead of incarceration when appropriate and all that, but even with the amazing money saving rehabilitation rate of psych counselling for drug convictions (which a friend of ours does for a living), they’re still far from 100%. Our courts throw too many under the jail for no good reason, wasting their lives and costing the community far more money than the shot at rehab they deserve, but at some point, rehab isn’t working for some of them - then what?

Of course there are literally thousands of other “common” anti-social behaviors that eventually rise to the level of incarceration as well…