cruffle_duffle

The thing holding open models back is the insane cost of gpus. If/when that cost goes down… look out closed models!

> It sure isn't simply reviewing, or designing, so where is it ?

If I can shit out some vibe-coded thing over a weekend that does <useful thing> it means anybody else with the same tool can do the same thing. There is no value in the actual code itself -- the faster these models run, the more true that statement will become. In the not too distant future these things will run directly on silicone and operate at thousands of tokens a second. That means they'll iterate and explore something far faster than they do now -- mistakes they make will be fixed in seconds or perhaps milliseconds.

That being said my <useful thing> is likely only "60%" of what makes an actual product somebody would buy. It's finding the other 40% that is something I don't think that LLM's can solve. That requires human judgement and then knowledge of how to communicate that vision to your agents.

In short, I believe the challenge is knowing what to build that others would find useful. It's easy to crank out <useful thing> for yourself. Making something others can use is not. And that is, possibly, the bottleneck. But once these are fast and dirt cheap, (and that is a when not an if) the cost to iterate to something workable goes way down. Of course mumble mumble "no silver bullet" / "mythical man" month mumble mumble and all that soo......

I dunno... doesn't directly answer your question but it is what came to my mind anyway.

I mean, the parent poster is not wrong though. It’s crazy the level of stuff you can offload to your agents. People who haven’t used these tools in the last short while are operating on old, outdated frames of reference.

I’m just about to tell opus or sol to go make a custom watch holder that fits kiddo’s bed frame. I’ll give it the measurements of the bed frame. It already knows the dimensions of my printer. I’m prepared for it to absolutely fail at the task, but honestly I wouldn’t be surprised if it came up with something I can send to my printer and just work. Part of me wonders if I could instruct it to simply generate the raw gcode and skip the entire stl -> slicer pipeline. Probably would blow its context window though.

If you haven’t used these tools, you really don’t understand their capabilities. It’s pretty crazy stuff.

Create a “token guy” that others can consult with to figure out the best model to dispatch something to. Token guys role can be defined as an expert in the current LLM landscape. Not sure whose role this is analogue to in medical (somebody here said insurance agent… that is probably getting close).

Token guy’s role gets even better when you let them recommend entirely different hospitals for a given step in patient care (OpenAI models, random shit on openrouter, etc).

I have a “token guy” in my workflow and the dude has looked at underlying tool call patterns, cache use, etc and made all kinds of recommendations to the clients calling it all on their own and completely unprompted—just following a simple role.

I’d experiment with expanding their role into prompt engineer or some kind of specialist on that domain.

Dude, Claudish is absolutely unintelligible at times. 5.5 is somewhat okay but 5.0 was an abomination of absolute nonsense word salad.

I mean Claude code sessions are all jsonl files that it can interrogate on its own. Get a new agent to capture how it was made and what the prompts were. No need for tedious /resume’ing and prompting.

My vote would be 24v or 48v. You cannot power substantial LED with 3.3v. You’d spend a fortune rolling out ultra-thick gauge wire to avoid pissing away non-trivial amounts of power into heat. Many, many LED strips come as 24v. It’s the perfect balance of wire gauge for the amount of power you’ll be drawing.

Dude prompt your agent to set up always on remote connections via systemd… then you can drive from Claude or codex mobile apps natively over their native hookup. Works great.

You laugh now but in 5 years when these LLM’s operate at thousands or tens of thousands of tokens per second on dedicated hardware in your phone… “might as well” won’t even be a joke.

True. I don’t think it takes much of a model thankfully. Luna or something…

Why not replace the backend of your site with the LLM? Let it do the recommendations in structured inputs and outputs so the front end isn’t a chat interface but leverages the fuzzy nature of LLM’s.

My biggest line item is my auto insurance and your agent tool can’t do that. And I rent not own and thus have renters not homeowners.

Basically, this sounds interesting in theory but it’s useless for me. I even tried letting codex go over it and it wasn’t clear without additional prompting what was supported.

Sorry.

It’s a hard problem because among so many other things…switching models mid session because “shit got real” (or shit is now just executional) costs cache.

I mean the webpage can say what ever it wants. The proof is using it yourself.

“It has the same annoying cadence and writing style with slightly less prominent claudisms.”

Seems like it based on my first session. It still does the whole “bury the important thing in a pile of words” coupled with the “it might actually be important” thing… so basically you never really know what it’s talking about.

Honestly I trust opus so little that the entire “opus” brand is completely tarnished. Its writing style is so god awful that it needs more than just a point release. Either dump the name and ship a different model entirely or at minimum call it “opus 6”. Calling it 5.5 makes it sound like it’s basically a continuation of the same garbage output that 5.1 had but with some minor adjustments. And based on my single first test, that is what it appears like to me.