Comment by Philip-J-Fry
13 hours ago
Feels better to me in the sense that it doesn't try to be too clever. When I ask it to do things it actually generates code that a developer would write. That still lets me build software in layers like I do mentally. And it keeps you in control of where abstractions are.
Some people would say that you don't need this at all with the frontier llms. But when extending existing software or writing software that fits into an existing ecosystem, I find I still need to offer a lot of guidance. And it's better that the context window isn't flooded with code that was reverted after I looked at it.
> Some people would say that you don't need this at all with the frontier LLMs
Hit the nail on the head. Development methodologies pivoting at the whim of frontier models have been one of the things that have been worrying me as of late.
it doesn't worry me too much tbh. This is really just cutting corners and going this route leaves you with some kind of additive ship of theseus piece of software you don't understand and require your frontier model to work with, which just... becomes more additive and gross. right now we're in some proverbial heyday of everyone seeing how absolutely grand of a thing they can make with frontier models. over time i expect some old design philosophies to take hold with people who use LLM's successfully and find a place again paired with developers who understand why such layered methodologies are important... because this whole AI thing wasn't the silver bullet CEO's and crazy execs thought it was. They will eventually get sick of the support burden of badly designed stuff. i anticipate a bit of pendulum swinging back towards "still using AI, but we need smart people to do it because it's expensive and we want it done right". Right now a lot of companies are being extremely reckless imo and they haven't fully realized/felt the cost implications of trainwreck software yet.
What if those “cost implications” are just asking Opus 6.7 to fix whatever issue cropped up?
4 replies →
> haven't fully realized/felt the cost implications of trainwreck software yet
As long as the API is private or not dependent on by automated systems and you have good test coverage, you can get away with bad code IMO since you can always refactor. If it looks stupid but it works (and can be refactored/fixed/improved later if needed) it aint stupid.
1 reply →
IMO the fundamental principles and "conflicts" are still the same, but many have forgotten about them with all the hype and sense of speed. The industry will slowly rediscover them.
Support burden? When did companies care about customer support or system maintenance?
They already did the bare minimum allocation to both despite engineering screaming at them.
I highly doubt executives are going to accept slower release and sale cycles in order to provide better support.
If anything, I expect this trend to accelerate. I’m imagining the opposite, a plethora of AI slop with increasingly burned-out engineers.
2 replies →
I've been using layered development techniques like the parent says even with frontier models, just one agent at a time with Zed.
I just use a suite of tests to enforce the mechanical standards (dry, encapsulation, etc) and arch documents for shape + reviews of new work against the two. I’m not seeing any significant drift.
Yea, this is one thing that gets me. Vibe coding and freestyle prompting gets wildly different results. Starting from a set of specs and expectations routinely gets me reviewable and we'll tested code. Sure, running the full test suite alot slows things down and uses more tokens, but when I hit a five hour or weekly window, I just roll up my sleeves and start working without the model because the code is layed out the way I expected and is designed to fit my meatbag workflow and pace.
You must not be looking very closely yourself then.
1 reply →
> Feels better to me in the sense that it doesn't try to be too clever. When I ask it to do things it actually generates code that a developer would write. That still lets me build software in layers like I do mentally. And it keeps you in control of where abstractions are.
yes!So it's not just me. In terms of how intelligent they want to be OpenAI is at the top, then Claude. And actually when I was playing with Sol it meant that many tasks it would just try to do something else because it thought I don't know what I want. Even trying to nudge it back in the right direction wouldn't work.
I guess the models just have the cultures of their creatures.
> it doesn't try to be too clever
DS is miles better than GLM in this regard. I wanted to add that as PC is planning to go to GLM next.
Interesting. I’ve been using GLM 5.3 Flash for a couple months and found it to be great. What has it been doing that you found to be “too clever”? It sometimes gets stuck in loops. I once asked it to do a code review of a simple pattern and it went off on a 45 odyssey trying to analyze too much. But other than those, it’s been a pretty good daily workhorse. And you can’t beat the price. Even Deepseek 4.1 Flash is more expensive. And I switched to GLM 5.3 Flash from DS 4 Flash, my previous daily driver, which I also found quite good. Anyway, my point isn’t to start a “which model is better” subthread, but was just curious what behavior you’ve seen where it’s “too clever.”
Anecdotal but I was doing regular adversarial reviews of new work using about 5 models (mostly the open Chinese ones like DS, GLM, Qwen, Kimi, etc) and GLM routinely took much longer for similar results and the same instructions.
It would just loop on things and go down rabbit holes more. DS flash has been so consistently good I've kinda settled on it for 90% of my work.
I still use adversarial review for bigger changes where I want to make sure any issues are caught early, but I genuinely feel as tho if no new models ever get made I'd be fine with just using DS forever.
1 reply →
It wastes my time. A lot! In different ways. That's the tl;dr. Really.
Keeps repeating things I settled as accepted or rejected. But what really gets on my nerves? The language it uses – that aweful, useless typical LLM languages! No AGENTS.md instruction, no per session instruction, no just two-prompts-prior instruction seems to work, hell sometimes instruction regarding that in same prompt doesn't work. GOOD/BAD example along with instruction don't work. That wastes time and zones you out! Not to mention the fatigue of to and fro. DS has been working much better. Hell Gemini 3.1 Pro works so much better than GLM. I don't have extended experience with OpenAI and Anthromic models but the free claide.ai works quite well and better than GLM 5.3 (not even flash). Albeit I must say I rarely use claude.ai for coding, but technical tasks? All the time!
GLM's best response: "No excuse. My fault. You exactly asked me not to do exactly that and so did the AGENTS.md. I failed". Love it. Except that if I press I will get that response every 5 prompts.
1 reply →
For me GLM thinks too much for no reason. It takes a 2-3 turns for DS where for GLM it takes up to 10.
I like giving precise instructions,”update function xx to support input yy, add tests”. There is no reason for 10 turns and 5 minutes of waiting.
I briefly tried GLM 5.3 Flash and the first thing I noticed was how slow it was followed quickly by its single-mindedness. With DS 4.1 Flash I find that it understands that not all the detail in my prompt is important. Sometimes I like to give a quick example of what I'm looking for and it can take the hint whereas GLM will seemingly reshape its worldview around this tiny detail.
DS just handles ambiguity and my bursts of detail with much more grace. I'm only using "max effort" and the CoT is reasonable compared to GLM.
> the first thing I noticed was how slow it
This really depends on the provider, it's a very fast model with ours
1 reply →
What harness are you using?