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.
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.
In a way, that’s the situation today. You can pay an experienced developer $$$$ to do it right, or you can outsource it and pay $$ but pay extra for more layers of review, have lots of unfixed bugs and more tech debt.
You are proposing that instead we pay the frontier model to clean it up. If it can, I expect it to cost somewhere in the neighborhood of an experienced dev because that’s the alternative. That’s the way the frontier AI companies will extracting the value of the work from you.
But there's two ways this could be true: Frontier models costs rise, or (more likely) experience dev costs fall
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.
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.
Sales are out of their control.
If there’s a market for quality and you don’t have it, sales are going down no matter what the PHB wants.
If there’s not, watch out for the vibe-coded competitor.
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.
DS is miles better than GLM in this regard. I wanted to add that as PC is planning to go to GLM next.
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.
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.
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.
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.
This really depends on the provider, it's a very fast model with ours
Wait... you just spent 3 paragraphs singing the praises of DS... and then you say you're going to subscribe to the one that makes GLM. ?
(I do tend to agree with you, though I tend to favor GLM-5.3-flash over DS 4.1 flash.)
debugging reasoning traces more effectively seems useful to find prompt issues writing custom harnesses even if it runs on stronger model later
It doesn’t have to be better, it just had to be nearly equivalent. I’ve been doing a lot of vulnerability work lately (trying to get ahead of the bots finding exploits in our code) and the frontier models feel like a chaperone at a church dance: get too close and they show up with the ruler and enforce the safety margin. Using deepseek and glm and qwen, I’ve gotten nearly equivalent results without the constraints. I still use frontier models for things but outside of a company environment, I’m happy to spend 1/10 the money, get the same results, and not have try and trick the AI into working.
Without getting too much into politics; over the past several decades, one country is going around starting wars, regime changes and other interventionism, or initiating economic warfare (Huawei sanctions because they were getting too competitive to US companies; GPU restrictions; tariffs; etc), and that's not one administration: it's US policy over the past few decades.
It's kinda like meeting someone on the street who's protesting about animal cruelty in slaughterhouses, slyly handing them a chart showing them the nutrition facts of beef, and then walking off without another word as though you've settled your case.
Like, I get the point you're trying to make, but I hope you're aware it's a non-sequitur in context?
Your hope should have been that amelius was not aware that it's a non-sequitur, because that is correctable. But then they kept posting and removed all doubt.
People like to bring up the Uyghurs immediately, but never really have the same smoke for the US. Also, do we include the prison population in human rights? Do we include access to healthcare? What about systems perpetuating poverty like lack of access to public transportation?
To pre-emptively be clear on my stance or potential bias, I stand by no country in the world. I was born in Iran, and my parents left when I was young to the US because they believed in the American Dream. The irony being that the dominos of why they left were initiated by the US and UK. And long before Trump was elected, but certainly immediately afterwards, I decided to leave the US once I had the means. (Also because of the lack of public transit, the $100k medical bills after an accident, being punched by a homeless person, and having my possessions stolen repeatedly and the cops giving me a :shrug:). I moved to Japan, where I experienced blatant racism for the first time. I moved to Hong Kong (and tbh no real complaints so far), but I am technically subject to the freedom of speech constraints on China.
I have no allegiance to any country or specific people. My allegiance is to all living beings.
And I don't care if they're reputable, I will look at the methodology and make my own conclusion either way. Appeals to authority are not a substitute for critical thinking.
No lessons to take from the US.
PS. https://en.wikipedia.org/wiki/Foreign_interventions_by_the_U...
I know the US thinks only its citizens are human, so "The United States has been involved in hundreds of interventions in foreign countries throughout its history, engaging in nearly 500 military interventions between 1776 and 2026, with half of these operations occurring since 1950 and over 25% occurring in the post-Cold War period” is not probably part of the reputable storytelling
If they actually said something, it would be much less effective.
> We're not talking about a few % difference here.
How may freedom percent is it again? Do you lose 1% or 20% for failing to allow foreign ownership of utilities? How many percent do you gain for stopping Russian misinformation?
I'm not going to discuss it further if you don't have anything substantial to bring to the table.
edit: wording
Yes, people have put thought into this. This is why I trust institutions and not random posters on the internet.
Just look at Russia - it's score is highest in the 90s - the most corrupt and destitute time in its modern history, also happens to be when it was deemed somewhat friendly to US and friends.
https://www.prisonstudies.org/world-prison-data/highest-lowe...
https://www.theguardian.com/us-news/2022/sep/27/slavery-loop...
In Arkansas, agricultural work accounted for 17% of all prison labor assignments in 2021. https://www.prisonlegalnews.org/media/publications/Captive_L...
17% of prison assignments might be 3 assignments for all I know.
I found a number in your document: "...1.2 million people incarcerated in state and federal prisons, over 65 percent work."
That's 780,000.
And there's 163 million working folks in the USA.
Prisoners power .4% of the economy.
> Have you ever wondered why Manus founders gave up billions of dollars from Meta and “voluntarily” moved back to Beijing?
This country wouldn't be such a shithole if we were able to bring the wealthy to heel. I suspect most people on earth see China as the more competently governed country. Understandably, this forum tends to select for people who view private investment positively, but this is clearly insufficient by itself to produce functional societies.
[0] https://openrouter.ai/deepseek/deepseek-v4.1-flash
(The badges next to companies indicate their privacy terms, i.e. green/yellow/red)
The "green" badges explicitly say the opposite:
> "Private: this provider does not train on prompts and does not retain prompt data."
I.e. for the example of Fireworks, this is the precise language they use to describe ZDR:
> "More technically: prompt and generation data exist only in volatile memory for the duration of the request. If prompt caching is active, some prompt data (and associated KV caches) can be stored in volatile memory for several minutes. In either case, prompt and generation data are not logged into any persistent storage."
https://docs.fireworks.ai/guides/security_compliance/data_ha...
Plus only one country has threatened to annex mine and it wasn't China.
The economic sovereignty of any nation, but especially smaller ones like canada and denmark, is harmed by sending our money out of the country when there are options available in country.
I respectfully disagree. I subscribe to the comparative advantage economic principle, so I'm fine with international trade.
The 'clean power' would factor in a well, but I haven't really looked in to it. I know China is installing a ton of renewable and has a carbon market while Canada's building a new pipeline and building Alberta data centres running on natural gas, so I'm not willing to take on faith that switching is cleaner.
Also, re: power. I said canada has lots of cheap power, clean was your word.
The unfortunate truth behind at least the solar PV boom in China is that it's driven in large part by the exploitation of Uyghurs in the Xinjiang region. One of the reasons they can sustain the pace they're going at is because while nowadays PV modules exports increasingly gravitate towards offshore fabs (Vietnam, Malaysia), domestically all the biggest PV players make full use of modern slavery within Xinjiang's re-education camps.
more on the topic: https://www.shu.ac.uk/-/media/home/research/helena-kennedy-c...
It might be a minor nitpick, but that part is just not how the macroeconomic principle works. Abundance of a good/service within a country is irrelevant to the concept of a comparative advantage - this is counter-intuitive but but that was David Ricardo's entire point. It's not about whether production is more efficient domestically, it's whether comparative allocation of resources between multiple countries is more efficient by specialization within the different goods.
Even if you can produce "good 1" cheaper in Country A vs Country B, if Country A can instead produce "good 2" that Country B cannot, it's better for Country A to produce "good 2" domestically and import "good 1".
Now, I still agree with the original notion. It's unnecessarily risky to use services from "democratically-challenged" nations (especially imperialist ones), when safer options are available. But there's no need to twist macro 101 concepts to justify that view. Modern trade policy takes more into account than basic market efficiency.
(But I'd concede that in the field of genocide China doesn't come up as bad)
I feel like I’m getting good value out of my subscription and they have the cheapest GLM-5.3 API rates I could find in Europe. I moved to them recently from Mistral for the better prices.
They do have a slightly different daily inference capacity allowance, but it’s actually not too bad once used to it. I was forced to up my game and figure out what GLM-5.3 Flash and Deepseek-4.1 Flash could actually do for me, rather than just throwing everything at GLM-5.3. Quite happily chugging along with a nice balance across them now.
I am hoping Melious make the new DS 4.1 Flash 1017 available soonest.
Dont assume that is how your model decides...
https://arxiv.org/abs/2305.04388
I can't tell if this is a really human thing to do, or if it's just a sycophantic thing to do when your master likes a good story.
I love COT and gives me a good idea about the thought process. Sure it doesn’t work always
Both OpenAI and Anthropic hides so much of that, that you don't really know if it misunderstood something or you got something wrong, until all the turns are complete or you directly see tool calls that implicitly tells you something is wrong.
You don't need to read any papers to see this effect concretely when developing together with LLMs.
Is DeepSeek: DeepSeek v4.1 Flash 1017 any better?
please don't subject yourself to this. z.ai subscriptions are not a good value. just pay API rates, or use a subscription like opencode go.
DeepSeek feels like an excellent sparring partner. The others feel like mind readers.
I don't have to worry about any complex context engineering / multi agent stuff if I am using DeepSeek because the sequential stuff is fast enough. I don't have to bother about touching a million context length because its cheap enough.
Where it seems to lag behind is 3d, computer use, motion graphics, design but the initial demos seem like the vision model has improved substantially.
If so OpenCode Go looks good and includes DeepSeek models (yet to try myself).
DeepSeek: ZDR agreement is renewed monthly. The current agreement is valid through October 31, 2026.
https://opencode.ai/go
DeepSeek V4.1 Flash / V4 Pro: Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday; all other hours, including weekends, are Off-Peak. Learn more.
https://opencode.ai/v2/docs/console/go/#usage-limits
I always find DS or Qwen to be very spot on on audit changes or plans for just a few cents.
It is very capable for a lot of tasks, not only coding like the other open models.
Just to test, will GLM – hosted by an infra provider other than Z.AI – also stop answering if I asked controversial questions from the pov of its country of origin? Or third party infra providers can perform some unhinging on the model before hosting? Or actually the blocking happens even before the question reaches the LLMs in that country?
(Note: my paid LLM access at this point are only via DS and GLM subs - 90-95% usage; and a free Gemini Pro access. I hope that adds some cred that I am not trying to take potshots at the their country of origin).
I usually default to ds, and if it gets stuck try mimo but usually just for fire and forget tasks. After a bit I come back and see what it did.
No, I don't care about license provenance for all the components of a 30 line script that i'm using in my house, for myself to sync up some folders. No, i'm not going to find and archive the licenses for all the previous projects leading to this for future legal disclosure.
Yes, I own the DVD for the movie i'm transcoding. No I don't want to talk about the legality of rebroadcast.
Deepseek (and most chinese models as a whole) don't care . oAI and Anthropic models want to make it point #1.
... which is extra hilarious knowing the laws broken to create the damn models in the first place.
there's a literal meme about this floating around the harness/AI chats for the past few months[0]
[0]: https://i.ibb.co/SwH4Sgwq/796779070-28483090404651613-920066...
Wait wut? You do realize that they are two different companies, right?
The Fable is the worst offender, and it is not purely Anthropic's fault.
Anthropic, whether Opus 5.5 or Fable, frequently will review for security, and the classifiers fire mid-way while it's fixing something. Authentication and credential related code, including implementation, also trigger safeguards.
Earlier today, I hit what appears to be a sticky "reasoning_extraction" block when attempting to reproduce a LLM paper at small scale, and I asked it ~"tell me more about the idea [X] you considered but dropped". I'm not sure how it works or how long it lasts, but Opus 5.5 has refused to do technical LLM work for me now, even in completely new sessions.
It reduces the speed and tends to overthink sometimes which is really annoying. For programming, 4.1/high is the bread and butter in my daily work.
I had a bad experience with MiMo from Xiaomi. I had it trigger unsafe content blocks while fixing visual bugs in a completely vanilla space game it just wrote from scratch in the same session. It also got triggered for changes in some other benign greenfield app.
Long-term & software engineering fronts: I hear occasional railing against “US hegemony”. Anyone concerned that if a new tech gives rise to a different country, they’ll become the source of ire instead? (This isn’t my field, at all.)
So I go with feeling, and that feeling says that China is worse. The feeling is partially supported by the fact that China is softly backing Russia in its attack against the west.
Trump's USA is obviously completely horrible, but he is also in large part a product of Russian influence, which again is part of the larger block in which China resides.
And China is not really supporting Russia, they make money off the war and want to keep it going as long as possible so both Russia and Ukraine become weak and more open to their influence. Is it evil? Absolutely, but you can't compare them to USA when Trump is actively doing regime change operations, deporting its own citizens and putting sanctions on the ICC.
Chinese communists want to stay in power at all costs but America wants to expand its power into the whole world. It's not the same level of danger at all.
note that the volume of EU and US support to Ukraine is obviously calibrated to the same goal.
Yes, yes they are.
https://en.wikipedia.org/wiki/China_and_the_Russo-Ukrainian_...
Trump’s a smelly orange shit but the US at least pretends to have data privacy protections and we can’t legally compel companies to have back doors… yet. I think. It’s very hard to keep up with all the nonsense these days.
In a post-Snowden world where the existence of PRISM is no longer a secret, I'm not sure how folks can keep stating this with a straight face
It's only for long running/very difficult cases that the differences are here, and they are small.