Not my experience, honestly. With a good code base for it to explore and good tooling, and a really good prompt I've had excellent results with frankly quite obscure things, including homegrown languages.
As others said, the key is feedback and prompting. In a model with long context, it'll figure it out.
I use GLM 4.7 with DeepInfra.com and it's extremely reasonable, though maybe a bit on the slower side. But faster than DeepSeek 3.2 and about the same quality.
It's even cheaper to just use it through z.ai themselves I think.
Given the mass of code out there, it strikes me it's only a matter of time before someone fine tunes one of the larger more competent coding models on COBOL. If they haven't already.
Personally I've had a lot of luck Opus etc with "odd" languages just making sure that the prompt is heavily tuned to describe best practices and reinforce descriptions of differences with "similar" languages. A few months ago with Sonnet 4, etc. this was dicey. Now I can run Opus 4.5 on my own rather bespoke language and get mostly excellent output. Especially when it has good tooling for verification, and reference documentation available.
The downside is you use quite a bit of tokens doing this. Which is where I think fine tuning could help.
I bet one of the larger airlines or banks could dump some cash over to Anthropic etc to produce a custom trained model using a corpus of banking etc software, along with tools around the backend systems and so on. Worthwhile investment.
In any case I can't see how this would be a threat to people who work in those domains. They'd be absolutely invaluable to understand and apply and review and improve the output. I can imagine it making their jobs 10x more pleasant though.
Which COBOL... This is a particular issue in COBOL is it's a much more fragmented language than most people outside the industry would expect. While a model would be useful for the company that supplied the data, the amount of transference may be more limited than one would expect.
I (Canadian) drive a Polestar 2. Chinese manufactured car by Chinese company Geely (tho with Volvo DNA).
It's the best winter driving car I've ever owned. A set of Michelin X-Ices on it's amazing. I've been driving for 35 years and I've never driven something with better winter handling, including Subaru I used to own, etc.
As an EV driver from Ontario it's amazing crossing over, or even getting close to the border, the EV situation is just so much better.
That said, while hate Tesla the company ... I'll take their chargers over the patchwork of various apps & cards I have to install in order to make use of things there. There's a notable absence of Tesla stations there, but a lot of variety of other things and I had bad luck installing half the apps and it was not fun trying to set that all up while standing in the -20C cold in a gas station parking lot while just trying to get to the ski hill.
The laughing party is the person taking the tariffs and living large off them. The American consumer is suffering.
In reality, the vast majority of Canadian exports are energy and potash, neither of which have any kind of tariffs applied.
Because if they did, Trump's supporters would lose their shit completely. Gas prices would go through the roof and farmers would be in big big trouble.
The message here is that western Canadian agricultural & energy interests are of potential more strategic value than a dying technically backwards auto sector led by three moribund regressive manufacturers who have shown their willingness to show their belly to Trump anyways.
As an Ontarian, I'm saddened. But I don't think the Big3 deserve anymore state support. They've pushed it too far.
Just earlier this week they were running editorials against the gov't on EV mandates. Again. Ok, here you go. Don't want to make EVs? Only want to sell giant Canyoneros?
It's ok. China will service that market. Have fun becoming irrelevant. If consumers really don't want EVs, like they said earlier in the week, then there's nothing to fear. Right? Right?
More that Ontario auto is projected to have no value since Trump has explicitly signaled he wants to kill Canadian auto and reshore to US. If Ontario auto is going, no sense in losing also agriculture especially if oil also going in 5-10 years if VZ ever works to US favor. The only hedge is to save Canadian auto is hoping for some sort PRC JV where Canadian plants keep some jobs and grab some margins, possibly a lot of margins (i.e. no truly cheap EVs) since PRC inputs cheap. Best case scenario is Canada has meaningfully cheaper EVs, but not Chinese cheap, get to hold onto some auto work, have access to worlds largest ag buyer, maybe free up an extra million barrels of oil to export since US will want VZ heavy instead of WCS from Alberta. Although US has many other ways to punish Canada.
total duration: 57.312624726s load duration: 83.851327ms prompt eval count: 31 token(s) prompt eval duration: 114.445187ms prompt eval rate: 270.87 tokens/s eval count: 1838 token(s) eval duration: 56.76373212s eval rate: 32.38 tokens/s
HEAD of ollama
Trying now with vLLM with BF16, and hopefully FP8 soon after.
BF16 predictably bad:
reply