Yeah, I got the email that they removed them from the student plan:
"As part of this transition, however, some premium models, including GPT-5.4, and Claude Opus and Sonnet models, will no longer be available for self-selection under the GitHub Copilot Student Plan."
They specifically said this was primarily for student plans. I'm surprised they did this for the normal pro plans too; it's likely a mistake since the plans page[1] still says that the models will be available.
However, TBH, I've never liked Microsoft's flavor of these; they always seem lobotomized compared to using the models directly in Claude Code / Codex. I rarely use AI in VS Code because it's just bad.
Corporate India is facing way too many weird problems, not just 90-day notice periods:
1. Refusal to provide leaving documents if you leave on less than excellent terms, but you absolutely need pristine docs and sometimes multiple references when joining
2. Salary expectations as compulsory form fields during job applications, but no salary ranges provided in job descriptions
3. An unhealthy approach to leaves - need doctor certificates, way too early notices for leaves more than a few days, too few leaves, etc.
4. A sudden leap in "immediate joining" requirements - you need to come at once, but you can only leave after at least 90 days
5. Playing games with insurance, salary deductions and compulsory contribution requirements to management's favorite CSR pots
In the past few years I've become so frustrated that I just don't bother with large company job applications, or messages from Indian recruiters, because there's a 99% chance there's a really crappy process involved. Smaller firms with good founders / non-Indian consulting roles are a lot more relaxing, and most of the times pay is higher as well.
Really interesting! Somewhat reminds me of the ending of H. P. Lovecraft's "The Rats in the Walls", where the main character, a scion of a very old family which has done some really bad things, goes mad and progressively starts speaking in older and older versions of English after every sentence.
Thanks, that's such a great detail. I was reading Lovecraft during highschool in locally translated print editions. Where such details didn't come through.
Do you know if there any other such language related eastereggs in other of Lovecraft's writing? should I chose to revisit them, in English this time around.
The Call of Cthulhu seemed to have a bit of language construction and world-building, if you are into that. But my knowledge of Lovecraft lore is limited, so I wouldn't know all details; I just read his short stories from Standard eBooks a few months ago, which was my first exposure to his work.
I'm sure S. T. Joshi might have a bit to say about the topic. Personally speaking from very limited exposure and knowledge of language games, and me not being from an environment which has European language roots, I might have missed quite a bit of such easter eggs in the atmosphere and writing. Like, for example, your comment prompted me to find out what "rue d'auseil" (from The Music of Erich Zann) meant, I didn't bother to find out until today.
I do recommend rereading Lovecraft in English either way, since you never know what gets lost in translation!
The problem comes when rote learning actually is the be-all-end-all. Too many Asian students experience rote learning without any focus on actual learning. Our job used to be regurgitating paragraphs from textbooks, exactly as they were, into our exam papers. In classrooms, we were told that war happened in year X, but there was no discussion and analysis as to actual reasons, the milieu at the time, and the understanding and takeaway from that piece of history.
Facts and memorization are important, but they need to be in service to actual learning and understanding.
```
Drive. The car needs to be at the car wash.
```
Gemini Thinking gives me 3-4 options. Do X if you're going to wash yourself. Do Y if you're paying someone. Do Z if some other random thing it cooked up. And then asks me whether I want to check whether the weather in my city is nice today so that a wash doesn't get dirtied up by rain.
Funnily enough, both have the exact same personal preferences/instructions. Claude follows them almost all the time. Gemini has its own way of doing things, and doesn't respect my instructions.
I saw this pattern a few years ago, with Trump and cryptocurrency.
Oh, Trump is a joke, on the cusp of crashing out. Then Trump is a danger to society. Cycle as convenient.
Oh crypto is a joke with no uses. Then crypto is a danger to society. Cycle as convenient.
None of this stopped Trump being president twice. Nor did it stop Bitcoin shooting up to tens of thousands of dollars. A few years in, I realized that the Atlantic and its ilk are just in the business of publishing articles people will read, or maybe in the business of hyping things up (negative articles seem to just increase the hype, not taint the subject). They don't really seem to change people's opinions, and they certainly don't believe in being consistent.
I am not American. I pay for news, specifically business news. I subscribe to US, UK and Indian news websites.
Both the US and UK feel free to show me ads even when I've paid a bomb in terms of subscription costs. Not subtle ads of their own products! Top banner ads, middle-of-page scrolling ads, and the like, of whichever fancy watch or lifestyle destination has paid the most money to them. And then they have the gall to write opinion pieces on how ad-based AI and streaming channels are the bane of the world. Plus they feel free to subscribe me to a bunch of their newsletters and podcasts which I have to manually unsubscribe from. One of them actually pedals courses on how to write good.
The Indian news sites have no barrier on what is a paid piece and what is actually news. Promoted pieces occupy the same slots as paid ones. I've seen blatant advertisements masquerading as actual reporting.
I understand that news has been gutted by tech. But there is a need to be honest to a paying customer; if not, they deserve whatever has come to them.
> It has a brutal work culture, and if you can survive it long enough
After my first few years of experience, I specifically started filtering out companies which have people from Amazon in a leadership role. Their work culture is poison, and I'd rather not join than see myself become a zombie for the next couple of years. This has helped me stay sane in my career, but there still are some early scars that remain.
I explicitly do not believe that you can "make it anywhere" if you survive some brutal culture, and that surviving a few years in that place will bless you with the ability to "live life at another company". Why ruin my health and sanity when I can directly join that other company? This glorification of bad environments needs to stop.
Today, if I had to do a regular job, which demands more and more out of me while slowly giving me less and less, then I better have a very solid offer to consider it, otherwise it's not worth the hassle, considering half of what they'll offer will be rug-pulled after some time anyway; we've already seen that happen across multiple companies.
I quit a couple years ago, had enough funds saved from the "golden bubble". Took a nice break, and now I'm doing a bunch of different things, a few earning me enough income to live a decent life.
> they’re just anchoring to a world that no longer exists or might quickly disappear
Good engineers aren't blind. If "value has changed shape" for the employers, then it has changed shape for the employees as well. Enough have figured out that they need to diversify out anyway, since companies today cannot and should not be trusted. Only the most desperate will stick around empty shells of regular jobs - there are enough options for talented people to pursue.
Given the current state of global work, for me the more lucrative stuff is internet-related, since it's possible to get developed country rates for software work. I do volunteer in the physical world, so that I can be connected to my local environment.
"As part of this transition, however, some premium models, including GPT-5.4, and Claude Opus and Sonnet models, will no longer be available for self-selection under the GitHub Copilot Student Plan."
They specifically said this was primarily for student plans. I'm surprised they did this for the normal pro plans too; it's likely a mistake since the plans page[1] still says that the models will be available.
However, TBH, I've never liked Microsoft's flavor of these; they always seem lobotomized compared to using the models directly in Claude Code / Codex. I rarely use AI in VS Code because it's just bad.
[1]: https://docs.github.com/en/copilot/get-started/plans
reply