Lots of articles on it... and I am not even talking about competitors like Benioff [1]. I am talking about user complaints like this [2]. Users expect Copilot to be fully integrated, like Cursor is into VSCode. Instead what you get is barely better than typing into standalone AI chats like Claude.AI.
The linked complaint is specifically about Microsoft Copilot, which despite the name is completely unrelated to the original GitHub Copilot. VS Code's integrated GitHub Copilot nowadays has the Copilot Edits feature, which can actually edit, refactor and generate files for you using a variety of models, pretty much exactly like Cursor.
Could you elaborate on your issues with LangChain?
We're kinda headed towards using it as it seems to be a flexible enough abstraction that is relatively stable to work with, so I'd like to know if I'm overlooking something..?
A big soup of untyped json blobs and python, all working highly asynchronously in complicated ways.
Is this variable available in my agent’s state?
Depends! Is it an agent that started an hour ago before you made that other change? Is it in the manager state or not? Did the previous agent run correctly or not? Did something else go wrong?
Interestingly enough, e.g. Elsevier accepts latex but has their own typesetting backend. Which typically means that the last editing steps are quite annoying, because even if one is using the provided latex templates, what actually happens for the final typesetting is done by some mechanical turk editor on a slightly different publishing system.
Most important to me was something not mentioned here: to make i/j les ambiguous, I but effort into explicitly add the right swoop at the bottom of the i, which allowed for the j to be a straight longish line dotted at the top.
Copilot, back when I used it, completely ignored context outside of the file I was working in. Copilot, as of a few weeks ago, the absolute dumbest assistant of all the various options available.
With cursor, I can ask it to make a change to how the app generates a JWT without even knowing which file or folder the relevant code is in. For very large codebases, this is very very helpful.
ya know what, after a couple times hearing this comment, I downloaded it literally yesterday. It does feel pretty different, at least the composer module and stuff. A bit improvement in ai tooling imo
This article makes it sound more complicated than it is, collecting all those edge cases. Many of them can be simply ignored, as in being no-op. Which is also what most cheap calculators do...
Traditionally this would be solved by inserting a comment into the code that insults the user (for example the classic "The user is a wanker" comment [1]).
> Shit. I was just about to launch into an explanation of our code review procedures. Every week we sit around a table and carefully and dispassionately analyse and constructively criticise each others code. And it works. We sit there and listen and take it all in. It works really well and team morale is excellent. One day we're even planning to do one when we're sober.
I think there is a lesson here. Don’t let edge cases define your complexity whenever possible. Check if it is in fits your valid expressions and if it doesn’t do nothing, otherwise you get a mess of if x then y
You don't need an ASIC in the way the article's challenge described the lunch problem. Do it with an MCU. The article goes into how the author did it at the register level, but that is not how a modern audience would do it. If you're doing registers, it's probably for your MCU's CORDIC peripheral etc for trig!
You can even go fancy with floating point if you want (Probably elims the $0.30 MCUs). Tiny, cheap, and you can model it using appropriately expressive data structures (like rust enums) that handle the edge cases at the CPU level.
I don't mean to dismiss your comment; it's important. I think this article is skirting a gray area of which constraints are applied. Then the conclusion is altered. Is it saying making a 4-function calculator is complicated!, or is it saying making a 4-function calculator is complicated if you add a number of specific restrictions and requirements (No modern CPU code, exact behavior replication to all combinations of user input etc). The latter is less interesting.
This was discussed in the thread a couple days ago about board game rulebook design. I had a similar question (https://news.ycombinator.com/item?id=42302038 - although I phrased it more like wondering if Procida had been ripped off) and got some insightful replies.
In short: it looks identical because it's substantially his work, that he did while employed there. In fact, he's the one giving the talk in the PyCon Australia video there.
I actually find the contrasting approaches (see discussion elsethread about the terminology used in the diagram) helpful and it's always good to have historical context.
I agree that the Divio site would be well-served by directing folks to/acknowledging the newer site (as the newer site does).
