Mark Dykun

[Mark Dykun @MarkDykun]

It may be time to cancel @github copilot. It looks like now every task is held back. it is now the least efficient way of getting development work done. It tells me its gonna start work, and just sits there until I tell it to get to work.

[Mark Dykun @MarkDykun]

I know no one will see this here on twitter, but this is one of the products I am currently building. UI is coming along nicely.

[Mark Dykun @MarkDykun]

@github @OpenAI

How can I contact support, I have see something concerning in a copilot refactor that might be wide spread and involve GPT-5.4 and or Copilot. It would be good to discuss what I have seen.

[Mark Dykun @MarkDykun]

When did Copilot in VS become confirm happy. Long running process my ass when I have to confirm a search string every couple of minutes. #GitHubCopilot

[Mark Dykun @MarkDykun]

@MUKIREEE How f'n stupid is this... AI all the way

[Mark Dykun @MarkDykun]

@mkristensen sir, thank you. Looking at some of your Visx code got me past a dumb issue with commands not working correctly in vs. Here is the fruits of my labor today. Voice prompting of AI chats with some fun Jarvis like text to speech feedback. @VisualStudio #MyFirstAddIn

[Mark Dykun @MarkDykun]

@luireigns oh So Adam looks like John Cena lol. So pathetic

[Mark Dykun @MarkDykun]

@kellabyte @davidfowl what is this language you speak? I feel dumber reading your posts.

[Mark Dykun @MarkDykun]

@mkristensen I guess I will try this as well, ChatGPT was acting wonky last week for me.

[Mark Dykun @MarkDykun]

@mkristensen Last couple of days Copilot chat when using the OpenAi models have been broken. In the chat window the responses were showing json and not well formatted responses and it was getting confused and not generating good code. I had to walk away from the desk. Claude Opus seems better

[Mark Dykun @MarkDykun]

@ElonMusk_LordX 2

[Mark Dykun @MarkDykun]

@rasmalai GO, RUST

[Mark Dykun @MarkDykun]

@SumitM_X sorry for my quick typing, ... return the least amount of records as possible.

[Mark Dykun @MarkDykun]

@SumitM_X 1This seems like a stupid question to me but no SAAS solution would allow the ability to scan, page through 1M user.

1. There would be smart querying with a goal of remove the least amount of rows as possible. Users would be segmented base on teams, departments, etc

[Mark Dykun @MarkDykun]

@SumitM_X if the user has selected a query that can bring back too many rows, I would let them know so that they can tighten the query.

this whole get back a million rows is a scalability concern with a SAAS application, multitenant, shared resources etc and should be avoided

[Mark Dykun @MarkDykun]

@SumitM_X 2. Any unit of work that needed to be done on a large number of users would be offloaded to a job server, including exporting, reporting, realignment.

3. Any user experience does not win by getting all the data but the right data.

Paging is fine, but I would the dataset

[Mark Dykun @MarkDykun]

@eurofounder The better question would be, what is the strike price the options are offerred at and what is the vesting schedule. Google is not a failing company but to figure out the compensation you need to know those things. Then you can determine the risk and reward.

[Mark Dykun @MarkDykun]

@davidfowl for those thinking AI is ready for primetime. I asked copilot to create a simple feature that would require adding a single method to a repository class, instead it duplicated the existing methods 3x breaking the build

[Mark Dykun @MarkDykun]

@davidfowl I think the right mix is AI with some extremely capable product builders (Devs, UX, QA). Honestly when you mix it with People driving the process its a force multiplier. When you let it do it all, its spaghetti of mixed abstractions and half successful implementations.

[Mark Dykun @MarkDykun]

@nickchapsas @Khoral666 @DD_DD_DD_DD is this still Relevant in 2026? Legit question given the state of affairs with AI, and the broader engineering community.