Andy Pilate

I made in 2012 a fork of Litecoin (which was already a fork of Bitcoin), with the aim of just having fun in understanding how the codebase worked and how the blockchain works. I named it BBQCoin (BQC), and it turned out to be the first memecoin in history.

@IsidorN @pierceboggan I think the bun.lock change is too big for your prompt/model

@pierceboggan @IsidorN After checking in copilot debug view, it's medium

If I understand correctly there is still an effort parameter you have to give (low, medium, high, xhigh) to guide this adaptive thinking. My question was what's it set to for gpt5.4 when the parameter is on default in vscode. Openai's API docs mention that gpt5.4's default value is none. For the avoidance of doubt I'm talking about github.copilot.chat.responsesApiReasoningEffort

@IsidorN @pierceboggan when keeping default for the reasoning level in vscode, does it adjust it in some smart way or is it just like medium all the time? I have it at xhigh but I'm wondering if I should go down to high or even default if it's clever enough (gpt5.4 only)

I miss before LLMs. Even though I use them daily I wish we could go back to a world before them.

@talkingj3ss As someone who played kcd2 then kcd1 I was like wtf I was so good at it before

using the grindstone in kcd2 is leagues better than in kcd1 bloody hell

@pierceboggan @IsidorN Copilot CLI doesn't seem to have adaptive reasoning, or did I misunderstand what you meant?

@Cubox_ @IsidorN We have adaptive reasoning enabled so it does change. Those settings are informed by what we see in offline evaluations pre-launch, and online evaluations post-launch. Generally xhigh is overkill and sometimes leads to regressions in quality, and will lengthen turn time

ngl one of the big reasons I won't even consider alternatives to Github Copilot and VSCode is the fact that you and other Github folks reply here (and quickly), on top of the transparency and the open source of the whole Copilot ecosystem. Thanks!

@pierceboggan I'll stick with 5.4 then, thanks! With the recent switch to weekly releases, I think I will switch to VSCode stable? Insiders was nice for the features, but daily updates is quite annoying (especially when using remote ssh, where you need to update twice).

@Cubox_ In general, I always recommend latest models. We have discussed publishing our benchmarks on our own vsc-bench so you can see for yourself.

@pierceboggan @IsidorN I had it on xhigh without really thinking and the long turn times were annoying, which is why I asked. I'll switch back to the default and see if I can feel a difference. Thanks for the reply!

@JoshCaughtFire I'd try with gpt models they have a higher context window (gpt5.4 has 400k)

So GitHub Copilot... how is it this bad? > Sonnet 4.6 plans... > Does 1 thing, but marks everything complete > I compact > Switch to Opus 4.6 > "Didn't do X,Y,Z, please audit" > Read 5 files > Auto compacts > "Everything was marked complete, what would you like me to do?" WTF?

@_Evan_Boyle Wish it was 0x and not 0.33x

Expect to see more 5.4 mini popping up in various subsystems of the Copilot CLI

@pierceboggan @IsidorN also while i'm at it, you can ignore the "Contact your admin" for 5.4, that's not part of the problem

@Cubox_ Hmm, @IsidorN any ideas?

@pierceboggan is it normal that I don't have Auto in my model picker? I might have disabled it previously, but I don't see any setting to do so anywhere.

@pierceboggan @IsidorN I think I found the issue, opened github.com/microsoft/vsco…

@pierceboggan maybe with the new 1M context window with Claude, we could have it with Copilot? 👀

@allisx86 google translate "new" into french :)

The way france is ruining everything becuase how tf do you spell the name of the open source nvidia drivers

@thekitze Look into Kyber, it's still in the early stages but it's promising. They have very low latency and it's fully open source. It's also made by the guys at VLC, and they're insane (in a good way) at their job.

if u have a gaming pc and wanna stream your games anywhere i'll throw a few buzzwords at you and i hope you're smart enough to figure it out ◈ sunshine + moonlight (legacy) ◈ artemis + apollo (fork) ◈ vibeshine + vibepollo (latest) gl hf it works AMAZINGLY well

@TerraDevlog Why does it look like these creatures will form a consciousness and then get plugged into my brain and end civilization in a Netflix episode

Me and my homies praying. store.steampowered.com/app/3447760/Lu… #gamedev #indiegame #pixelart #indiedev