ricochet

programmer, reverse-engineer interested in x86 emulation and security

C++, Rust, Windows

[ricochet @mkovavoda]

gpt work is no longer free :( rip unlimited usage

[ricochet @mkovavoda]

@desphixs same here, 4 concurrent chats (mix of sol and luna max), dropped 15% in a day versus 50%

[x86byte @x86byte]

Static Devirtualization of Tencent VM by @BackEngineerLab. Great read for anyone interested in VM-based obfuscation and devirtualization.

https://t.co/z15HqIezJc

[ricochet @mkovavoda]

@kimmonismus The entire opus 5 series, with the exception of fable, was an absolute benchmaxxing shitshow, just to get back on the spotlight as subscribers migrated to OpenAI. Neither Opus 5 or Sonnet 5 are good models, whether on release or now.

[ricochet @mkovavoda]

people are sleeping on 5.6 pro

It can code within your github repository, if you have runners/workers, it can compile and test the code. It's not included in codex usage.

It's an amazing, thoughtful model.

[ricochet @mkovavoda]

@iuditg Terra is a very good model. I use it for majority of my tasks now. Rarely do I need Sol intelligence

[ricochet @mkovavoda]

@hodler_investor @Warczech @Ceskasporitelna Objednaj fyzicku.

[ricochet @mkovavoda]

@willdepue connect github, setup runners and it can code for you in your repo.

[ricochet @mkovavoda]

@thsottiaux fast mode?

[ricochet @mkovavoda]

@notch I'd suggest getting both claude and gpt on subscription plans first. They both have their strong points. Usually I begin tooling development with claude, and heavily refine it with gpt5.6

I think computer use on gpt5.6 shines too for actual auto-testing on a GUI.

[ricochet @mkovavoda]

@Da7_Tech ignoring resets, the plan value is about the same as claude; model a bit better than opus

however fyi i did less than 1b tokens on claude, and over 40b on gpt5.6 sol

[ricochet @mkovavoda]

@thsottiaux Large chats in the app on Windows still lag out for a good while.

[ricochet @mkovavoda]

@jturntdev omp is great, but ultra works best in codex app

[ricochet @mkovavoda]

@lukaszstarosta thankfully tibo's been making them weekly again!

[ricochet @mkovavoda]

>go to sleep with 10% usage, praying for @thsottiaux to reset
>wake up, usage reset
>prompt and make coffee

life simple. life good.

[ricochet @mkovavoda]

@thsottiaux It would fe fantastic if it stayed removed, and for weekly management if we could set artificial caps dynamically based on how much we use codex in a week; I can surely go through 100% in a day but that's just Sol deslopping my code from claude at scale.

[ricochet @mkovavoda]

gpt computer use has to be the craziest shit i've seen

[Crackmes.one @CrackmesOne]

We built crackmes-RE: 4,598 crackmes labeled with a flag and/or a runnable verifier script (2,172 have one), plus normalized obfuscation/anti-debugging/protections tags. For benchmarking LLMs' and decompilers' reverse-engineering ability. https://t.co/R9CGvziTUK

[ricochet @mkovavoda]

They're winning a lot of good will from me. GPT5.6 Sol rocks. Glad I resubscribed. Opus 5 will have a tough enemy.

[ricochet @mkovavoda]

@thsottiaux I used all my resets :( sol too good..