NL

I used this workflow: Main Host > VMware Windows > Remnux MCP/Ghidra installed on Remnux with the plugin configured. SSH to the Remnux MCP from the VM and HTTP on port 8080 to Ghidra, Claude or Codex Installed on the VM, I created this additional step for security, but limiting the interaction might not be necessary.

@NLx64 Interesting… I’m usually not going that deep on reversing but I’ll try that!

Testing the new GPT-5.4 inside REMnux with real malicious artifacts, memory dumps, and the custom analysis skill I built for it. Honestly… this is next level. The speed at which it pivots through artifacts, correlates behavior, and surfaces relevant findings is something I haven’t seen before. Still early testing, but very impressive so far.

@__mystudylog These notes are great, where are they coming from? iPad? Any specific app?

another day 1: learning how to learn and think - critical thinking ed.ted.com/lessons/5-tips… - logical fallacies ed.ted.com/ted_ed_collect… - reframe problem hbr.org/2024/01/to-sol… - 5 whys method interaction-design.org/literature/top… - inductive & deductive reasoning khanacademy.org/math/algebra-h…

@Kostastsale @fr0gger_ Of course, I will do it!

@NLx64 @fr0gger_ That's great, but why not contribute it to our repo so we can all benefit from this skill?

Tested an Agent Skill I created for PDF triage vs VirusTotal. VT: 0/63 detections Claude Skill: 95/100 risk score AI isn't replacing analysts it's amplifying them. The more you know about a subject the better your skill will be. nlx64.github.io/blog/agent-ski…

@amorriscode Appreciate the effort

@NLx64 yes, we should make them available and easy to use everywhere

We shipped A LOT of updates to Claude Code on desktop in the last week. - Plan mode (coming soon to web) - Notifications for permissions - Perf improvements - Fixed slash commands - Improved env access - Tons of polish We'll be shipping something cool later this week too. 👀

@Ceoz_1 @VictorTaelin If you find out, let me know.

@VictorTaelin TAELIN BENCH VERIFIED 2.0 btw, consegue testar uma coisa ? no pro vc consegue compartilahr com membros da sua familia, no ultra também consegue ? to considerando assinar e rachar com os campangas

For these wondering, and as expected, Gemini 3 Deep Think solves the stack underflow bug that cost me a few days. The answer is a more decisive than Opus 4.5, the only other public model to solve it (even Gemini 3 Pro fails). It even points the exact location confidently. It takes forever though... I don't have harder tests for now, most my benchmarks are saturated and I'm super busy with SupGen stuff, so that's all I have to say about this one

@thsottiaux @floraiNx9 That's what I was talking about.

@floraiNx9 Increased limits by 50% across the board

We gave gpt-5-codex a small sibling, you can now try gpt-5-codex-mini in Codex. Smaller, so up to 4X more usage included. $ codex -m gpt-5-codex-mini Also increased usage limits by 50% for Plus, Edu and Team plans and gave Pro priority processing for a faster experience.

@AIwithJai @Sauers_ O.O How did you do that?

@Sauers_ That's the way to go, also having CC controlling Gemini Cli is a superpower.

Sonnet 4.5 is currently controlling about 20 Codexes at once

@iannuttall @bheemreddy181 What did you experience with the Codex via Claude Code combination? Did you find it superior?

@bheemreddy181 playwright, sometimes postgres, repo prompt but often I don’t bother with them any more and use cli or bun commands for stuff

Not sure what everyone else is experiencing but Claude Code has been crushing it for me today 🔥

@harshith what is the parameter for that?

Using GPT-5 in Codex with bypass permissions mode. It’s insanely fast and smoother than Claude Code Opus 4.1. UI, infra, logic - it rips through everything. If you haven’t tried Codex lately, you’re missing out.

@BrettFromDJ what do you think about gigapixel AI for image upscaling?

My entire design/AI tech stack (25 tools): 1. Figma (design) 2. Jitter (animation) 3. Framer (currently learning) 4. Bolt (ui exploration) 5. Streamline (icons) 6. Creative Market (mockups) 7. Midjourney (Image + video gen) 8. Runway (cinematic video gen) 9. Veo 3 (video + sound gen) 10. ChatGPT 4o (everything) 11. Claude (coding help) 12. Astra (4k video upscaling + detail) 13. Bloom (image upscaling + detail) 14. Landbook (landing page inspo) 15. Pinterest (general design inspo) 16. Pixel Buddah (photoshop effects) 17. Contra (brand payments) 18. Stripe (agency invoicing) 19. Loom (client communication) 20. Screen Studio (screen recording) 21. Artlist (sound effects) 22. Restream (live streaming) 23. Figma Slides (slide decks) 24. Trello (project management) 25. Spark (email)

@iannuttall @alexchristou_ In your use, what works best? A file formatted for AI ingestion on the directory folder, via git ingest for example. Or the direct link like uithub.com/example

@alexchristou_ MCP server like Context7 or custom built works well. Or just giving individual links to single docs

Claude Code 1.0.86 has a new /context command to visualize your context window and token usage It doesn't seem to be accurate yet compared to the auto-compact notice, ccusage, or my status line, but it also shows MCP tool use and CLAUDE.⁠md memory files so could be v useful!

@RelamEditor @gustxvo_editor @ChriSSSeditor Ainda tem o link? vou dar um confere é minha área.

@gustxvo_editor o @ChriSSSeditor sem querer, disponibilizou um tutorial que fazia geral baixar 1 plugin, o mocha no premiere, e ele de fato funcionava, mas em troca ele pegava teu discord, teu twitter, e todas as contas logadas no navegador

pra quem baixou (igual eu) eu fiz a seguinte parada pra n formatar o pc -passar antivírus (de preferência uns top pique malware bytes) -limpa o cache do navegador -desinstala -procura pastas do navegador no %appdata% %localappdata% +

@__ambrosio @ardepcec Google é o lab quase perfeito para AI, se eles conseguirem escalar os TPUs...

@ardepcec O Checkpoint realmente bom do GPT5 foi para o plano mais caro pq eles não tem GPU suficiente para todo mundo usar, então isso frustrou muita gente. Espero que o Google não faça isso.

A AGI pode não estar nem perto de acontecer, mas o Gemini 3.0 alcançar 32.4% em HLE é algo MUUUITO grande! Repito: Gemini 3.0 >>>> GPT5

@Space_Sausages @briongoudreau @iannuttall he wouldn't do that to me. would he?

@briongoudreau @iannuttall it will cheat on those tests and strip out working code for simplified stubs to pass unit tests

It's amazing how effective this one sentence prompt is to stop Claude from overengineering or refactoring your codebase

@4ster_light I'm also using Fedora, which surprised me to be quite honest, after a long time switching between Arch and Debian.

So this is the end result, I just love simplicity

@NathanWilbanks_ @SIGKITTEN I liked the design, what is it?

@SIGKITTEN claude and mcp are gey and their tools don't even work is this a "its a good agent, sir" joke?

I can't fucking believe noone can get their shit together with coding tools and are letting Anthropic completely dominate