Architect&Designer, Musician&Music lover, PhD candidate/ Room acoustics, Crypto enthusiast, Investor& Trader, Interested in Arts, Technology & Philosophy
OUTer spACE
Joined April 2017
- Tweets2.3K
- Following1.1K
- Followers187
- Likes1.4K
๐๐๐๐๐ โซ retweeted
holy f*ck, someone just dropped an entire reverse engineering degree on GitHub for free.
Now you can build agents that can reverse engineer almost anything:
1. Ghidra
How to read any binary.
The NSAโs reverse engineering framework. Where almost everyone starts.
github.com/NationalSecurityAโฆ
2. radare2
The same skill, but in the terminal.
UNIX-style reverse engineering framework and command-line toolset.
github.com/radareorg/radare2
3. LLM4Decompile
An LLM that decompiles.
Turns binary code into readable C. The 9B v2 model hits 64.9% re-executability on its own benchmark.
github.com/albertan017/LLM4Dโฆ
4. r2ai
An LLM inside radare2.
LLM-powered reversing for radare2. This is where AI starts reading the binary with you.
github.com/radareorg/r2ai
5. ida-pro-mcp
Let the agent drive the tool.
Bridges IDA Pro to language models through MCP.
github.com/mrexodia/ida-pro-โฆ
6. reverse-engineering-assistant
The same thing for Ghidra.
An MCP server for reverse engineering tasks inside Ghidra.
github.com/cyberkaida/reversโฆ
7. REA
Agents that reverse engineer anything.
35 target types: native binaries, APKs, firmware, websites, Electron apps, and more. Every finding is backed by evidence.
github.com/morluto/rea
Read the full REA guide below โ
๐๐๐๐๐ โซ retweeted
THIS IS F**KING Dangerous
HACKERS CAN USE THIS TO MAP AN ENTIRE ATTACK SURFACE.
BBOT PACKS 100+ RECON MODULES INTO ONE FRAMEWORK THAT CAN HELP MAP AN ENTIRE ATTACK SURFACE.
Hereโs what it can do:
โ Discover subdomains and hidden assets
โ Identify exposed ports, services, and cloud storage
โ Fingerprint websites and capture screenshots
โ Search for vulnerabilities and leaked secrets
โ Export findings into reports and databases
It also supports API integrations, custom scan profiles, and Python automation.
Instead of juggling dozens of different reconnaissance tools, BBOT brings them together in one framework.
If youโre into ethical hacking, bug bounties, or penetration testing, this deserves a spot in your toolkit.
BOOKMARK IT.
Only scan systems you own or have explicit permission to test.
REPO BELOW.
๐๐๐๐๐ โซ retweeted
We wrote 150,000 CVE exploits and trained an LLM on them.
Hinoki Offensive One scores 88.4% on CyberGym, 2.8 points above Daybreak, with zero refusals.
Pentesting and exploit development with no limits. Available now.
๐๐๐๐๐ โซ retweeted
Paste this prompt into Claude Code.
It will build you an Opus 5.5 + Jev trading bot.
It will probably make you a lot of money...
๐๐๐๐๐ โซ retweeted
Jev has been exploding in popularity recently.
If you already have access to the Jev API but arenโt sure how to start experimenting with it, just copy this checklist:
1. agent-desktop
Desktop automation. Read the system's accessibility tree, judge which button, menu, or input field to click next. github.com/lahfir/agent-deskโฆ
2. typesafe-mario
Have Jev play Super Mario. No screenshotsโjust read the structured state in the emulator's RAM, then decide to run, jump, or dodge. github.com/fhshaik/typesafe-โฆ
3. jev-drone
Use Jev to control a drone. The underlying flight control still handles stability and safety; Jev just does higher-level judgments like climbing, braking, and navigating obstacles. github.com/RomanSlack/jev-drโฆ
4. OneVOneJev
1v1 FPS in the browser. Every decision tick, judge movement, view angle, aiming, firing, and jumping. github.com/emrickgarrett/Oneโฆ
5. jev-trader
High-frequency market making on Monad testnet. Jev judges the next buy or sell based on spreads and trade direction, with model latency around 81ms. github.com/jarrodwatts/jev-tโฆ
6. Prism
Doesn't directly have Jev place orders. It judges states like toxic flow, market pressure, mean reversion, etc., then hands off to the original strategy. github.com/irfndi/prism-liquโฆ
7. neo4jev
Stuff Jev into a knowledge graph. At each node, judge the most worthwhile edge to take next, then follow it all the way. github.com/jexp/neo4jev
8. jev-curate
Use Jev to screen training data. For JSONL / Parquet, first judge quality, relevance, and risk, then decide which ones go into the next training round. github.com/AkashPriyadarshiiโฆ
9. Canny
Prevents Coding Agents from stubbornly claiming they're done. Look at tool outputs, code diffs, and test results, then judge if the completion claim is reliable. github.com/qkal/Canny
10. killmyidea
Input a startup idea, and Jev scores it from multiple angles, finally giving you KILL, FIX, or SHIP. github.com/monteduro/killmyiโฆ
Copy these complete Jev blueprints - then read full Jev setup below โ โ
๐๐๐๐๐ โซ retweeted
Just tried the wildest AI agent I've seen in months
It's called EvoX and it self-evolves every single time you use it
Every task you complete, it extracts the useful pattern and saves it as a reusable skill. Next time a similar task comes up, it starts from what worked before, not from zero
This is the future of AI agents.
Hereโs how it works ๐
๐๐๐๐๐ โซ retweeted
Sam Altman (CEO of OpenAI):
"You don't need to write prompts anymore. You need loops and graphs that write them for you."
In 78 minutes he explains how to use LLMs better than almost anyone does
Prompts โ Agents โ Loops โ Graphs
most people are still rewriting the same message and calling it work
he stopped writing messages and built the thing that writes them instead
same model, same tokens, completely different week
watch it today
the full graph engineering guide is below, save it while it is still early โ
๐๐๐๐๐ โซ retweeted
Anthropic engineer:
"You're not supposed to prompt Claude. You're supposed to build a system that prompts itself."
In 45 minutes she shows exactly how Anthropic builds agents that remember, fix their own mistakes and get smarter with every run.
This beats any paid course on agents I've seen.
Watch it, then read the guide on building loops below.
๐๐๐๐๐ โซ retweeted
Anthropic engineer:
"90% of our engineers were already using self-improving loops
Now everyone is moving toward agentic graphs"
"Prompting is basically over"
In just 10 minutes, she builds her full Claude Code system live from a completely empty terminal
Agents โ Loops โ Graphs โ Self-Improving Systems
Prompting was the old workflow
Graph engineering is the next one
This 10-minute breakdown is worth more than most $1,000 agent engineering courses
Watch it today
Then save the full guide below before everyone starts catching on โ
๐๐๐๐๐ โซ retweeted
Most people use about a fifth of Claude and think they've seen it. They open a chat, type a question, copy the answer. That is level one of five.
Level one is chat done properly: pick the model for the job, brief it like a colleague instead of prompting it like a search bar, and turn on extended thinking for anything with more than one step. Most people leave thinking off and call the answers shallow.
Level two is memory. Projects hold the files, memory holds the facts, connectors read your Drive and calendar so you stop pasting.
Level three is where it becomes yours. Skills turn any task you keep re-explaining into a slash command. A voice file built from twenty real pieces kills the AI sound.
Level four is Cowork: one folder for the business, and scheduled tasks that run while you're closed.
Level five is Claude Code. You describe the tool in plain English and it builds it.
Bookmark this and find your level. Then go one up.
๐๐๐๐๐ โซ retweeted
I hate to say it, but Omarchy is life changing.
sadly there is a huge problem.
Learning the keybinds are 1000% ESSENTIAL to actually unlocking the operating system.
I see countless people "try omarchy" and give up right away because it overwhelms them.
too new and uncomfortable.
they never get to experience even 2% of the magic.
I built this mini game called Learn Omarchy it teaches you to absolutely fly around your operating system in a way that is NOT possible on mac or windows.
lessons start off easy and progress to full psychopath levels.
they are timed and graded.
you can only progress to the next level once you get three stars on the previous one. The game hides keybinds as you learn them to reinforce your memory.
its free and open source.
I still use mac os daily, but I find myself incredible annoyed that I cannot control my computer at the speed of thought.
Omarchy is fast, organized, and insanely fun.
It has completly changed the way I think about interfacing with computers, we are currently doing things WRONG in every way.
๐๐๐๐๐ โซ retweeted
Anthropic engineer said something that completely changes how you think about AI agents:
"Don't prompt Claude. Build a system that prompts itself."
In just 26 minutes, he walks through exactly how that works from the core idea to building the entire system step by step.
Most people are still manually prompting AI for every task.
The next generation of builders is designing systems that can plan, execute, evaluate, and improve on their own.
Watch the talk first, then save the guide below you'll look at AI agents very differently afterward.
๐๐๐๐๐ โซ retweeted
Google Engineers just showed how their AI engineers are building Context Graphs from 0% to 100%:
10% โ 4:01 - how to build a Context Graph
30% โ 15:00 - Graph Engineering explanation
55% โ 30:00 - Agentic search Engineering
80% โ 35:48 - Graph Engineering practice
100% โ 47:06 - self-improving agents in graphs
54-minute Google workshop, and itโs one of the clearest hands-on examples of the shift from.
Watch it today, then read the full Context Graphs Playbook below.
๐๐๐๐๐ โซ retweeted
Introducing TranslatePsy-EuroNano and TranslatePsy-AfriNano, two new additions to the TranslatePsy family: 9 European languages and 8 African languages, in models small enough to run in a browser.
One multilingual checkpoint replaces a folder of bilingual models.
Model cards: huggingface.co/collections/qโฆ
Blog: huggingface.co/blog/qvac/traโฆ
GitHub: github.com/tether-ai-researcโฆ
and github.com/tether-ai-researcโฆ
๐๐๐๐๐ โซ retweeted
Anthropic hired this engineer at $250K-$750K a year because he knows how to build harnesses for multi-agent systems
In this 15-minute workshop, he shows exactly how to build one from scratch
AI โ Agents โ Harness โ Loops โ Graphs
step 1 โ start with the Claude Agent SDK - the harness handles loops, context, and sandboxing
step 2 โ separate the brain from the hands - reasoning in one place, tools in a sandbox, 60% faster to first token
step 3 โ run it server-side and log every step - close your laptop and it keeps running, crashes resume from the log
step 4 โ make failure cheap - retry dead sandboxes and replay lost context instead of starting over
step 5 โ turn yesterday's logs into new memory and skills - the harness wakes up smarter
Anthropic calls this "dreaming"
Most people spend weeks building this by hand
You don't have to
Bookmark and watch it
Then read the full harness engineering guide below โ
๐๐๐๐๐ โซ retweeted
Stop renting a chatbot and calling it your companyโs brain.
Utopia is the first open-source enterprise world model Iโve seen that actually treats knowledge like a system, not a vibe.
It's a Self-hosted, open source and built around the one question every other knowledge base ignores: when.
Every fact in it carries a start date and an end date. Nothing gets overwritten. When something changes, the correction closes the old version and opens a new one. The history stays intact.
So the graph isn't a snapshot anymore. It's a recording you can rewind.
You drag a timeline on the screen and watch the whole thing redraw itself to show what was true on that date. Who owned what. What the policy said before it changed. All of it, replayable.
And every fact points back to the exact sentence it came from. Nothing gets in without a receipt.
Here's the part that got me.
The whole thing is one binary and one Postgres database. That's it. No Elasticsearch, no separate vector service and no message queue. Most RAG stacks are six services duct-taped together. This is two. It even runs fully offline on an air-gapped network with any model you want.
github.com/deeplethe/utopia
๐๐๐๐๐ โซ retweeted
Andrej Karpathy revealed what survives after prompting
"Harness engineering is taking over"
"Delete everything else and keep the harness"
In a 1-hour lecture, he explains how to build harnesses and why they become the layer every agent system depends on
LLMs โ Prompts โ Agents โ Harnesses
Most people are still trying to perfect prompts
The harness is the part that survives
Everything before it is just a step in the process
This free lecture is worth more than most $600 harness engineering courses
Watch it today
Then read the full guide on harness engineering below
๐๐๐๐๐ โซ retweeted
The ultimate Grok @bot set up for me is to filter EVERYTHING - absolutely everything through one agent and hide everything else.
That agent is in charge of as many agents as needed that will own specific areas for your process/business/personal life. You can even ask the master agent to spin up agents that make sense to own each area.
Then ask your master agent to create chat rooms for all the agents where it makes sense for them to get together and chat about whatever they need to. Remind your master agent to do this periodically so the chats/agents are optimized to whatever the latest iteration of what you're doing is.
Then - most importantly - tell your master agent to PING YOU whenever something needs your attention or needs your action. And tell it to ping you ONLY ONE AT A TIME, and once you clear an item, THEN it can ping you for the next item.
So essentially ask the Master agent to keep a running list of items it needs you to action on.
This massively helps with clutter/distractions/doing too many things at once, which is STUPIDLY EASY to do with a tool that can give you quasi-infinite agents.
OFFLOAD all the complex/multitasking stuff to your agents, and empower them to be as fully autonomous as you feel comfortable, and TRUST THEM with as many things you can trust them with.
The more you trust them with things, and the more autonomy you give them, the less you have to do/worry about it, and with this set up, whatever you need to do will COME TO YOU, instead of YOU GOING TO IT.
Basically treat them as you would a team of ultra talented people - give them the keys to the kingdom, allow them to experiment, and be available for them when they need your help. I think Grok Bot is at a point where this can be done because of how consistent and stable it is, especially relative to other agentic harnesses.
If you want your own set up like this, you can copy the link to this post and give it to your Grok Bot, or you can click on the link here to copy my own: x.ai/bot/j7B5LHnEIPTuPQZxxQwโฆ
๐๐๐๐๐ โซ retweeted
This is the first open-sourced AI classroom.
Not a chatbot that answers questions. A full classroom.
It's called OpenMAIC. You type one topic. "Teach me quantum physics."
Then it builds the whole thing for you:
- An AI teacher who lectures out loud
- AI classmates who ask questions and get called on
- A whiteboard where the teacher draws diagrams and solves equations step by step
- Interactive 3D simulations you can play with
- Quizzes that grade you in real time
The classmates are what got me. There's a roundtable mode where multiple AI agents with different personalities debate the topic while you watch. You learn from the argument, not the lecture.
MIT licensed so free for anyone.
You can export the whole lesson as PowerPoint or an interactive webpage and keep it.
Repo: github.com/THU-MAIC/OpenMAIC
