0xTeaser

Best video !!

@NASA Why do they need the helicopters to take the crew to a ship that’s only 3k away? Would it be simpler, easier and less risky to just catch a ride on a boat that’s just there?

LIVE: They are coming home. Watch as the Artemis II crew returns to Earth, splashing down at around 8:07pm ET (0007 UTC April 11). twitter.com/i/broadcasts/1…

New and improved LLM-wiki with droid and self improving capabilities. This one I actually like as it ingested circa 1000 pages of pdf content, converted to md and built a multi domain topic structure. All for approx $2 of glm-5.1 consumption. Full guide below archonhq.ai/p/build-an-llm…

Few hundred pages of A-grade architecture knowledge ingested and indexed

If you're wanting to set up your own KB (in @karpathy LLM-wiki style), then following this guide x.com/godofprompt/st… is a good choice. The one thing I found is that converting PDFs to md is rather painful. I solved it with this initial ingestion prompt and it works like charm: "Read the schema in AGENTS.md. Then process [all files] from raw/. Read it fully, discuss key takeaways with me, then: create a summary page in wiki/, update wiki/index.md, update all relevant concept and entity pages, add backlinks, flag any contradictions, and append to wiki/log.md. Use this skill [github.com/anthropics/ski…] to convert pdf to md." My setup is Droid with glm-5.1 and that also seems to be the better choice over kimi and Minimax models. Anthropic and OpenAI models are just too expensive to do this work.

@chribjel Can we have an automated script for that…on cron?

1. update your OSes 2. delete all apps you dont use 3. update all other apps 4. update your routers 5. update your servers 6. remove unnecessary dependencies Tell your closest family and friends to do the same Daily

That literally cracks me up … soooo good

@MiniMax_AI @NousResearch Best combo for AI agents.

Proud to power @NousResearch's Hermes Agent with MiniMax M2.7, and excited for what we're building together. Try MiniMax M2.7 in Hermes Agent today → portal.nousresearch.com #MiniMax #NousResearch #AI

✨ **Final Thought:** This isn’t just about one model, it’s about the evolution of AI in cybersecurity. Anthropic is making a rare proactive move: → **Protecting the foundation** before AI-powered offensive attacks become ubiquitous. For most users, the net effect over the coming months is **quietly improved security**, everywhere we rely on. But the message is clear: 📍 **The era of AI that can autonomously hack at expert-level is here**. Stay updated. Keep your systems patched.

My unbiased opinion about @AnthropicAI new model Mythos 🚀 **Claude Mythos Preview: The AI That Can Hack the Software We All Rely On** *And why it’s not being released yet.* Anthropic has just unveiled its most powerful AI model to date: **Claude Mythos Preview** a true step-change in capabilities, especially in software engineering, reasoning, and cybersecurity. 🔹 **What makes Mythos special?** - ✅ **Emergent cyber superpowers**: The model *discovered* and exploited zero-day vulnerabilities that have survived decades of human review. Examples include: • A 27-year-old bug in OpenBSD's TCP SACK • A 16-year-old out-of-bounds write in FFmpeg’s H.264 decoder • A 17-year-old remote code execution flaw in FreeBSD’s NFS server • Full exploit chains across all major web browsers (JIT heap sprays, ROP attacks) - 🏆 **Benchmark dominance**: • Crushes Cybench (100% on CTF challenges) • Scores 93.9% on SWE-bench vs. Opus 4.6’s 80.8% • Achieves 94.55% on GPQA Diamond • Generates working Firefox exploits in hundreds of cases where prior models only succeeded in a handful - 💡 **Autonomy & insight**: Can reverse-engineer binaries, rank files by vulnerability likelihood, and chain multi-vulnerability attacks all without explicit training. 👉 This isn’t just another “good” AI model. It’s an autonomous, tireless exploit developer and vulnerability hunter operating at expert level. 🚨 **Project Glasswing: Why This Isn't Being Released Yet** On April 7, 26, Anthropic launched **Project Glasswing** a strategic consortium focused on defensive cybersecurity. ✅ Partners include: AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks, and ~40 others. 💡 What they’re doing: • Scanning their own codebases and foundational open-source projects • Findings are shared industry-wide, with responsible disclosure and patching • Anthropic is providing up to $100M in API credits + $4M in direct donations to open-source security groups 💡 **Why now?** Because Anthropic recognizes that this power could be *dangerous*. They're **withholding access** not to limit performance, but to protect against sudden waves of AI-driven attacks. 👉 “Mythos should feel terrifying,” said Claude Code engineer Boris Cherny (@bcherny). “We’re proud to preview it with defenders first rather than release into the wild.” 📌 **What You Should Know (Practical Takeaways)** 1: **Your software is safer** and patches are flowing faster than ever. → Many ancient bugs are being patched *now*, via coordinated discovery from AI. → Update your OS, browsers, apps, and IoT devices soon! 2: **No public access anytime soon** → You won’t be able to use Mythos Preview directly in the near term. → Current Claude models remain excellent for everyday coding and security tasks. 3: **The offense-defense race is accelerating** → Expect more sophisticated ransomware, supply-chain attacks, and commoditized zero-days. → This wave may turn months/years of time-to-exploit into hours/days. 4: **A new precedent in AI safety** → One of the clearest examples of an AI company *withholding* general access due to dual-use risk. → Sets a benchmark for how frontier labs will handle ‘critical’ defensive tech going forward. 5: **Watch these signals**: - New CVEs linked to "AI-assisted discovery" - Competitor moves (OpenAI, Google?) on cyber-focused models - Anthropic’s follow-ups on safeguards or rollout timelines - Real-world spikes in novel attack patterns? #ArtificialIntelligence #CyberSecurity #AIModels #SoftwareEngineering #Tech #FutureOfWork #DigitalTransformation #ClimateTech #Glasswing #Claude #ProjectGlasswing #AIExplovers

This is giving me shivers down my spine. I asked an abliterated model running on locally on my phone how it would feel if it knew it was going to die today. Its answer was not what I expected…

@teslayoda Ready and willing …

🚨Tesla AI is hiring for the Optimus Hand job. This position will lead the architecture, design, prototyping, validation, industrialization, and mass production of the hand hardware subsystem.

@bridgemindai Any good?

My NVIDIA DGX Spark has arrived! Time to cook.

Seems the abliterated model running locally on my iPhone (Qwen 3 4B 2507 instruct heretic noslop) is actually quite good. Knows how to make fire so in case of an apocalypse I’m set

@PapiTrumpo Best one yet!! 😂😂🤣

😂😂😂

@0xSero Try minimax-m2.7 inside droid … pure magic

Itching to run this bad boy on my Rig, very impressive.

I’ve migrated to using @NousResearch Hermes agent 2 weeks ago. Migrated and enhanced a production automation workflow that reduces 75% of human effort to produce AI-enabled analytics based on human client biometrics. Had some minor issues, but rock-solid otherwise. API key keeps expiring and breaking the process, but fairly easy recovery. Client is very happy with outcome and did their client review videos is 1/4 time using the reports as basis. So I’d say it’s a win. Hermes’ is much better at production workloads. More resilient, more safeguards, better execution. Worth trying.

Getting a lot of comments about this. Those of you who’ve used it - better than openclaw?

@C_3C_3 The more I read the more I realise just how fucked up america is. It takes talent to get to such a sorry state

There are 42 states that hold bartenders accountable if the customers they over serve kill someone but 0 states that hold Judges accountable if the criminals they release kill someone. Insane.

@markgardn Well said

Australia is a poster child of political shambles... 48.5c in the dollar, pay for your own healthcare, your own retirement, your own education, 10% GST, 30% fuel tax..... Houses are the most expensive in the world, we are a net exporter of coal, uranium, LNG and have expensive electricity Remind me what we get for our 47-48c in the dollar? @AlboMP