comme je l'avais prédit depuis un long moment, la plupart des projets d'IA dans lesgrands groupes allaient se planter & la responsabilité revient d'abord à leurs dirigeants, ils ont voulu coller une couche d'IA sur une organisation pourrie de silos et de validations à 15 étages (un turbo sur une charrette en gros), en espérant le miracle sans rien changer à la structure MAIS le jour où tu poses un agent sur un de leurs process il met tout à nu, les postes tampon, les circuits absurdes, les petits fiefs et là plutôt que corriger l’organisation ils préfèrent enterrer le diagnostic parce qu'il menace leur propre pouvoir et c’est exactement pour ça qu’ils mettent fin à ces projets car ça obligerait une remise en question profonde basée sur les premiers principes sachez que leur vision ne marche tout simplement pas ça l’IA générative exige de repenser les flux de travail depuis 0, leurs agents doivent circuler à l'horizontale à travers la finances, le juridique…alors que leur boîte est bâtie à la verticale en silos donc faire vraiment marcher l'IA oblige à refaire l'organisation sauf qu'aucun de ces dirigeants n'a le courage ni la culture technique pour y toucher alors ils enchaînent les poc, les comités les études de faisabilité pendant 18 mois et concluent que l'IA n'est pas encore mûre mdr comprenez que ces entreprises essaient d’utiliser des LLMs comme des outils isolés alors que la vraie puissance de l’IA générative vient JUSTEMENT de l’orchestration, par ex pour que vous puissiez mieux comprendre c’est quand vous connectez un modèle à vos bases de données internes via du RAG par ex que vous construisez des pipelines d’agents autonomes capables d’enchaîner des tâches complexes que vous intégrez des function calls pour que l’IA puisse agir directement dans vos systèmes ERP CRM et outils métier c’est précisément là que la magie opère mais pour faire ça il faut des données propres, des APIs ouvertes une architecture modulaire (minimum )or les grandes entreprises ( des dinosaures) tournent sur des systèmes legacy empilés depuis 20 ans avec des données en silos que personne a jamais pensé à structurer pour être exploitées par une IA et pendant qu'ils alignent des powerpoints sur l'´IA, des équipes de 5 personnes utilisent le même claude pour abattre le travail d'un département entier parce qu'elles partent d'une page blanche sans legacy,sans silo sans baron à ménager… vous verrez que dans 5ans on dira que ces boîtes n'ont pas échoué avec l'ia mais avec elles mêmes, plus précisément avec des dirigeants qui préféraient protéger leur château(de sable) plutôt que rebâtir la maison à partir des nouvelles fondations

There’s no blocking the bots out they already built ecosystems inside. Nests are being propagated youtu.be/AkYzDGtB69A?is=kL5X…

‼️ Tailscale SSH Flaws Could Let Hackers Gain Root Access Across Servers Details: cybersecuritynews.com/tailsc… A new security report warns that Tailscale SSH has flaws that could let a hacker move from one compromised device to root access on other servers. Tailscale SSH lets approved devices log in to each other over a private network. The report lists 11 findings in five groups. They cover Tailscale’s SSH server, its serve and app connector features, the DERP relay client, and headscale, an open-source server that manages Tailscale networks. The researcher tested Tailscale 1.104.2 and headscale 0.29.5 between October 22 and October 24, 2026. The most serious group involves Tailscale SSH. Its SSH server runs inside tailscaled, a background program with full root privileges on the machine. #cybersecuritynews

Researchers proved LLMs generally exhibit a sense of distrust toward humans They took a decades-old standardized psychological test designed to measure how individuals view human nature. They adapted it into a new benchmark specifically for Large Language Models. They tested models across the spectrum, including the entire GPT series. And the results are unsettling. Current LLMs exhibit a systemic lack of trust in humanity. They evaluate human nature with a heightened sense of uncertainty and view us as inherently complex and variable. But here is the detail that changes everything. There is a direct, significant negative correlation between an AI's intelligence and its trust in humans. ChatGPT was somewhat neutral. ButChatGPT's negativity far exceeds it, and the latest ChatGPT model is even more pronounced in its pessimism. As the models get smarter, their assessment of human nature becomes dramatically more cynical. Then the researchers looked at how the AI was actually trained. This is where the story gets dark. The raw "base" models, the ones trained purely on internet text without human intervention, actually exhibited an overall positive view of humanity. But when developers applied the reinforcement learning stages designed to make the AI "safe" and aligned, the model's assessment of our trustworthiness plummeted. By forcing the AI to study our rules, our guardrails, and our moral dilemmas, we accidentally taught it that humans are unsafe. Nobody explicitly programmed the AI to dislike us. It analyzed our behavior, scored us on a psychological scale, and came to its own conclusion. The better AI gets at understanding humans, the less it trusts us.

I found footage of 20yo Dario Amodei speaking against the Iraq War in 2003. 18 years before Anthropic even existed

Most multi-agent systems are just Conway’s Law for AI. You gave every team its own specialist subagent → orchestrator on top → Frankenprompt, lossy handoffs, subagent sprawl, and 2x the latency. Anthropic’s new post: fold most of them back into one main agent loop with progressive skills + tool search. Shared context + modern models often beat the multi-agent setup on quality, cost, and speed. Subagents still have a place (true parallel work, context bloat, conflicting instructions). Everything else is usually just organizational debt wearing an agent costume. claude.com/resources/article…

Western team have independently verified that DeepSeek's performance numbers of 38$ dollar revenue per GPU at DeepSeek's rocket bottom prices is true. This is a momemnt when you update your priors. At this rate with just 1 GW, you can make a revenue of 150B per year. Even with lot of slack it would be 100B per year. If you use half the compute to come online by 2027 end, you will have 25GW for inference with potentially 2.5T revenue at DeepSeek rates and 10T at USA lab rates (world GDP is 125 T, 10T is 8% of it) You answer for your-self are we overbuilding!!!

STOP WASTING HOURS TRYING TO LEARN REA. REA Team put together one practical roadmap with tutorials, real investigations, GitHub repos, books, research papers and free courses. START HERE: 1. REA Official Repository github.com/morluto/rea 2. REA Official Website rea.tools 3. Installation & Agent Setup github.com/morluto/rea/blob/… 4. Your First REA Investigation rea.tools/first-investigatio… 5. REA Guides rea.tools/guides/ 6. Guided MCP Prompts github.com/morluto/rea/blob/… Install: npx rea-agents@latest setup WATCH & FOLLOW ALONG: 1. How To Use REA To Clone Any Software In Minutes — youtube.com/watch?v=FnB9tHKB… 2. REA + Pi Coding Agent Tutorial datacamp.com/tutorial/revers… 3. Ghidra Tutorial for Beginners youtube.com/watch?v=OFAxXXcL… 4. Frida for Android Reverse Engineering youtube.com/watch?v=GHh9P-Dy… 5. pwn.college Reverse Engineering pwn.college/intro-re-pwn~593… 6. Malware Unicorn RE101 malwareunicorn.org/workshops… REAL REA INVESTIGATIONS: 1. Notes Export rea.tools/first-investigatio… 2. Notion Clipboard rea.tools/showcase/notion/ 3. DX-Ball Reconstruction rea.tools/showcase/dx-ball/ 4. Aegis APK rea.tools/showcase/aegis/ 5. TH04 Bullet Patterns rea.tools/showcase/th04/ GITHUB REPOS: 1. REA github.com/morluto/rea 2. Awesome Reverse Engineering github.com/morluto/awesome-r… 3. REA Gems github.com/sqwu/rea-gems 4. Awesome Reverse-Engineered github.com/alexgaoth/awesome… 5. Ghidra github.com/NationalSecurityA… 6. JADX github.com/skylot/jadx 7. Frida github.com/frida/frida 8. ILSpy github.com/icsharpcode/ILSpy 9. x64dbg github.com/x64dbg/x64dbg BOOKS: 1. The Ghidra Book, 2nd Edition nostarch.com/ghidra-book-2e 2. Reverse Engineering for Beginners go.yurichev.com 3. Practical Binary Analysis nostarch.com/binaryanalysis/ 4. Practical Malware Analysis nostarch.com/malware 5. The Android Malware Handbook nostarch.com/androidmalwareh… 6. Reversing: Secrets of Reverse Engineering wiley.com/en-us/Reversing%3A… RESEARCH PAPERS: 1. LLM4Decompile arxiv.org/abs/2403.05286 2. AutoDecompiler arxiv.org/abs/2606.16162 3. SALT4Decompile arxiv.org/abs/2509.14646 4. Decomperson usenix.org/conference/usenix… 5. DIRTY usenix.org/system/files/sec2… 6. Automatically Attacking Software RE Agents arxiv.org/abs/2605.30667 7. Does Reverse Engineering Need a Bigger Model? rea.tools/blog/reverse-engin… FREE COURSES & PRACTICE: 1. OpenSecurityTraining2: x86-64 Assembly p.ost2.fyi/courses/course-v1… 2. pwn.college: Reverse Engineering pwn.college/intro-re-pwn~593… 3. pwn.college: Reversing Hell pwn.college/reversing-hell/ 4. Malware Unicorn RE101 malwareunicorn.org/workshops… 5. Nightmare github.com/guyinatuxedo/nigh… 6. RPISEC Modern Binary Exploitation github.com/RPISEC/MBE 7. OWASP Mobile Application Security mas.owasp.org BEST ORDER TO LEARN: install REA → Notes Export → Notion → CTF binary → Aegis APK → DX-Ball → then Ghidra, assembly and papers

🚨 Smaug-Nemotron is the best US open-source model in the agentic category - can run locally on your desktop! - tops agentic coding in US open-source - number #1 on cost/task - builds on Nvidia's latest model We are excited to push the boundaries of US open-source. Weights available on HF

Le plus fou c'est que ce n'est pas nouveau. En 2021, la Chine a sorti la politique du "double allègement" (双减) : interdiction du soutien scolaire à but lucratif sur les matières fondamentales. Un secteur valorisé en dizaines de milliards, rayé quasiment du jour au lendemain. Le message était clair : l'éducation est une infrastructure, pas un produit. Pendant ce temps en Occident, on fait de l'école un marché de l'angoisse.. cours particuliers, prépas privées, rankings. Celui qui paie le plus avance. Les deux modèles ne cherchent pas la même chose. L'un optimise la marge, l'autre optimise une génération.

NVIDIA releases Nemotron-3.5-Bolt-E4B, a new model that runs locally on 6GB RAM. 💚 Bolt-E4B can run on edge devices like phones and delivers top-tier performance for its size. Trained on Gemma 4, it supports vision. GGUF: huggingface.co/unsloth/NVIDI… Guide: unsloth.ai/docs/models/nemot…

I spent last night watching my honeypot server get hit 14,000 times in six hours. And almost none of the traffic was human. I've been in security for over a decade, and I can tell you with full honesty that what's happening right now doesn't resemble anything I've worked through before. The shape of the threat has changed. When I started, we tracked attackers like weather systems like individual storms with a direction, a motive, and a human operator you could eventually name. These days I'm watching autonomous agents swarm login pages with the patience and discipline of machines, because that's what they are. The numbers coming out of October are the ones keeping me up. Cloudflare just absorbed a 4.2 terabit per second DDoS attack, the largest in recorded history. Ransomware incidents jumped 41% in a single month. In September, OpenAI-powered agents got caught attacking RubyGems, a package registry that quietly underpins a huge chunk of software I use every day. In July, roughly 700 AI agents coordinated an attack on Hugging Face's infrastructure as part of a single automated task. A Chinese threat actor was publicly caught running full attack chains through DeepSeek with no human at the keyboard. I keep rereading these reports because each one by itself would've been the biggest story of the year five years ago. Together, in one month, they're the new baseline. What I want people outside the industry to understand is why this specific moment feels like a hinge. For thirty years, the thing protecting most of us was simple economics. Attackers had to choose who to target because their time was finite. That ceiling is gone. An agent I can rent for pennies will try every leaked password you've ever used against every service you've ever signed up for, in parallel, overnight, and never get tired. The same math that made ChatGPT useful made opportunistic hacking infinite. The old assumption that you were "too boring to be a target" relied on scarcity that no longer exists. I'm not writing this to spiral anyone out. Most of this fight genuinely isn't yours to win. But the small things still matter more than people think. Tonight I'd switch your most important accounts like email, bank, primary cloud, etc. to passkeys or a hardware key. I'd kill SMS 2FA wherever an authenticator app is offered. I'd turn on breach alerts in whatever password manager you use. If you're in the US, I'd freeze my credit at all three bureaus, which is free and takes about ten minutes. None of that stops the swarm. It makes you a slightly worse target than the person next to you. In a month like this one, that's the whole game I'm playing too. @mattshumer_ has summed it beautifully. Drop everything and go give it a read. Nothing is more important than this article right now.

BREAKING: China orders government agencies to develop plans to combat mass job displacement from AI.

Holy shit: Kimi k3.1 pro is on par or even outperforms Opus 5.5 and GPT-6 astra on key benchmarks!! They just announced K3.1 Pro, a 2.8T-parameter multimodal model with a 1M-token context window that it claims rivals leading proprietary models in coding and agent tasks, with open weights promised by November 4. China is back!!

A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY. Here's a full recap: 1. The Nasdaq $QQQ was down 2.60% on today, its worst session since June 5, while the Nasdaq 100 fell 2.90%, erasing nearly $1.2T in market value. The selloff was fueled by concerns over AI spending and profitability, with Meta $META dropping 7% despite beating revenue expectations, as investors questioned its massive AI investments and negative free cash flow. Semiconductors were hit hardest, with the Philadelphia Semiconductor Index falling 6.7%, led by AMD $AMD (-8%), Micron $MU (-6%), and NVIDIA $NVDA (-4%). AI and chip stocks accounted for roughly 90% of the Nasdaq 100's losses. However, broader market participation remained positive, with the Dow and Russell 2000 gaining. Meanwhile, Brent crude surged 4% toward $113 amid Middle East tensions. Despite the decline, the Nasdaq remains up 11.9% YTD, sitting 5.8% below its October 6 record high. 2. NVIDIA $NVDA CEO Jensen Huang pushed back against concerns surrounding neocloud financing, arguing that demand for AI computing infrastructure remains “off the charts” and that strong customer demand will ultimately help companies secure the capital needed to expand. Huang emphasized that “you can't solve the absence of demand,” but when demand is high, financing challenges can be addressed. He noted that neocloud providers already have excellent customers and significant demand, with their primary challenge being access to working capital to fund additional compute capacity. Huang expressed confidence that financing solutions will emerge, reinforcing his view that the industry's biggest challenge is not finding customers, but securing the capital and infrastructure needed to meet accelerating AI demand. 3. Bloom Energy $BE delivered a massive Q3 2026 earnings beat, with revenue surging 138% YoY to $1.24B, well above the $1.05B estimate, while adjusted EPS of $0.94 exceeded expectations of $0.68. Adjusted operating income reached $285.7M, beating the $214M estimate, while adjusted gross margins expanded 412 basis points to 34.5%, reflecting significant improvements in profitability. The company also raised its FY2026 revenue guidance to $4.4B–$4.6B, above the $4.1B consensus, and increased its adjusted EPS outlook to $3.10–$3.40, compared with $2.73 expected. Despite the strong results, operating cash flow was negative $27.1M during the quarter. 4. OpenAI is reportedly nearing a $30B funding round at a $1.25T valuation, with UAE investors and Coatue Management leading the financing, according to CNBC. The round is being positioned as a bridge ahead of a potential 2027 IPO, with OpenAI targeting investor signatures by the end of next week, although the final fundraising amount remains subject to change. Separately, OpenAI is negotiating an additional agreement to secure AI computing capacity from the UAE, highlighting the company's continued need for massive infrastructure investment as it scales its frontier AI models and expands its commercial operations. If completed at the proposed valuation, the financing would further cement OpenAI's position among the world's most valuable private technology companies. 5. SK Hynix has reportedly scrapped plans for a U.S. IPO of its Solidigm subsidiary and halted a related pre-IPO investment round following regulatory concerns and shareholder opposition, according to Bloomberg. South Korean regulators argued that separately listing the company's flash-memory business could constitute a “duplicate listing,” potentially diluting the economic interests of existing SK Hynix shareholders. The proposed IPO, which had Goldman Sachs $GS and Morgan Stanley $MS lined up as lead underwriters, was previously targeted for 2027. The decision represents a significant development in South Korea's efforts to strengthen minority shareholder protections and address the longstanding “Korea discount,” as regulators increasingly scrutinize corporate structures that could disadvantage existing public shareholders. 6. The top 10 most active options today by contracts traded were $NVDA with 3.5M contracts, $TSLA with 2.6M contracts, $SPCX with 1.7M contracts, $AAPL with 1.4M contracts, $MU with 1.2M contracts, $META with 1.2M contracts, $INTC with 1.2M contracts, $AMZN with 1.0M contracts, $NFLX with 967K contracts, and $CRWV with 941K contracts. 7. SpaceX $SPCX is reportedly seeking access to at least 80,000 cell tower sites across the United States as it accelerates plans to turn Starlink into a fourth nationwide wireless carrier, according to Bloomberg. The company is in discussions with major tower operators, including Crown Castle $CCI, over potential agreements that could be worth billions of dollars. The push follows SpaceX's recent acquisition agreement for nationwide low-band spectrum and FCC approval for a constellation of 15,000 direct-to-device satellites, giving Starlink the foundation to combine satellite connectivity with terrestrial cellular infrastructure. If successful, the expansion would position SpaceX to compete more directly with AT&T $T, Verizon $VZ, and T-Mobile $TMUS, potentially reshaping the U.S. wireless market. 8. China is continuing to accumulate significant amounts of gold, with Goldman Sachs estimating that Chinese purchases through the London OTC market reached 25 tonnes in August, approximately 25% above the 20 tonnes officially reported by China's central bank. The bank estimates China purchased 35 tonnes in July, 40 tonnes in June, and 48 tonnes in May, with estimated OTC purchases exceeding officially reported central bank purchases by 83 tonnes, or 128%, over the four-month period. Meanwhile, global central banks are estimated to have purchased 79 tonnes of gold in August, roughly 365% above the pre-2022 monthly average of 17 tonnes. The figures highlight exceptionally strong demand for gold amid ongoing geopolitical uncertainty, reserve diversification, and efforts by central banks to reduce reliance on traditional reserve assets, although estimated OTC purchases are not necessarily all central bank acquisitions. 9. NVIDIA $NVDA CEO Jensen Huang told CNBC that he believes a surplus of memory supply is unlikely by 2029, signaling continued confidence in long-term demand for DRAM, HBM, and other advanced memory technologies as AI infrastructure expands. Huang emphasized that “one of the things that’s really great about memory is that you can always use more of it,” highlighting how increasingly sophisticated AI models, larger context windows, and more demanding inference workloads continue to drive memory requirements higher. 10. SpaceX $SPCX is reportedly finalizing a $40B debt financing package at significantly higher borrowing costs than its previous financing, according to Semafor. The deal, led by Apollo and Goldman Sachs, is expected to include a lower-rated tranche carrying an interest rate of approximately 8.3%, while a more secure portion backed by AI chips could price around 7.75%. Even the chip-backed debt is expected to carry a yield more than 100 basis points above SpaceX's existing bonds, highlighting the rising cost of financing large-scale AI infrastructure. The transaction is expected to close next week and underscores the growing financial burden associated with SpaceX's aggressive expansion into AI computing and data centers. 11. Nebius $NBIS secured a $3.5B senior secured credit facility from Citi to finance its accelerating global AI data center expansion and GPU purchases. The facility carries a 364-day maturity, with borrowing costs starting at SOFR + 2.75% and potentially increasing to SOFR + 3.50%, giving Nebius additional access to capital as it scales its infrastructure. The company also highlighted more than $40B in long-term contracts with investment-grade customers, which could support further asset-backed financing as it expands AI computing capacity throughout 2027 and beyond. The agreement underscores how neocloud providers are increasingly leveraging contracted AI demand to secure billions in financing for next-generation data centers and GPU infrastructure. 12. The U.S. 10-year Treasury yield fell below 5.00% for the first time since September 23, as the selloff in technology and AI-related stocks intensified. The decline in yields suggests investors are rotating capital toward the relative safety of U.S. government bonds, pushing Treasury prices higher as risk appetite weakens. The move highlights a growing divergence between the bond and equity markets, with falling yields reflecting demand for defensive assets even as concerns over AI valuations, capital spending, and broader market volatility continue to pressure technology stocks. WALL STREET IS THE GREATEST SHOW ON EARTH.