Find answers instead of hunting for files - Google With Workspace Intelligence, Drive becomes a living knowledge base. Easily locate files or info with AI Overviews, and ask deeper, conversational questions about your files with Ask Gemini in Drive. https://lnkd.in/dVyxdPHX

Find answers instead of hunting for files - Google With Workspace Intelligence, Drive becomes a living knowledge base. Easily locate files or info with AI Overviews, and ask deeper, conversational questions about your files with Ask Gemini in Drive. https://lnkd.in/dVyxdPHX

AI is changing threat hunting, but not every AI tool is helping defenders. Coalfire's Cory Marchand discusses the ideas behind Threat Hunting 101 (in an AI World!), including how threat hunting is adapting as AI becomes part of the enterprise and the risks of unsanctioned AI tools. Watch Cory’s interview and explore the full Black Hat wrap-up: https://lnkd.in/eV_9Diei

AI is changing threat hunting, but not every AI tool is helping defenders. Coalfire's Cory Marchand discusses the ideas behind Threat Hunting 101 (in an AI World!), including how threat hunting is adapting as AI becomes part of the enterprise and the risks of unsanctioned AI tools. Watch Cory’s interview and explore the full Black Hat wrap-up: https://lnkd.in/ejDr-V9K

Spent this week hunting a quiet but critical bug in Aura's scoring logic — one where genuinely urgent items could slip through as "no action needed." Root-caused it, added consistency checks, wrote tests around the exact failure cases, and verified the fix against a real connected inbox. This is the unglamorous part of building AI products: not the demo, but making sure the system tells the truth when it matters. More of this to come.  #AgenticAI #AIInfrastructure #StartupFounders #Privacyfirst

My 30-sec AI firefight workflow: (GPT img 2.0 + Seedance 2.5) → Create the main character sheet → Design one consistent weapon reference → Build the full battleground topology → Lock all enemy positions + attack directions → Write the 30-sec action with cuts, reactions, and counterattacks → Generate only after continuity, geography, and performance are fixed

Silicon Valley may end mathematical trophy hunting by making sport of it. But the modern mathematical landscape is so much broader than that.... Read the article on - https://lnkd.in/dw5fjr_f #algorithms #artificialIntelligence #at #books #data #openai #reviews #science

One idea from LLMs keeps coming back to me: word embedding. Language models became much more powerful when words stopped being treated as isolated symbols and started being represented through context and relationships. I think geography may need a similar step. Not just semantic labels like “river”, “road” or “forest”, but geospatial embeddings where places, objects, terrain and their relations are represented inside the same knowledge space. That could give robots something much richer than a map: a way to reason about geography through context, similarity and relationships.

10 things people have already built with it: 1. Jev-Trader: a live trading bot on Monad, 1,911 stars t.co/ZXeOzS5zDJ 2. Jev-Ultrafast: a browser agent that decides every click, 16,758 stars t.co/QuXLmNQcF1 3. Jev-Doom-Agent: plays Doom by making tactical decisions every frame t.co/Itgfre75QR 4. Jev-T-Rex-Runner: plays the Chrome dinosaur game on its own t.co/gjJNZxaQR0 5. Typesafe-Chess: Jev vs a search engine in chess t.co/crzaW2KwCo 6. Jev-Drone: flies a simulated drone through an obstacle course t.co/uP4SAD8kFT 7. Tax-Doc-Classifier: sorts IRS forms at roughly $0.001 per page t.co/LkiR61i6r1 8. KillMyIdea: scores startup ideas and returns kill, fix or ship t.co/cQ113cIfCC 9. Jev-Curate: filters 1,500+ data rows per second t.co/Y2eyrlnf8i 10. PG-Jev: lets you ask your Postgres tables questions in plain English t.co/xJTqYqhS6N #jev #aiagent #aitools #trending

Your AI agents are the new privileged insiders. I just open-sourced a KQL hunting pack for Defender XDR: "Configured ≠ observed ≠ validated." It inventories every AI agent Defender can see, flags the ones shared with your entire tenant, and mines the exposure graph for AI-agent findings — one row per asset, signal not noise. 6 queries, syntax-validated, MIT licensed: → Agent inventory → Tenant-wide sharing (the blast-radius list) → Exposure graph discovery → Findings per asset → Access posture summary https://lnkd.in/dWHksjMm If you're running Copilot agents in production, query 02 shows which ones every user can reach. That's the list I'd review first. #ThreatHunting #KQL #DefenderXDR #AISecurity #ExposureManagement