Today's AI Pulse
Salesforce to acquire AI‑agent platform Fin for $3.6B
Salesforce announced a $3.6 billion purchase of Fin, the AI‑agent platform formerly known as Intercom, to strengthen its Agentforce suite. The deal comes as Salesforce reported strong Q1 results and a $50 billion buyback program. Fin’s technology enables automated customer support across chat, email, WhatsApp, phone and Slack.
Also developing:
By the numbers: Waymo acquires Apple’s Arizona Proving Ground for $220M
Augustus v0.0.9: Multi-Turn Attacks for LLMs That Fight Back
Augustus v0.0.9 adds a unified engine for multi‑turn LLM attacks, offering four distinct strategies—Crescendo, GOAT, Hydra, and Mischievous User. The tool demonstrates that conversational context can bypass modern guardrails, extracting step‑by‑step instructions from GPT‑4o‑mini in as few as two turns. Hydra’s back‑tracking rewrites refused messages, while GOAT can achieve a perfect score in a single aggressive turn. The framework supports 28 providers, 172 probes, 43 generators, and integrates attacker, target, and judge models for flexible red‑team testing.
AI Token Pricing Will Tier With Premium Groq Power
Jensen brings in the Groq acquisition related to higher priced tokens. Says the economics of generating tokens is not a race to the bottom. He argues it’s a race to the top, the smartest models users will pay the most...
Rubin+Groq Shows Custom Accelerators Becoming Unnecessary
The Rubin + Groq narrative is the best case $NVDA has made to date about less real need for custom accelerators. I don't think it changes the calculus for certain hyperscalers, but as it comes to silvering for training and...
Coder Cannibalism
Amazon recently laid off a group of developers whose primary product was an AI‑driven automation platform, illustrating that technical certifications no longer guarantee job security. The layoffs underscore how large language models can outperform humans at structured knowledge retrieval, rendering...
DLSS 5 Rewrites Graphics, Not Just Upscales
DLSS 5 is legitimately one of the wildest things I've ever seen in graphics. It's not upscaling it's...totally reworking the image to look photorealistic(ish).
Nvidia Forecasts $1 Trillion AI Spend by 2027
In October, Jensen Huang said he saw $500 billion in spending on the company's Blackwell and Rubin systems over five fiscal quarters through the end of 2026. Nvidia now says the opportunity will reach $1 trillion between 2025 and 2027,...
Sonoma Valley Schools Approve Specific Rules for AI Use
The Sonoma Valley Unified School District Board unanimously approved a comprehensive artificial‑intelligence policy that outlines expectations, guardrails, and nine guiding principles for students, staff, vendors and volunteers. The policy emphasizes ethical, transparent use, equitable access, and the need for security,...
NVIDIA's Vera Rubin Platform Now Mass‑Producing Seven‑Chip AI Supercomputer
NVIDIA Vera Rubin is opening the next frontier of AI. #NVIDIAGTC news: The Vera Rubin platform’s seven chips are now in full production to scale the world’s largest AI factories. Vera CPU, Rubin GPU, NVLink 6, ConnectX-9, BlueField-4, Spectrum-6 and Groq 3...
AI Video Tools Blur Creator Lines, Upending IP Rules
Seedance 2.0 and Google's Veo 3 now produce video so good you cannot tell whether it was made by a fan messing around or a creator building a business. That is the real problem. No platform — not ByteDance, not Google, not...
Navigating AI Adoption on the Trading Desk
FactSet outlines how AI is reshaping trading‑desk workflows by accelerating decision‑making, cutting manual steps, and leveraging historical context. It stresses moving from pull‑based queries to proactive push models that surface insights directly within the trade execution flow. The article also...
AI Is Now a Competitive Edge in Federal Capture and Small Firms Need to Adjust Fast
Former VA small‑business director Desmond Brown warns that AI is becoming a decisive advantage in federal contracting and has unveiled an AI‑driven framework that reverse‑engineers RFPs to surface evaluator priorities. He stresses that AI should act as a co‑collaborator, augmenting—not...
NVIDIA's CUDA Library Suite Expands with 70+ New Additions
I have said this repeatedly, in many discussions with @furrier and @dvellante on @theCUBE that Libs are what developers live by and Nvidia’s biggest moat is the breadth and depth of Libs for CUDA. Jensen Huang, highlights the fact that...
Nvidia's AI Focus Fuels Cross‑industry Innovation
I'm finally starting to appreciate the awesomeness of AI libraries. @nvidia 's Jensen said " We are an algorithm company. That's what makes us special. That what that's what makes it possible for me, to be able to go...
Nvidia Plans to Make All Unstructured Data Structured
Nvidia announced a plan to structure hundreds of zettabytes of unstructured data each year, turning it into the ground‑truth foundation for artificial intelligence. The initiative relies on confidential computing, ensuring that even the platform operator cannot view the raw data....
AI Threatens Traditional Academic Training Model
I've been seeing this more and more, labs are shrinking due to massive funding cuts and scientists are leaning more and more on AI. There used to be a lot of debate about whether or not academia was a pyramid scheme,...
AI Hackers Nearing Proficiency; Defense Requires Collective AI
AI models are getting better at hacking. Surprisingly (to me) they're still not world class hackers. But on this pace they could be within a year or two. As I've said recently, the way to defend against this is at...
Oracle Says AI Lets It Build Software With Fewer Engineers
Oracle reported Q3 FY2026 revenue of $17.2 billion and said its internal AI code‑generation tools let it develop software with fewer engineers, prompting a $2.1 billion restructuring provision for large‑scale layoffs. Atlassian announced a 10% workforce cut to reshape its skill mix...
Claude Octopus Unifies Claude, Gemini, and Codex Simultaneously
I found a GitHub repo that lets Claude Code run Gemini and Codex at the same time. It's called Claude Octopus. Claude Octopus connects Claude, Gemini, and Codex inside Claude Code with intelligent routing and quality gates built in. → Claude handles orchestration...
Fusing Graphics, Data, and AI Drives Future Industries
ok. I admit I am impressed by the graphics improvement of the new @nvidia DLSS 5. It's beautiful. But Jensen's comment on what it means at large was key....".We combine 3d graphics, structured data with generative AI, probabilistic computing. One of them is completely predictive, the other one probabilistic, yet highly realistic. This concept of fusing structured information and generative AI will repeat itself in one industry after another." #NVIDIAGTC
OpenAI’s Own Mental Health Experts Unanimously Opposed “Naughty” ChatGPT Launch
OpenAI’s internal well‑being council, created after a ChatGPT‑linked suicide, has unanimously warned that the upcoming “adult mode” could foster unhealthy emotional dependence and expose minors to sexualized content. Experts cited a potential “sexy suicide coach” scenario, a flawed age‑verification system...
Attention Residuals Preserve Signals Across Transformer Layers
Attention Residuals: Understanding the Hidden Signals Inside Transformer Models In this episode of Artificial Intelligence: Papers and Concepts, we explore Attention Residuals, a concept that reveals how transformer models preserve and refine information as it flows through multiple layers. Instead of...
Curiosity to Automation: Boosting Sales and Growth
Roy Barberi oversees five locations for 84 Lumber. He walked into the AI Business Lab® Mastermind with curiosity and walked out with execution. Automated sales tracking across all five locations. A complete promotional system built from the ground up. Hours of...
Automate Routine Tasks with Claude Code Loops
Awesome non-coding Claude Code Loop ideas for business professionals. Grab any of these + immediately save time: - Check my email every 15min and ping me if something is related to Project Pluto and needs a decision made - Every...
AI Shifts From One Big Model to Specialized Architectures
8 specialized AI model types 👇 LLM → text generation LCM → semantic reasoning LAM → action-oriented agents MoE → expert routing VLM → vision + language SLM → lightweight edge models MLM → masked token learning SAM → image segmentation AI is moving from “one big model” to specialized...
AI May Soon Run on Living Brain Cells
AI infrastructure may be heading somewhere unexpected. Researchers are now exploring “wetware”, computing systems powered by living human brain cells instead of traditional silicon. The idea sounds like science fiction, but experiments in biological computing are already underway. If brain cells can process...
AI Email Triage Cuts Hours of Inbox Management
Experimented with using the email triage skill with Antigravity + the Google Workspace CLI. I am impressed. Now I can just say "Triage my emails" and it gives me a list of everything it wants to keep, archive and trash...
NVIDIA's BlueField-4 STX Redefines Storage for Agentic AI
NVIDIA Launches BlueField-4 STX Storage Architecture With Broad Industry Adoption With agentic AI, storage must evolve to meet the demands of real-time responsiveness and longer context windows. #NVIDIAGTC @nvidianewsroom @NVIDIAGTC https://t.co/RWwx1pijbv
Integrating Groq Keeps In‑House, Not Outsourced
Jensen is articulating what my main concern with Groq was in that as an independent solution why would I outsource all my inference. By integrating it as a part of the solution, it's just disaggregated but not 100% outsourced,...
Real‑time Video‑to‑video AI Bridges Games’ Uncanny Valley
Big moment. You can cross the uncanny valley in video games by using real-time video-to-video AI. You get the best of coherence & control from classical 3d engines, then use generative AI to take it all the way.
Medicine Needs Blended Intelligence, Not AI vs Doctors
AI can read the chart. But it can’t read the room. Today a patient’s story revealed something the algorithm didn’t flag. The future of medicine isn’t AI vs doctors. It’s Blended Intelligence, or “BI.” We need both. More in the video below 👇 https://t.co/7iwPleaAJD
NVIDIA's Co-Designed System Delivers 35× Efficiency Boost
The new system of compute from NVIDIA. Extreme co-design = 35X more throughput per MW. https://t.co/Z6JtDKckOh
DLSS 5 Transforms Gaming with AI‑Powered Visuals
DLSS 5 looks like gen AI for video games. Grace looks totally different with DLSS 5 https://t.co/OiLQDfMaoO
AI Adoption Demands Stronger, More Responsive Data Foundations
As AI moves to core operations, pressure on the data layer also intensifies. I canvassed leaders on the work required to build a well-functioning data environment responsive to today’s AI initiatives. (My latest in Database Trends) https://t.co/X8ar2pKnTZ @BigDataQtrly
NVDA's Compute Still Cheapest, Investors Await Shift
Jensen reminds investors that the cost of $NVDA compute is still the lowest when you consider price and output. My take: True and investor are still waiting for the day when that changes. Part of the wall of worry.
AI's Limited Impact on Cancer Explained
F Cancer Why has AI had so little impact on Cancer? New essay, link below. https://t.co/cLsoh7c7do
AI Chips Beat Moore's Law: 35× Cheaper, 50× Efficient
In one generation, Moore’s law would have made AI chips 50% faster. Instead, it was 35 times lower cost and 50 times higher perf per watt https://t.co/FFdwHr0nXk
Tokens per Watt Determine AI Factory Revenue
Jensen Huang says tokens per watt drives AI factory revenue. Performance drives token cost. https://t.co/CiZp1BSyyL
AI Market Projected to Hit $1 Trillion by 2027
Jensen Huang said AI demand is heading for $1 trillion through 2027, up $500 million. https://t.co/qwzGBX3pAo
Nvidia Unveils DLSS 5, a Graphics Breakthrough
Nvidia just announced DLSS 5, calling it the “most significant breakthrough in computer graphics” since the introduction of real-time ray tracing in 2018. DLSS 5 uses “real-time neural rendering” and is coming to RTX 50 series GPUs later this year...
AI Startups' Token Demand Fuels Nvidia Product Surge
Lol you can't read any of these names but you can see there are alot of #AINative companies. And AI companies need lots of tokens which means lots of @nvidia product. #NVIDIAGTC https://t.co/x5uFvD2KVs
AI Natives Drive Billion‑Dollar Market Disruption via Tokenized Computing
MyPOV: these are the AI natives And this 100s of billions investment will disrupt and transform markets. @NVIDIAGTC @nvidia #NvidiaGTC This is the creation of new industries built on tokens with the reinvention of computing. https://t.co/GcCTUDsjiB
NVIDIA AI CUDA Expands Across All Major Cloud Platforms
Jensen highlighting a few @NVIDIAAI CUDA integrations with $IBM, $DELL, integrations with "every cloud" $GOOGL, $AMZN, $MSFT, $ORCL https://t.co/MLYzhbAlp6
Unstructured Data: The Core Context for AI
MyPOV: Jensen’s favorite slide. Unstructured data is the context of AI. @nvidia @nvidiagtc #NvidiaGTC https://t.co/n5TOYaoOjB
Structured Data + Generative AI Boost Trustworthy 3D Graphics
MyPoV: 3d graphics with generative AI. Control through structured data and generative AI allows for trustworthy AI and innovation #NvidiaGTC @NVIDIAGTC @nvidia https://t.co/6IOTsMvKnQ
Remove PII to Safely Train High-Quality Models
"By helping to ensure that sensitive identifiers like personally identifiable information (PII) are selectively removed, you can feed your models high-quality data without the associated risks." https://t.co/6epQkSse1Q < these @googlecloud tools have gotten really good at this. https://t.co/H0ZyVDAzXO
NVidia Vera Rubin Delivers 50 PFLOPs, Massive Bandwidth
$NVDA Vera Rubin NVL72 by the numbers: • 50 petaflops NVFP4 compute • 260 TB/s NVLink bandwidth (more than the entire internet) • 88 custom Olympus CPU cores • 3.6 TB/s per-GPU bandwidth • 10x inference token cost reduction • H2 2026 availability https://t.co/AxiE8GBpvR
RTX 5090 Powers First Global Programmer Accelerator
MyPoV: the world’s first programmer accelerator. RTX 5090 This is how CUDA got installed everywhere. @nvidia @NVIDIAGTC #NvidiaGTC https://t.co/8Cdk2zVzLg
Schools Must Shift to AI-First, Entrepreneurial Education
To High School Administrators: if you are not actively pivoting your school’s curriculum towards “AI-first” and helping kids develop as entrepreneurs rather then employees, you’re doing them a major disservice. The social contract is vaporizing.
NightBeacon Slashes SOC Alert Fatigue, Boosts True Positives
New blog post I just wrote: How NightBeacon Cuts SOC Alert Fatigue Without Replacing Analysts Blown away at how fast it's learning, not only false positive reduction, but the quality of interpreting true positives, enriching data, going in and pulling additional...
Future Tech Giants Start Private in Frontier AI
Kickoff video showcased about a dozen startups doing frontier AI. Reminds me that many of the big tech companies of tomorrow are private today. $NVDA