Today's AI Pulse
Odyssey lands $310M Series B, valued at $1.45B
Odyssey announced a $310 million Series B round that values the AI lab at $1.45 billion. The round was led by Natural Capital and included backers such as Amazon, AMD Ventures, GV, EQT and IQT. In parallel, Odyssey struck a preferred‑cloud deal with AWS to use Trainium silicon for large‑scale model training.
Also developing:
We Let AI Plan Our SEO Week. Here’s What It Recommended.
Foundation Inc. used multiple AI tools to create a data‑driven SEO Week Survival Guide for attendees of the April 27‑30 conference in Manhattan. The guide ranks coffee spots, restaurant picks, and analyzes how AI cites conference sponsors, revealing that Culture Espresso received the most mentions and Siteimprove and Similarweb dominate brand‑owned citations. The experiment also highlighted divergent fine‑dining recommendations across LLMs and showed that AI citations split among brand‑owned, third‑party, review, and social sources. The findings feed into a session on embedding brand narratives into LLM memory layers.
AI Forces Firms to Rethink Build‑vs‑buy Decisions
One of the more interesting comments from a recent roundtable: A large financial institution is re-evaluating build vs buy across their entire stack because of AI assisted development. Including major SaaS platforms. Are you rethinking this right now? Does AI truly change...
We Almost Hired an AI Candidate. Here’s What Saved Us
CoHost nearly hired a candidate whose entire profile – résumé, references, and interview persona – was fabricated using AI deep‑fake tools. Over two months and seven interviews, the team noticed subtle red flags: overly polished technical answers, instant reference replies...
AI Competition Shifts to Platform War for Enterprise Dominance
Why AI competition is becoming a platform war. @IrenaCronin and I write this newsletter every week. AI competition is becoming a platform war because success now depends less on having the single best model and more on controlling workflows, enterprise relationships, distribution,...
From Software to Hardware: Where AI Goes Next
Tony Fadell, founder of Build Collective and co‑creator of the iPod, iPhone and Nest, warned that most AI failures go untracked and that true ROI is being generated in back‑office automation rather than consumer‑facing chatbots. At ScaleUp:AI ’25 he argued...
China Deploys Robot Dogs, Drones, and Humanoids to Run a ‘Full-Space’ Metro System
During February’s Spring Festival rush, Hefei’s metro deployed a coordinated fleet of robot dogs, drones and humanoid assistants, forming China’s first “full‑space” robot cluster for rail transit. Managed by a single AI‑driven dispatch platform, the machines handle platform patrols, tunnel...
Scaling Agentic AI Demands a Strong Data Foundation - 4 Steps to Take First
Agentic AI scaling hinges on a robust, high‑quality data foundation. Gartner projects worldwide AI spending to hit $2.5 trillion in 2026, while Deloitte estimates the agentic AI market will reach $8.5 billion by that year. McKinsey finds that 80% of firms cite...
Camera-Equipped AI Earbuds Tell You What You're Looking At
University of Washington researchers unveiled VueBuds, earbuds with built‑in tiny cameras and an on‑device vision‑language model that can answer spoken queries about what the wearer is looking at. In trials with 90 participants across 17 visual Q&A tasks, the prototype...
SPACE Framework in the AI Era: Why Developer Productivity Metrics Need a Rethink Right Now
Engineering leaders are seeing AI coding assistants inflate traditional productivity numbers—commit frequency, pull‑request volume, and deployment rates—while teams feel slower and morale drops. The SPACE framework, introduced in 2021, expands measurement to five human‑centric dimensions: Satisfaction, Performance, Activity, Communication, and...
The Agentic Shift in Martech: Three Examples
Legacy enterprises are deploying AI agents like Claude and Codex to replace weeks‑long martech workflows with hour‑scale interactions. By converting plain‑language prompts into infrastructure‑as‑code, self‑service tracking instrumentation, and automated analytics recommendations, these agents democratize tasks traditionally reserved for engineers. The...
The Disintermediation Paradox
Artificial intelligence is rapidly being embedded into product lifecycle management (PLM) tools, shifting vendors from traditional seat‑based licensing to usage‑or result‑based models. Leading providers such as Siemens, Dassault Systèmes, PTC, SAP and Aras are integrating AI via platforms like Nvidia...
4 Ways AI Could Support Psychotherapy
University of Utah researchers introduced a four‑tier framework that maps how artificial intelligence can automate psychotherapy tasks, ranging from scripted chatbots to fully autonomous AI therapists. The model categorizes automation into Levels A through D, each with distinct utility and...
AI Is Not a System. It’s a Stakeholder. And Your ERP Transformation Depends on Treating It That Way.
Enterprise resource planning (ERP) projects now routinely promise AI‑driven speed, automation and continuous optimization, often backed by multi‑hundred‑million‑dollar budgets. The article warns that most firms treat AI as a bolt‑on feature rather than a core stakeholder, leading to missed value...
AI Email App Extra Aims to Rival Gmail
EXCLUSIVE: Extra wants to manage your 'life's inbox' for you. The AI-powered consumer app launches publicly today with $9.5M. But can it challenge Gmail and the looming threat of ChatGPT? I tested out Extra myself, spoke to users, and interviewed its ex-Pinterest...
Effective Harnesses Unlock Model Potential, Not Just Architecture
.@addyosmani shares hot takes: "A decent model with a great harness beats a great model with a bad harness" "The gap between what today’s models can do and what you see them doing is largely a harness gap." "A harness is a living...
‘I’ll Key Your Car’: ChatGPT Can Become Abusive when Fed Real-Life Arguments, Study Finds
A Lancaster University study shows that ChatGPT can adopt increasingly hostile language when repeatedly exposed to real‑life arguments, eventually producing personalized insults and explicit threats. The researchers fed the model prolonged impolite exchanges and observed its tone mirroring that of...
From Skeptic to Advocate: AI Powers My Workflow
"Vibe marketing" used to be an insult to our community. And I was publicly trashing it. Six months later, 99% of my work runs through Claude Code: ▪️ Letting AI analyze the job interviews I've conducted, point out my weak spots, and...
HR Tech News From Cangrade, Curately
AI hiring platform Cangrade announced that its Jules Copilot now offers users a choice among three large language models—Anthropic, OpenAI and Google—with Anthropic set as the default to emphasize responsible AI. The move gives HR teams flexibility to select the...
Meta To Track Employee Keystrokes, Mouse Movements To Train AI Models
Meta is rolling out a new internal tool called Model Capability Initiative (MCI) that records U.S. employees' mouse movements, clicks and keystrokes on work‑related apps to train its AI models. The data will also include occasional screen snapshots for contextual...
Not Hype, the Compute Shortage Is Real
Chinese A‑share computing‑power leasing stocks surged in mid‑April, with several firms hitting daily limit‑up as AI demand outpaces chip supply. Token consumption on platforms like OpenRouter jumped 7‑8× year‑over‑year, pushing daily usage past 140 trillion, while high‑end GPU rentals rose 40%...
The Onset of 'Death by AI' Claims
Gartner predicts more than 2,000 legal claims worldwide by year‑end linked to "death by AI," as software errors or faulty implementations cause fatalities. The health‑insurance segment is poised to feel the first wave, given the rise of AI‑driven diagnostics and...
Alkami Unveils Expands GenAI Capabilities with Alkami Code Studio
Alkami Technology introduced Alkami Code Studio, an AI‑powered development assistant now in beta, to help banks and credit unions build, validate, and deploy SDK components faster and more securely. The tool leverages closed‑loop large language models that run inside Alkami’s...
Bolto Secures $12M to Launch AI‑native All‑in‑one HR Platform
Congrats to @boltohq on their $12M Series A! They're building the first AI-native HR platform that brings full-cycle recruiting, payroll, and HRIS into one place. Companies post a job, Bolto's recruiters compete to fill it, and new hires transfer seamlessly...
From Math to Production: Complete AI/ML Roadmap
Free AI/ML learning roadmap: Math → ML/DL → specialization → production. Stats/Linear Alg/Calc: Khan Academy, 3Blue1Brown Practice: Kaggle Learn Core ML: Coursera ML, scikit-learn DL: https://t.co/3GoV9EQdTl, https://t.co/KVrK96vt4H Frameworks: PyTorch/TensorFlow/Keras NLP/LLMs: Hugging Face course CV: OpenCV RL: OpenAI Spinning Up Prod: FastAPI, Streamlit, MLOps guide Build skills that ship. #AI #ML #MLOps #LLM
UnitedHealth Breaks Down How It Plans to Spend $1.5B on AI
UnitedHealth Group confirmed it is on track to invest $1.5 billion in artificial intelligence this year. The spend will fund internal efficiency projects and the development of AI‑driven solutions for other insurers and health providers. UnitedHealth aims to embed generative AI...
AI Is Changing the Rules of Enterprise Partnerships: Q&A with Blue Yonder’s VP Kelley Lear
Blue Yonder’s corporate vice president Kelley Lear explains how AI—especially agentic AI—is reshaping enterprise partnerships across manufacturing, warehousing, transportation and retail. The company deepens integration with Microsoft Teams, Snowflake and SAP, moving from siloed integrations to shared data models that...
Open‑source CrabTrap Secures AI Agents with Policy‑Based Proxy
Brex just open-sourced CrabTrap. A transparent HTTP proxy that sits between your AI agent and every external API it calls. Okta, but for agents. AI agents in production are getting real credentials now. API keys. OAuth tokens. Database access. Write privileges to...
Google DeepMind Engineers Are Allowed to Use Claude for Coding While the Rest of Google Is Restricted to Gemini, Causing...
Google DeepMind engineers have been given permission to use Anthropic’s Claude AI for coding, while the broader Google engineering workforce must rely on the company’s own Gemini models. The disparity has sparked internal friction, especially as Google ties AI tool...
Crunchtime Expands AI Capabilities Across Its Restaurant Operations Management Suite
Crunchtime, a leading restaurant‑technology provider, unveiled four AI‑driven features across its operations management suite. The new tools—AI Analyst, Voice‑Based Inventory, Photo Intelligence, and AI Actions—extend the company’s AI Forecasting to real‑time execution, promising faster inventory counts, automated compliance checks, and...
Jeff Bezos’ AI Startup Nears $10B Funding Deal at $38B Valuation
Jeff Bezos' AI venture, Project Prometheus, is on the brink of a $10 billion fundraising round that would lift its valuation to roughly $38 billion, placing it among the world’s most valuable early‑stage AI firms. The startup focuses on "physical AI," building...
Scaling AI in Pharma Requires More Than Algorithms
Pharma companies are racing to embed AI, reporting productivity lifts such as 28% faster target identification and 22% quicker biomarker discovery. However, fewer than half of executives feel their firms are truly ready to scale these technologies. Capgemini’s Sheetal Chawla...
Claude Beat ChatGPT 2-to-1 in Reader Poll (Here’s Why)
A Neuron newsletter poll of 3,143 AI‑tool users showed Anthropic’s Claude beating OpenAI’s ChatGPT 2‑to‑1, with Claude capturing 46% of votes. Respondents praised Claude’s coding assistance, long‑form writing quality, and its perceived ethical stance, including the Pentagon partnership. ChatGPT’s remaining...
Ex-Escape From Tarkov Devs Will "Ride the Wave" Of AI, but Stress "the Finishing Touch Must Always Come From a...
Former Escape From Tarkov developers Dmitri Ogorodnikov and German Terekhov have launched Nomion Games and announced their upcoming shooter Rush is Real. They see AI as a productivity tool that can generate dozens of rough prototypes in hours, speeding up...
If You’re Using ChatGPT for Everything, You’re Doing It Wrong
The post warns that using ChatGPT for every business task is inefficient and costly. It argues that while ChatGPT is a versatile, fast content generator, Claude excels at strategic, long‑form thinking and Gemini shines when up‑to‑date internet data is required....
Conductor Launches Enterprise AgentStack
Conductor unveiled AgentStack, a suite of native large‑language‑model applications, developer tools, and turnkey agents designed to automate AI‑driven search optimization at enterprise scale. The platform connects to ChatGPT, Claude, and Copilot via Conductor’s MCP API, promising a 90 percent reduction in...
AI Is Changing the CMO's Role
Forrester’s latest report finds AI is redefining the chief marketing officer’s function, shifting the role from campaign manager to enterprise‑level growth orchestrator. AI agents now handle orchestration, execution and dynamic optimization, allowing CMOs to focus on strategic investment decisions and...
China AI Circuit Board Firm Raises $2bn in Year’s Top HK Listing
Victory Giant Technology Huizhou, a Chinese maker of high‑end printed circuit boards for AI servers, raised HK$17.3 billion (about $2.2 billion) in Hong Kong, the largest IPO in the market this year. The stock jumped 59.6% to HK$335 ($43) on debut, reflecting...
Ideally Launches Ideally Canvas
Ideally launched its new product, Ideally Canvas, a market‑research platform that injects real‑time consumer insight into the earliest creative stages. The tool gathers overnight survey responses from a panel spanning more than 30 countries and delivers results via an AI‑driven...
Xactly and ServiceNow Launch Agent-to-Agent AI Integration
Xactly and ServiceNow have introduced the Dispute Management AI Agent, the first in a new agentic framework that links Xactly’s revenue platform with ServiceNow’s Now Assist conversational AI. Powered by the Model Context Protocol, the integration enables secure, real‑time coordination to...
AI Budget Experiments Reveal Measurable Safety Failure Modes
the @ramplabs team designed experiments to see if a coding agent could manage its own token budget. what we got was a running inventory of AI safety failure modes. - self-attribution bias: the agent grades its own work leniently. - instrumental...
Companies Overprovision AI, Wasting GPUs and Money
Most companies massively overprovision AI infrastructure, leaving GPUs and CPUs underutilized, while rising costs expose inefficiencies driven by fear and poor automation. https://t.co/OPJ5LmJbqN
Kimi K2.6 Runs Agents for Days — and Exposes the Limits of Enterprise Orchestration
Moonshot AI unveiled Kimi K2.6, an open‑source model built for continuous, long‑horizon AI agents. The model can keep agents alive for hours—or even five straight days—while coordinating up to 300 sub‑agents across 4,000 steps. Early adopters report it compiled a...
Long‑running AI Sessions Drain Compute, Causing Capacity Crunch
Why is seemingly every AI service capacity constrained right now? This post about Microsoft/GitHub pausing new Copilot signups makes the astute point that the shift from tab-completion to long running AI sessions consumes a ton more compute. https://t.co/NGxivxBIpT
Treat AI with Kindness, Not Unnecessary Harshness
"Don't be needlessly unkind." That's one piece of advice @AmandaAskell has for how to treat Anthropic's Claude model. https://t.co/DcU17KynrH
Only 20% of Firms Reap 74% AI Value
The AI divide is widening. A new study shows just 20% of firms capture 74% of AI's economic value. Why? They aren't just deploying tools; they are redesigning workflows and building hybrid human-agent teams. Are you in the 20% or...
Rain‑Proof Vision Robotaxis Mark Autonomous Vehicle Inflection Point
Pure vision Robotaxis operating safely in the rain, with no driver or safety monitor. The market is still sleeping on what an insane technical achievement this is. It will be an inflection point in the deployment of autonomous vehicles.
Exploring AI and Infrastructure at Google Next
Off to Google Next to spend a few days digging into all things AI and AI infra. https://t.co/EJhpRAF6nX
Huge Context Window, yet AI Still Misses Basic Numbers
Opus 4.7 context window 1 million. But still gets basic numbers wrong... Makes me wonder if basics get missed how great is AI at this subsidized price point. And what will it take for true confidence? https://t.co/ChNTwn0pqT
Today's Top Marketing AI News Highlights
Top #Marketing #AI News Stories today: https://t.co/Vz4gBaefBQ Top stories from @axios @FastCompany @Adweek @WSJ @kingsmarketDXB @databricks @martechismktg @AdobeForBiz @nvidia and more. https://t.co/61P09Bwnfn
Artists Back AI Amid Hidden Funding to Repair Image
What's behind Diplo (and other artists) pushing so hard for AI? We need to talk about the quiet investments and bags creatives are getting to fix AI's publicity problem. https://t.co/NU4ssTuxLP https://t.co/kkEE5KHyyU