Oracle Soars As AI Enabler; Publishers Battle Back; Humanoids Get Less 'Oid'

Oracle Soars As AI Enabler; Publishers Battle Back; Humanoids Get Less 'Oid'

💰 Oracle’s AI Jackpot: Ellison Leaps To #1 As “Picks & Shovels” Pay Off

Oracle says it has $455B in future AI infrastructure contracts, including a $300B multiyear compute deal with OpenAI (roughly $60B/yr starting 2027).

The disclosure sent Oracle up ~40% in a day, and Larry Ellison reportedly vaulted past Elon Musk as the world’s richest person. Oracle projects cloud infrastructure revenue from $18B → $144B within five years, most already locked by signed agreements.

Bigger picture: AI’s wealth cascade is increasingly favoring infrastructure landlords - chips (Nvidia), power, data centers and now the capacity brokers who parcel it out. With hyperscalers racing to secure compute (see projects like “Stargate”), Oracle is positioning itself as a long-term toll collector on the AI highway... Not a competitor but an enabler.

Read more: The Rundown AI

🚀 Replit Triples Valuation And Unveils Agent 3

Replit raised $250M at a $3B valuation and launched Agent 3, its “most autonomous agent yet,” which can build and test applications for ~200 minutes without human oversight (10× Agent 2).

Agent 3 can spin up other agents, generate Slack/email automations, and auto-test web apps via simulated user sessions. Replit claims revenue climbed from $3M → $150M in a year - rocket-fuel growth for the agentic coding sector next to Cognition, Lovable, Cursor and others.

If the 200-minute figure holds up in the wild, we’re entering the shift from autocomplete to autonomy phase for dev work: less “co-pilot,” more “contractor who ships v1.” Expect tougher questions around test coverage, code provenance, and the new definition of “done.”

Read more: The Rundown AI


🌐 Web Publishers Band Together On AI Licensing Standard

A coalition including Reddit, Yahoo and Medium launched Real Simple Licensing (RSL) - a protocol layered on robots.txt that lets publishers advertise payment options for AI crawling (from free to pay-per-crawl). A nonprofit RSL Collective would negotiate rates, borrowing the ASCAP/BMI playbook from music royalties. RSS co-creator Eckart Walther is leading technical development, and Reddit’s Steve Huffman endorsed the effort despite Reddit’s existing big-ticket licensing.

The promise: a clearer market for AI training rights; the challenge: universal adoption, enforcement, and pricing power. If major sources converge on RSL, model builders could finally get predictable licensing pipelines - otherwise it’s just another flag in a very crowded standards field.

Read more: The Rundown AI


🧠 Musk v Wikipedia: Grok’s “Truth Engine” Takes Aim At The Web’s Commons

Elon Musk says Grok will scan sources like Wikipedia, flag falsehoods, patch half-truths, and insert missing context - possibly crystallizing into a rival “Grokipedia.”

This escalates a long-running feud with Wikipedia’s governance model. The philosophical split is sharp: Wikipedia’s messy, transparent consensus vs. a centralized model-as-arbiter that learns its own corrections.

If Grok’s “truth layer” scales, power tilts from distributed editorship to a single AI oracle - raising alarms for researchers and regulators about epistemic capture and accountability. Even if intended as an overlay, it will be judged as a replacement by many readers who just want one answer.

Read more: AI Secret


🤖 Ant Group Enters The Humanoid Wars With “Robbyant R1”

Ant Group introduced Robbyant R1, a humanoid that can guide tours, sort meds, consult on healthcare and cook basics. Instead of competing on brawn, Ant is outsourcing hardware and focusing on its in-house BaiLing large model ... a cognitive OS to make robots useful companions rather than showroom mascots.

This reframes humanoids as AI distribution channels tied to finance, healthcare and daily services. With the U.S.-China robotics race heating up (Optimus, Unitree, Boston Dynamics), the battleground shifts from factory agility to ecosystem lock-in - who owns the robot’s mind and marketplace.

Read more: AI Secret


🧩 ClickUp Bets Big On “Ambient AI”

ClickUp crossed $300M ARR and is rolling out a fleet of features - AI Agents (task automation), Autonomous Projects (self-moving roadmaps), AI Meetings (scheduling/notes), enterprise AI Search & Ask and an AI Creator for docs/visuals/tasks.

CEO Zeb Evans frames this as the shift from add-on tools to ambient AI that quietly reshapes work around each person.

If orchestration becomes the interface, point-solution SaaS gets squeezed. The winners will be platforms that invisibly keep teams in flow - less dashboard spelunking, fewer tickets to herd, more “wake up to progress.”

Read more: AI Secret


🧰 How-To Tricks: Pull Reddit Insights With Gumloop

Here's how to use Gumloop’s Reddit Insights Generator to harvest posts & comments and auto-cluster them into themes and sentiment for quick market research.

You set subreddit, topic, timeframe then let the pipeline compile examples and patterns (feature complaints, why users switch tools, how launches land). Schedule it as a loop to keep a fresh pulse on your niche.

It’s scrappy, but the dashboards are surprisingly actionable - good for PMs and founders who would rather build than doom-scroll.

Read more: AI Secret


🛠 Tools & Launches

🖼 Seedream 4.0

ByteDance’s unified image generation + editing model aims for 4K quality, complex prompt handling and faster inference - positioned against top-tier creative models.

🧠 ERNIE X1.1

Baidu’s upgraded reasoning model reports strong benchmark gains with reduced hallucinations, part of the intensifying China - U.S. model race on reliability.

🎥 Veo 3 (Price Cut, Vertical HD)

Google slashed costs ~50%, added 1080p and vertical formats, making Veo 3 friendlier for mobile-first creators and ad teams.

🧑‍💻 Latitude 2.0

A platform to create/connect/run production AI agents from a single prompt; pitched as ops-grade, not demo-ware.

🛍 Genstore.ai

Spin up an AI-run storefront in minutes - agents help with setup, design, and marketing automation.

🔬 Incredible (Agentic Models Preview)

Research preview of agentic AI models that can take 1,000+ actions and process large datasets - ambitious claims in the autonomy race.

🏗 Mocha

All-in-one app builder with backend/auth/payments/hosting/db included and built-in AI features; one-click deploy from prototype to production. Don't they all claim that now?)

⭐ VoiceType

Voice-first writing system that claims 9× faster drafting; now a recurring favorite in tool round-ups.

🎓 Resea AI

An academic research agent stitching lit review → data → analysis into one workflow (accuracy and sourcing are key).

Read more: AI Secret, The Rundown AI


⚡ Quick Hits

  • Sam’s Club rolls out floor-level GenAI for managers - an early model for in-store decision support (Walmart is also embracing AI to train employees)
  • CuspAI raises $100M for materials-discovery “search.”
  • Adobe ships its first AI agents inside Experience Platform.
  • “Dead internet” fears resurface as bot traffic climbs and AI slop spreads.

Today’s Sources:

The Internt
AI Secret
The Rundown AI

Subscribe to AI Slop

Sign up now to get access to the library of members-only issues.
Jamie Larson
Subscribe